Update aosp/master LLVM for rebase to r222494.

Change-Id: Ic787f5e0124df789bd26f3f24680f45e678eef2d

13 files changed, 1918 insertions, 63 deletions

diff --git a/lib/ProfileData/Android.mk b/lib/ProfileData/Android.mk
index f4b3fa9..1e1d5f2 100644
--- a/lib/ProfileData/Android.mk
+++ b/lib/ProfileData/Android.mk
@@ -1,9 +1,15 @@
 LOCAL_PATH:= $(call my-dir)
 
 profiledata_SRC_FILES := \
-  InstrProf.cpp          \
-  InstrProfReader.cpp    \
-  InstrProfWriter.cpp    \
+  CoverageMapping.cpp \
+  CoverageMappingReader.cpp \
+  CoverageMappingWriter.cpp \
+  InstrProf.cpp \
+  InstrProfReader.cpp \
+  InstrProfWriter.cpp \
+  SampleProf.cpp \
+  SampleProfReader.cpp \
+  SampleProfWriter.cpp
 
 # For the host
 # =====================================================
diff --git a/lib/ProfileData/CMakeLists.txt b/lib/ProfileData/CMakeLists.txt
index aefb16c..b9d472d 100644
--- a/lib/ProfileData/CMakeLists.txt
+++ b/lib/ProfileData/CMakeLists.txt
@@ -2,4 +2,10 @@ add_llvm_library(LLVMProfileData
   InstrProf.cpp
   InstrProfReader.cpp
   InstrProfWriter.cpp
+  CoverageMapping.cpp
+  CoverageMappingWriter.cpp
+  CoverageMappingReader.cpp
+  SampleProf.cpp
+  SampleProfReader.cpp
+  SampleProfWriter.cpp
   )
diff --git a/lib/ProfileData/CoverageMapping.cpp b/lib/ProfileData/CoverageMapping.cpp
new file mode 100644
index 0000000..0ccebc2
--- /dev/null
+++ b/lib/ProfileData/CoverageMapping.cpp
@@ -0,0 +1,475 @@
+//=-- CoverageMapping.cpp - Code coverage mapping support ---------*- C++ -*-=//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file contains support for clang's and llvm's instrumentation based
+// code coverage.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/ProfileData/CoverageMapping.h"
+
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/Optional.h"
+#include "llvm/ADT/SmallSet.h"
+#include "llvm/ProfileData/CoverageMappingReader.h"
+#include "llvm/ProfileData/InstrProfReader.h"
+#include "llvm/Support/Debug.h"
+#include "llvm/Support/ErrorHandling.h"
+
+using namespace llvm;
+using namespace coverage;
+
+#define DEBUG_TYPE "coverage-mapping"
+
+Counter CounterExpressionBuilder::get(const CounterExpression &E) {
+  auto It = ExpressionIndices.find(E);
+  if (It != ExpressionIndices.end())
+    return Counter::getExpression(It->second);
+  unsigned I = Expressions.size();
+  Expressions.push_back(E);
+  ExpressionIndices[E] = I;
+  return Counter::getExpression(I);
+}
+
+void CounterExpressionBuilder::extractTerms(
+    Counter C, int Sign, SmallVectorImpl<std::pair<unsigned, int>> &Terms) {
+  switch (C.getKind()) {
+  case Counter::Zero:
+    break;
+  case Counter::CounterValueReference:
+    Terms.push_back(std::make_pair(C.getCounterID(), Sign));
+    break;
+  case Counter::Expression:
+    const auto &E = Expressions[C.getExpressionID()];
+    extractTerms(E.LHS, Sign, Terms);
+    extractTerms(E.RHS, E.Kind == CounterExpression::Subtract ? -Sign : Sign,
+                 Terms);
+    break;
+  }
+}
+
+Counter CounterExpressionBuilder::simplify(Counter ExpressionTree) {
+  // Gather constant terms.
+  llvm::SmallVector<std::pair<unsigned, int>, 32> Terms;
+  extractTerms(ExpressionTree, +1, Terms);
+
+  // If there are no terms, this is just a zero. The algorithm below assumes at
+  // least one term.
+  if (Terms.size() == 0)
+    return Counter::getZero();
+
+  // Group the terms by counter ID.
+  std::sort(Terms.begin(), Terms.end(),
+            [](const std::pair<unsigned, int> &LHS,
+               const std::pair<unsigned, int> &RHS) {
+    return LHS.first < RHS.first;
+  });
+
+  // Combine terms by counter ID to eliminate counters that sum to zero.
+  auto Prev = Terms.begin();
+  for (auto I = Prev + 1, E = Terms.end(); I != E; ++I) {
+    if (I->first == Prev->first) {
+      Prev->second += I->second;
+      continue;
+    }
+    ++Prev;
+    *Prev = *I;
+  }
+  Terms.erase(++Prev, Terms.end());
+
+  Counter C;
+  // Create additions. We do this before subtractions to avoid constructs like
+  // ((0 - X) + Y), as opposed to (Y - X).
+  for (auto Term : Terms) {
+    if (Term.second <= 0)
+      continue;
+    for (int I = 0; I < Term.second; ++I)
+      if (C.isZero())
+        C = Counter::getCounter(Term.first);
+      else
+        C = get(CounterExpression(CounterExpression::Add, C,
+                                  Counter::getCounter(Term.first)));
+  }
+
+  // Create subtractions.
+  for (auto Term : Terms) {
+    if (Term.second >= 0)
+      continue;
+    for (int I = 0; I < -Term.second; ++I)
+      C = get(CounterExpression(CounterExpression::Subtract, C,
+                                Counter::getCounter(Term.first)));
+  }
+  return C;
+}
+
+Counter CounterExpressionBuilder::add(Counter LHS, Counter RHS) {
+  return simplify(get(CounterExpression(CounterExpression::Add, LHS, RHS)));
+}
+
+Counter CounterExpressionBuilder::subtract(Counter LHS, Counter RHS) {
+  return simplify(
+      get(CounterExpression(CounterExpression::Subtract, LHS, RHS)));
+}
+
+void CounterMappingContext::dump(const Counter &C,
+                                 llvm::raw_ostream &OS) const {
+  switch (C.getKind()) {
+  case Counter::Zero:
+    OS << '0';
+    return;
+  case Counter::CounterValueReference:
+    OS << '#' << C.getCounterID();
+    break;
+  case Counter::Expression: {
+    if (C.getExpressionID() >= Expressions.size())
+      return;
+    const auto &E = Expressions[C.getExpressionID()];
+    OS << '(';
+    dump(E.LHS, OS);
+    OS << (E.Kind == CounterExpression::Subtract ? " - " : " + ");
+    dump(E.RHS, OS);
+    OS << ')';
+    break;
+  }
+  }
+  if (CounterValues.empty())
+    return;
+  ErrorOr<int64_t> Value = evaluate(C);
+  if (!Value)
+    return;
+  OS << '[' << *Value << ']';
+}
+
+ErrorOr<int64_t> CounterMappingContext::evaluate(const Counter &C) const {
+  switch (C.getKind()) {
+  case Counter::Zero:
+    return 0;
+  case Counter::CounterValueReference:
+    if (C.getCounterID() >= CounterValues.size())
+      return std::make_error_code(std::errc::argument_out_of_domain);
+    return CounterValues[C.getCounterID()];
+  case Counter::Expression: {
+    if (C.getExpressionID() >= Expressions.size())
+      return std::make_error_code(std::errc::argument_out_of_domain);
+    const auto &E = Expressions[C.getExpressionID()];
+    ErrorOr<int64_t> LHS = evaluate(E.LHS);
+    if (!LHS)
+      return LHS;
+    ErrorOr<int64_t> RHS = evaluate(E.RHS);
+    if (!RHS)
+      return RHS;
+    return E.Kind == CounterExpression::Subtract ? *LHS - *RHS : *LHS + *RHS;
+  }
+  }
+  llvm_unreachable("Unhandled CounterKind");
+}
+
+void FunctionRecordIterator::skipOtherFiles() {
+  while (Current != Records.end() && !Filename.empty() &&
+         Filename != Current->Filenames[0])
+    ++Current;
+  if (Current == Records.end())
+    *this = FunctionRecordIterator();
+}
+
+ErrorOr<std::unique_ptr<CoverageMapping>>
+CoverageMapping::load(ObjectFileCoverageMappingReader &CoverageReader,
+                      IndexedInstrProfReader &ProfileReader) {
+  auto Coverage = std::unique_ptr<CoverageMapping>(new CoverageMapping());
+
+  std::vector<uint64_t> Counts;
+  for (const auto &Record : CoverageReader) {
+    Counts.clear();
+    if (std::error_code EC = ProfileReader.getFunctionCounts(
+            Record.FunctionName, Record.FunctionHash, Counts)) {
+      if (EC != instrprof_error::hash_mismatch &&
+          EC != instrprof_error::unknown_function)
+        return EC;
+      Coverage->MismatchedFunctionCount++;
+      continue;
+    }
+
+    assert(Counts.size() != 0 && "Function's counts are empty");
+    FunctionRecord Function(Record.FunctionName, Record.Filenames,
+                            Counts.front());
+    CounterMappingContext Ctx(Record.Expressions, Counts);
+    for (const auto &Region : Record.MappingRegions) {
+      ErrorOr<int64_t> ExecutionCount = Ctx.evaluate(Region.Count);
+      if (!ExecutionCount)
+        break;
+      Function.CountedRegions.push_back(CountedRegion(Region, *ExecutionCount));
+    }
+    if (Function.CountedRegions.size() != Record.MappingRegions.size()) {
+      Coverage->MismatchedFunctionCount++;
+      continue;
+    }
+
+    Coverage->Functions.push_back(std::move(Function));
+  }
+
+  return std::move(Coverage);
+}
+
+ErrorOr<std::unique_ptr<CoverageMapping>>
+CoverageMapping::load(StringRef ObjectFilename, StringRef ProfileFilename) {
+  auto CounterMappingBuff = MemoryBuffer::getFileOrSTDIN(ObjectFilename);
+  if (auto EC = CounterMappingBuff.getError())
+    return EC;
+  ObjectFileCoverageMappingReader CoverageReader(CounterMappingBuff.get());
+  if (auto EC = CoverageReader.readHeader())
+    return EC;
+  std::unique_ptr<IndexedInstrProfReader> ProfileReader;
+  if (auto EC = IndexedInstrProfReader::create(ProfileFilename, ProfileReader))
+    return EC;
+  return load(CoverageReader, *ProfileReader);
+}
+
+namespace {
+/// \brief Distributes functions into instantiation sets.
+///
+/// An instantiation set is a collection of functions that have the same source
+/// code, ie, template functions specializations.
+class FunctionInstantiationSetCollector {
+  typedef DenseMap<std::pair<unsigned, unsigned>,
+                   std::vector<const FunctionRecord *>> MapT;
+  MapT InstantiatedFunctions;
+
+public:
+  void insert(const FunctionRecord &Function, unsigned FileID) {
+    auto I = Function.CountedRegions.begin(), E = Function.CountedRegions.end();
+    while (I != E && I->FileID != FileID)
+      ++I;
+    assert(I != E && "function does not cover the given file");
+    auto &Functions = InstantiatedFunctions[I->startLoc()];
+    Functions.push_back(&Function);
+  }
+
+  MapT::iterator begin() { return InstantiatedFunctions.begin(); }
+
+  MapT::iterator end() { return InstantiatedFunctions.end(); }
+};
+
+class SegmentBuilder {
+  std::vector<CoverageSegment> Segments;
+  SmallVector<const CountedRegion *, 8> ActiveRegions;
+
+  /// Start a segment with no count specified.
+  void startSegment(unsigned Line, unsigned Col) {
+    DEBUG(dbgs() << "Top level segment at " << Line << ":" << Col << "\n");
+    Segments.emplace_back(Line, Col, /*IsRegionEntry=*/false);
+  }
+
+  /// Start a segment with the given Region's count.
+  void startSegment(unsigned Line, unsigned Col, bool IsRegionEntry,
+                    const CountedRegion &Region) {
+    if (Segments.empty())
+      Segments.emplace_back(Line, Col, IsRegionEntry);
+    CoverageSegment S = Segments.back();
+    // Avoid creating empty regions.
+    if (S.Line != Line || S.Col != Col) {
+      Segments.emplace_back(Line, Col, IsRegionEntry);
+      S = Segments.back();
+    }
+    DEBUG(dbgs() << "Segment at " << Line << ":" << Col);
+    // Set this region's count.
+    if (Region.Kind != coverage::CounterMappingRegion::SkippedRegion) {
+      DEBUG(dbgs() << " with count " << Region.ExecutionCount);
+      Segments.back().setCount(Region.ExecutionCount);
+    }
+    DEBUG(dbgs() << "\n");
+  }
+
+  /// Start a segment for the given region.
+  void startSegment(const CountedRegion &Region) {
+    startSegment(Region.LineStart, Region.ColumnStart, true, Region);
+  }
+
+  /// Pop the top region off of the active stack, starting a new segment with
+  /// the containing Region's count.
+  void popRegion() {
+    const CountedRegion *Active = ActiveRegions.back();
+    unsigned Line = Active->LineEnd, Col = Active->ColumnEnd;
+    ActiveRegions.pop_back();
+    if (ActiveRegions.empty())
+      startSegment(Line, Col);
+    else
+      startSegment(Line, Col, false, *ActiveRegions.back());
+  }
+
+public:
+  /// Build a list of CoverageSegments from a sorted list of Regions.
+  std::vector<CoverageSegment> buildSegments(ArrayRef<CountedRegion> Regions) {
+    for (const auto &Region : Regions) {
+      // Pop any regions that end before this one starts.
+      while (!ActiveRegions.empty() &&
+             ActiveRegions.back()->endLoc() <= Region.startLoc())
+        popRegion();
+      if (Segments.size() && Segments.back().Line == Region.LineStart &&
+          Segments.back().Col == Region.ColumnStart) {
+        if (Region.Kind != coverage::CounterMappingRegion::SkippedRegion)
+          Segments.back().addCount(Region.ExecutionCount);
+      } else {
+        // Add this region to the stack.
+        ActiveRegions.push_back(&Region);
+        startSegment(Region);
+      }
+    }
+    // Pop any regions that are left in the stack.
+    while (!ActiveRegions.empty())
+      popRegion();
+    return Segments;
+  }
+};
+}
+
+std::vector<StringRef> CoverageMapping::getUniqueSourceFiles() const {
+  std::vector<StringRef> Filenames;
+  for (const auto &Function : getCoveredFunctions())
+    for (const auto &Filename : Function.Filenames)
+      Filenames.push_back(Filename);
+  std::sort(Filenames.begin(), Filenames.end());
+  auto Last = std::unique(Filenames.begin(), Filenames.end());
+  Filenames.erase(Last, Filenames.end());
+  return Filenames;
+}
+
+static Optional<unsigned> findMainViewFileID(StringRef SourceFile,
+                                             const FunctionRecord &Function) {
+  llvm::SmallVector<bool, 8> IsExpandedFile(Function.Filenames.size(), false);
+  llvm::SmallVector<bool, 8> FilenameEquivalence(Function.Filenames.size(),
+                                                 false);
+  for (unsigned I = 0, E = Function.Filenames.size(); I < E; ++I)
+    if (SourceFile == Function.Filenames[I])
+      FilenameEquivalence[I] = true;
+  for (const auto &CR : Function.CountedRegions)
+    if (CR.Kind == CounterMappingRegion::ExpansionRegion &&
+        FilenameEquivalence[CR.FileID])
+      IsExpandedFile[CR.ExpandedFileID] = true;
+  for (unsigned I = 0, E = Function.Filenames.size(); I < E; ++I)
+    if (FilenameEquivalence[I] && !IsExpandedFile[I])
+      return I;
+  return None;
+}
+
+static Optional<unsigned> findMainViewFileID(const FunctionRecord &Function) {
+  llvm::SmallVector<bool, 8> IsExpandedFile(Function.Filenames.size(), false);
+  for (const auto &CR : Function.CountedRegions)
+    if (CR.Kind == CounterMappingRegion::ExpansionRegion)
+      IsExpandedFile[CR.ExpandedFileID] = true;
+  for (unsigned I = 0, E = Function.Filenames.size(); I < E; ++I)
+    if (!IsExpandedFile[I])
+      return I;
+  return None;
+}
+
+static SmallSet<unsigned, 8> gatherFileIDs(StringRef SourceFile,
+                                           const FunctionRecord &Function) {
+  SmallSet<unsigned, 8> IDs;
+  for (unsigned I = 0, E = Function.Filenames.size(); I < E; ++I)
+    if (SourceFile == Function.Filenames[I])
+      IDs.insert(I);
+  return IDs;
+}
+
+/// Sort a nested sequence of regions from a single file.
+template <class It> static void sortNestedRegions(It First, It Last) {
+  std::sort(First, Last,
+            [](const CountedRegion &LHS, const CountedRegion &RHS) {
+    if (LHS.startLoc() == RHS.startLoc())
+      // When LHS completely contains RHS, we sort LHS first.
+      return RHS.endLoc() < LHS.endLoc();
+    return LHS.startLoc() < RHS.startLoc();
+  });
+}
+
+static bool isExpansion(const CountedRegion &R, unsigned FileID) {
+  return R.Kind == CounterMappingRegion::ExpansionRegion && R.FileID == FileID;
+}
+
+CoverageData CoverageMapping::getCoverageForFile(StringRef Filename) {
+  CoverageData FileCoverage(Filename);
+  std::vector<coverage::CountedRegion> Regions;
+
+  for (const auto &Function : Functions) {
+    auto MainFileID = findMainViewFileID(Filename, Function);
+    if (!MainFileID)
+      continue;
+    auto FileIDs = gatherFileIDs(Filename, Function);
+    for (const auto &CR : Function.CountedRegions)
+      if (FileIDs.count(CR.FileID)) {
+        Regions.push_back(CR);
+        if (isExpansion(CR, *MainFileID))
+          FileCoverage.Expansions.emplace_back(CR, Function);
+      }
+  }
+
+  sortNestedRegions(Regions.begin(), Regions.end());
+  FileCoverage.Segments = SegmentBuilder().buildSegments(Regions);
+
+  return FileCoverage;
+}
+
+std::vector<const FunctionRecord *>
+CoverageMapping::getInstantiations(StringRef Filename) {
+  FunctionInstantiationSetCollector InstantiationSetCollector;
+  for (const auto &Function : Functions) {
+    auto MainFileID = findMainViewFileID(Filename, Function);
+    if (!MainFileID)
+      continue;
+    InstantiationSetCollector.insert(Function, *MainFileID);
+  }
+
+  std::vector<const FunctionRecord *> Result;
+  for (const auto &InstantiationSet : InstantiationSetCollector) {
+    if (InstantiationSet.second.size() < 2)
+      continue;
+    for (auto Function : InstantiationSet.second)
+      Result.push_back(Function);
+  }
+  return Result;
+}
+
+CoverageData
+CoverageMapping::getCoverageForFunction(const FunctionRecord &Function) {
+  auto MainFileID = findMainViewFileID(Function);
+  if (!MainFileID)
+    return CoverageData();
+
+  CoverageData FunctionCoverage(Function.Filenames[*MainFileID]);
+  std::vector<coverage::CountedRegion> Regions;
+  for (const auto &CR : Function.CountedRegions)
+    if (CR.FileID == *MainFileID) {
+      Regions.push_back(CR);
+      if (isExpansion(CR, *MainFileID))
+        FunctionCoverage.Expansions.emplace_back(CR, Function);
+    }
+
+  sortNestedRegions(Regions.begin(), Regions.end());
+  FunctionCoverage.Segments = SegmentBuilder().buildSegments(Regions);
+
+  return FunctionCoverage;
+}
+
+CoverageData
+CoverageMapping::getCoverageForExpansion(const ExpansionRecord &Expansion) {
+  CoverageData ExpansionCoverage(
+      Expansion.Function.Filenames[Expansion.FileID]);
+  std::vector<coverage::CountedRegion> Regions;
+  for (const auto &CR : Expansion.Function.CountedRegions)
+    if (CR.FileID == Expansion.FileID) {
+      Regions.push_back(CR);
+      if (isExpansion(CR, Expansion.FileID))
+        ExpansionCoverage.Expansions.emplace_back(CR, Expansion.Function);
+    }
+
+  sortNestedRegions(Regions.begin(), Regions.end());
+  ExpansionCoverage.Segments = SegmentBuilder().buildSegments(Regions);
+
+  return ExpansionCoverage;
+}
diff --git a/lib/ProfileData/CoverageMappingReader.cpp b/lib/ProfileData/CoverageMappingReader.cpp
new file mode 100644
index 0000000..6476d28
--- /dev/null
+++ b/lib/ProfileData/CoverageMappingReader.cpp
@@ -0,0 +1,553 @@
+//=-- CoverageMappingReader.cpp - Code coverage mapping reader ----*- C++ -*-=//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file contains support for reading coverage mapping data for
+// instrumentation based coverage.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/ProfileData/CoverageMappingReader.h"
+#include "llvm/ADT/DenseSet.h"
+#include "llvm/Object/ObjectFile.h"
+#include "llvm/Support/Debug.h"
+#include "llvm/Support/LEB128.h"
+
+using namespace llvm;
+using namespace coverage;
+using namespace object;
+
+#define DEBUG_TYPE "coverage-mapping"
+
+void CoverageMappingIterator::increment() {
+  // Check if all the records were read or if an error occurred while reading
+  // the next record.
+  if (Reader->readNextRecord(Record))
+    *this = CoverageMappingIterator();
+}
+
+std::error_code RawCoverageReader::readULEB128(uint64_t &Result) {
+  if (Data.size() < 1)
+    return error(instrprof_error::truncated);
+  unsigned N = 0;
+  Result = decodeULEB128(reinterpret_cast<const uint8_t *>(Data.data()), &N);
+  if (N > Data.size())
+    return error(instrprof_error::malformed);
+  Data = Data.substr(N);
+  return success();
+}
+
+std::error_code RawCoverageReader::readIntMax(uint64_t &Result,
+                                              uint64_t MaxPlus1) {
+  if (auto Err = readULEB128(Result))
+    return Err;
+  if (Result >= MaxPlus1)
+    return error(instrprof_error::malformed);
+  return success();
+}
+
+std::error_code RawCoverageReader::readSize(uint64_t &Result) {
+  if (auto Err = readULEB128(Result))
+    return Err;
+  // Sanity check the number.
+  if (Result > Data.size())
+    return error(instrprof_error::malformed);
+  return success();
+}
+
+std::error_code RawCoverageReader::readString(StringRef &Result) {
+  uint64_t Length;
+  if (auto Err = readSize(Length))
+    return Err;
+  Result = Data.substr(0, Length);
+  Data = Data.substr(Length);
+  return success();
+}
+
+std::error_code RawCoverageFilenamesReader::read() {
+  uint64_t NumFilenames;
+  if (auto Err = readSize(NumFilenames))
+    return Err;
+  for (size_t I = 0; I < NumFilenames; ++I) {
+    StringRef Filename;
+    if (auto Err = readString(Filename))
+      return Err;
+    Filenames.push_back(Filename);
+  }
+  return success();
+}
+
+std::error_code RawCoverageMappingReader::decodeCounter(unsigned Value,
+                                                        Counter &C) {
+  auto Tag = Value & Counter::EncodingTagMask;
+  switch (Tag) {
+  case Counter::Zero:
+    C = Counter::getZero();
+    return success();
+  case Counter::CounterValueReference:
+    C = Counter::getCounter(Value >> Counter::EncodingTagBits);
+    return success();
+  default:
+    break;
+  }
+  Tag -= Counter::Expression;
+  switch (Tag) {
+  case CounterExpression::Subtract:
+  case CounterExpression::Add: {
+    auto ID = Value >> Counter::EncodingTagBits;
+    if (ID >= Expressions.size())
+      return error(instrprof_error::malformed);
+    Expressions[ID].Kind = CounterExpression::ExprKind(Tag);
+    C = Counter::getExpression(ID);
+    break;
+  }
+  default:
+    return error(instrprof_error::malformed);
+  }
+  return success();
+}
+
+std::error_code RawCoverageMappingReader::readCounter(Counter &C) {
+  uint64_t EncodedCounter;
+  if (auto Err =
+          readIntMax(EncodedCounter, std::numeric_limits<unsigned>::max()))
+    return Err;
+  if (auto Err = decodeCounter(EncodedCounter, C))
+    return Err;
+  return success();
+}
+
+static const unsigned EncodingExpansionRegionBit = 1
+                                                   << Counter::EncodingTagBits;
+
+/// \brief Read the sub-array of regions for the given inferred file id.
+/// \param NumFileIDs the number of file ids that are defined for this
+/// function.
+std::error_code RawCoverageMappingReader::readMappingRegionsSubArray(
+    std::vector<CounterMappingRegion> &MappingRegions, unsigned InferredFileID,
+    size_t NumFileIDs) {
+  uint64_t NumRegions;
+  if (auto Err = readSize(NumRegions))
+    return Err;
+  unsigned LineStart = 0;
+  for (size_t I = 0; I < NumRegions; ++I) {
+    Counter C;
+    CounterMappingRegion::RegionKind Kind = CounterMappingRegion::CodeRegion;
+
+    // Read the combined counter + region kind.
+    uint64_t EncodedCounterAndRegion;
+    if (auto Err = readIntMax(EncodedCounterAndRegion,
+                              std::numeric_limits<unsigned>::max()))
+      return Err;
+    unsigned Tag = EncodedCounterAndRegion & Counter::EncodingTagMask;
+    uint64_t ExpandedFileID = 0;
+    if (Tag != Counter::Zero) {
+      if (auto Err = decodeCounter(EncodedCounterAndRegion, C))
+        return Err;
+    } else {
+      // Is it an expansion region?
+      if (EncodedCounterAndRegion & EncodingExpansionRegionBit) {
+        Kind = CounterMappingRegion::ExpansionRegion;
+        ExpandedFileID = EncodedCounterAndRegion >>
+                         Counter::EncodingCounterTagAndExpansionRegionTagBits;
+        if (ExpandedFileID >= NumFileIDs)
+          return error(instrprof_error::malformed);
+      } else {
+        switch (EncodedCounterAndRegion >>
+                Counter::EncodingCounterTagAndExpansionRegionTagBits) {
+        case CounterMappingRegion::CodeRegion:
+          // Don't do anything when we have a code region with a zero counter.
+          break;
+        case CounterMappingRegion::SkippedRegion:
+          Kind = CounterMappingRegion::SkippedRegion;
+          break;
+        default:
+          return error(instrprof_error::malformed);
+        }
+      }
+    }
+
+    // Read the source range.
+    uint64_t LineStartDelta, CodeBeforeColumnStart, NumLines, ColumnEnd;
+    if (auto Err =
+            readIntMax(LineStartDelta, std::numeric_limits<unsigned>::max()))
+      return Err;
+    if (auto Err = readULEB128(CodeBeforeColumnStart))
+      return Err;
+    bool HasCodeBefore = CodeBeforeColumnStart & 1;
+    uint64_t ColumnStart = CodeBeforeColumnStart >>
+                           CounterMappingRegion::EncodingHasCodeBeforeBits;
+    if (ColumnStart > std::numeric_limits<unsigned>::max())
+      return error(instrprof_error::malformed);
+    if (auto Err = readIntMax(NumLines, std::numeric_limits<unsigned>::max()))
+      return Err;
+    if (auto Err = readIntMax(ColumnEnd, std::numeric_limits<unsigned>::max()))
+      return Err;
+    LineStart += LineStartDelta;
+    // Adjust the column locations for the empty regions that are supposed to
+    // cover whole lines. Those regions should be encoded with the
+    // column range (1 -> std::numeric_limits<unsigned>::max()), but because
+    // the encoded std::numeric_limits<unsigned>::max() is several bytes long,
+    // we set the column range to (0 -> 0) to ensure that the column start and
+    // column end take up one byte each.
+    // The std::numeric_limits<unsigned>::max() is used to represent a column
+    // position at the end of the line without knowing the length of that line.
+    if (ColumnStart == 0 && ColumnEnd == 0) {
+      ColumnStart = 1;
+      ColumnEnd = std::numeric_limits<unsigned>::max();
+    }
+
+    DEBUG({
+      dbgs() << "Counter in file " << InferredFileID << " " << LineStart << ":"
+             << ColumnStart << " -> " << (LineStart + NumLines) << ":"
+             << ColumnEnd << ", ";
+      if (Kind == CounterMappingRegion::ExpansionRegion)
+        dbgs() << "Expands to file " << ExpandedFileID;
+      else
+        CounterMappingContext(Expressions).dump(C, dbgs());
+      dbgs() << "\n";
+    });
+
+    MappingRegions.push_back(CounterMappingRegion(
+        C, InferredFileID, LineStart, ColumnStart, LineStart + NumLines,
+        ColumnEnd, HasCodeBefore, Kind));
+    MappingRegions.back().ExpandedFileID = ExpandedFileID;
+  }
+  return success();
+}
+
+std::error_code RawCoverageMappingReader::read(CoverageMappingRecord &Record) {
+
+  // Read the virtual file mapping.
+  llvm::SmallVector<unsigned, 8> VirtualFileMapping;
+  uint64_t NumFileMappings;
+  if (auto Err = readSize(NumFileMappings))
+    return Err;
+  for (size_t I = 0; I < NumFileMappings; ++I) {
+    uint64_t FilenameIndex;
+    if (auto Err = readIntMax(FilenameIndex, TranslationUnitFilenames.size()))
+      return Err;
+    VirtualFileMapping.push_back(FilenameIndex);
+  }
+
+  // Construct the files using unique filenames and virtual file mapping.
+  for (auto I : VirtualFileMapping) {
+    Filenames.push_back(TranslationUnitFilenames[I]);
+  }
+
+  // Read the expressions.
+  uint64_t NumExpressions;
+  if (auto Err = readSize(NumExpressions))
+    return Err;
+  // Create an array of dummy expressions that get the proper counters
+  // when the expressions are read, and the proper kinds when the counters
+  // are decoded.
+  Expressions.resize(
+      NumExpressions,
+      CounterExpression(CounterExpression::Subtract, Counter(), Counter()));
+  for (size_t I = 0; I < NumExpressions; ++I) {
+    if (auto Err = readCounter(Expressions[I].LHS))
+      return Err;
+    if (auto Err = readCounter(Expressions[I].RHS))
+      return Err;
+  }
+
+  // Read the mapping regions sub-arrays.
+  for (unsigned InferredFileID = 0, S = VirtualFileMapping.size();
+       InferredFileID < S; ++InferredFileID) {
+    if (auto Err = readMappingRegionsSubArray(MappingRegions, InferredFileID,
+                                              VirtualFileMapping.size()))
+      return Err;
+  }
+
+  // Set the counters for the expansion regions.
+  // i.e. Counter of expansion region = counter of the first region
+  // from the expanded file.
+  // Perform multiple passes to correctly propagate the counters through
+  // all the nested expansion regions.
+  SmallVector<CounterMappingRegion *, 8> FileIDExpansionRegionMapping;
+  FileIDExpansionRegionMapping.resize(VirtualFileMapping.size(), nullptr);
+  for (unsigned Pass = 1, S = VirtualFileMapping.size(); Pass < S; ++Pass) {
+    for (auto &R : MappingRegions) {
+      if (R.Kind != CounterMappingRegion::ExpansionRegion)
+        continue;
+      assert(!FileIDExpansionRegionMapping[R.ExpandedFileID]);
+      FileIDExpansionRegionMapping[R.ExpandedFileID] = &R;
+    }
+    for (auto &R : MappingRegions) {
+      if (FileIDExpansionRegionMapping[R.FileID]) {
+        FileIDExpansionRegionMapping[R.FileID]->Count = R.Count;
+        FileIDExpansionRegionMapping[R.FileID] = nullptr;
+      }
+    }
+  }
+
+  Record.FunctionName = FunctionName;
+  Record.Filenames = Filenames;
+  Record.Expressions = Expressions;
+  Record.MappingRegions = MappingRegions;
+  return success();
+}
+
+ObjectFileCoverageMappingReader::ObjectFileCoverageMappingReader(
+    StringRef FileName)
+    : CurrentRecord(0) {
+  auto File = llvm::object::ObjectFile::createObjectFile(FileName);
+  if (!File)
+    error(File.getError());
+  else
+    Object = std::move(File.get());
+}
+
+namespace {
+/// \brief The coverage mapping data for a single function.
+/// It points to the function's name.
+template <typename IntPtrT> struct CoverageMappingFunctionRecord {
+  IntPtrT FunctionNamePtr;
+  uint32_t FunctionNameSize;
+  uint32_t CoverageMappingSize;
+  uint64_t FunctionHash;
+};
+
+/// \brief The coverage mapping data for a single translation unit.
+/// It points to the array of function coverage mapping records and the encoded
+/// filenames array.
+template <typename IntPtrT> struct CoverageMappingTURecord {
+  uint32_t FunctionRecordsSize;
+  uint32_t FilenamesSize;
+  uint32_t CoverageMappingsSize;
+  uint32_t Version;
+};
+
+/// \brief A helper structure to access the data from a section
+/// in an object file.
+struct SectionData {
+  StringRef Data;
+  uint64_t Address;
+
+  std::error_code load(SectionRef &Section) {
+    if (auto Err = Section.getContents(Data))
+      return Err;
+    Address = Section.getAddress();
+    return instrprof_error::success;
+  }
+
+  std::error_code get(uint64_t Pointer, size_t Size, StringRef &Result) {
+    if (Pointer < Address)
+      return instrprof_error::malformed;
+    auto Offset = Pointer - Address;
+    if (Offset + Size > Data.size())
+      return instrprof_error::malformed;
+    Result = Data.substr(Pointer - Address, Size);
+    return instrprof_error::success;
+  }
+};
+}
+
+template <typename T>
+std::error_code readCoverageMappingData(
+    SectionData &ProfileNames, StringRef Data,
+    std::vector<ObjectFileCoverageMappingReader::ProfileMappingRecord> &Records,
+    std::vector<StringRef> &Filenames) {
+  llvm::DenseSet<T> UniqueFunctionMappingData;
+
+  // Read the records in the coverage data section.
+  while (!Data.empty()) {
+    if (Data.size() < sizeof(CoverageMappingTURecord<T>))
+      return instrprof_error::malformed;
+    auto TU = reinterpret_cast<const CoverageMappingTURecord<T> *>(Data.data());
+    Data = Data.substr(sizeof(CoverageMappingTURecord<T>));
+    switch (TU->Version) {
+    case CoverageMappingVersion1:
+      break;
+    default:
+      return instrprof_error::unsupported_version;
+    }
+    auto Version = CoverageMappingVersion(TU->Version);
+
+    // Get the function records.
+    auto FunctionRecords =
+        reinterpret_cast<const CoverageMappingFunctionRecord<T> *>(Data.data());
+    if (Data.size() <
+        sizeof(CoverageMappingFunctionRecord<T>) * TU->FunctionRecordsSize)
+      return instrprof_error::malformed;
+    Data = Data.substr(sizeof(CoverageMappingFunctionRecord<T>) *
+                       TU->FunctionRecordsSize);
+
+    // Get the filenames.
+    if (Data.size() < TU->FilenamesSize)
+      return instrprof_error::malformed;
+    auto RawFilenames = Data.substr(0, TU->FilenamesSize);
+    Data = Data.substr(TU->FilenamesSize);
+    size_t FilenamesBegin = Filenames.size();
+    RawCoverageFilenamesReader Reader(RawFilenames, Filenames);
+    if (auto Err = Reader.read())
+      return Err;
+
+    // Get the coverage mappings.
+    if (Data.size() < TU->CoverageMappingsSize)
+      return instrprof_error::malformed;
+    auto CoverageMappings = Data.substr(0, TU->CoverageMappingsSize);
+    Data = Data.substr(TU->CoverageMappingsSize);
+
+    for (unsigned I = 0; I < TU->FunctionRecordsSize; ++I) {
+      auto &MappingRecord = FunctionRecords[I];
+
+      // Get the coverage mapping.
+      if (CoverageMappings.size() < MappingRecord.CoverageMappingSize)
+        return instrprof_error::malformed;
+      auto Mapping =
+          CoverageMappings.substr(0, MappingRecord.CoverageMappingSize);
+      CoverageMappings =
+          CoverageMappings.substr(MappingRecord.CoverageMappingSize);
+
+      // Ignore this record if we already have a record that points to the same
+      // function name.
+      // This is useful to ignore the redundant records for the functions
+      // with ODR linkage.
+      if (!UniqueFunctionMappingData.insert(MappingRecord.FunctionNamePtr)
+               .second)
+        continue;
+      StringRef FunctionName;
+      if (auto Err =
+              ProfileNames.get(MappingRecord.FunctionNamePtr,
+                               MappingRecord.FunctionNameSize, FunctionName))
+        return Err;
+      Records.push_back(ObjectFileCoverageMappingReader::ProfileMappingRecord(
+          Version, FunctionName, MappingRecord.FunctionHash, Mapping,
+          FilenamesBegin, Filenames.size() - FilenamesBegin));
+    }
+  }
+
+  return instrprof_error::success;
+}
+
+static const char *TestingFormatMagic = "llvmcovmtestdata";
+
+static std::error_code decodeTestingFormat(StringRef Data,
+                                           SectionData &ProfileNames,
+                                           StringRef &CoverageMapping) {
+  Data = Data.substr(StringRef(TestingFormatMagic).size());
+  if (Data.size() < 1)
+    return instrprof_error::truncated;
+  unsigned N = 0;
+  auto ProfileNamesSize =
+      decodeULEB128(reinterpret_cast<const uint8_t *>(Data.data()), &N);
+  if (N > Data.size())
+    return instrprof_error::malformed;
+  Data = Data.substr(N);
+  if (Data.size() < 1)
+    return instrprof_error::truncated;
+  N = 0;
+  ProfileNames.Address =
+      decodeULEB128(reinterpret_cast<const uint8_t *>(Data.data()), &N);
+  if (N > Data.size())
+    return instrprof_error::malformed;
+  Data = Data.substr(N);
+  if (Data.size() < ProfileNamesSize)
+    return instrprof_error::malformed;
+  ProfileNames.Data = Data.substr(0, ProfileNamesSize);
+  CoverageMapping = Data.substr(ProfileNamesSize);
+  return instrprof_error::success;
+}
+
+ObjectFileCoverageMappingReader::ObjectFileCoverageMappingReader(
+    std::unique_ptr<MemoryBuffer> &ObjectBuffer, sys::fs::file_magic Type)
+    : CurrentRecord(0) {
+  if (ObjectBuffer->getBuffer().startswith(TestingFormatMagic)) {
+    // This is a special format used for testing.
+    SectionData ProfileNames;
+    StringRef CoverageMapping;
+    if (auto Err = decodeTestingFormat(ObjectBuffer->getBuffer(), ProfileNames,
+                                       CoverageMapping)) {
+      error(Err);
+      return;
+    }
+    error(readCoverageMappingData<uint64_t>(ProfileNames, CoverageMapping,
+                                            MappingRecords, Filenames));
+    Object = OwningBinary<ObjectFile>(std::unique_ptr<ObjectFile>(),
+                                      std::move(ObjectBuffer));
+    return;
+  }
+
+  auto File = object::ObjectFile::createObjectFile(
+      ObjectBuffer->getMemBufferRef(), Type);
+  if (!File)
+    error(File.getError());
+  else
+    Object = OwningBinary<ObjectFile>(std::move(File.get()),
+                                      std::move(ObjectBuffer));
+}
+
+std::error_code ObjectFileCoverageMappingReader::readHeader() {
+  const ObjectFile *OF = Object.getBinary();
+  if (!OF)
+    return getError();
+  auto BytesInAddress = OF->getBytesInAddress();
+  if (BytesInAddress != 4 && BytesInAddress != 8)
+    return error(instrprof_error::malformed);
+
+  // Look for the sections that we are interested in.
+  int FoundSectionCount = 0;
+  SectionRef ProfileNames, CoverageMapping;
+  for (const auto &Section : OF->sections()) {
+    StringRef Name;
+    if (auto Err = Section.getName(Name))
+      return Err;
+    if (Name == "__llvm_prf_names") {
+      ProfileNames = Section;
+    } else if (Name == "__llvm_covmap") {
+      CoverageMapping = Section;
+    } else
+      continue;
+    ++FoundSectionCount;
+  }
+  if (FoundSectionCount != 2)
+    return error(instrprof_error::bad_header);
+
+  // Get the contents of the given sections.
+  StringRef Data;
+  if (auto Err = CoverageMapping.getContents(Data))
+    return Err;
+  SectionData ProfileNamesData;
+  if (auto Err = ProfileNamesData.load(ProfileNames))
+    return Err;
+
+  // Load the data from the found sections.
+  std::error_code Err;
+  if (BytesInAddress == 4)
+    Err = readCoverageMappingData<uint32_t>(ProfileNamesData, Data,
+                                            MappingRecords, Filenames);
+  else
+    Err = readCoverageMappingData<uint64_t>(ProfileNamesData, Data,
+                                            MappingRecords, Filenames);
+  if (Err)
+    return error(Err);
+
+  return success();
+}
+
+std::error_code
+ObjectFileCoverageMappingReader::readNextRecord(CoverageMappingRecord &Record) {
+  if (CurrentRecord >= MappingRecords.size())
+    return error(instrprof_error::eof);
+
+  FunctionsFilenames.clear();
+  Expressions.clear();
+  MappingRegions.clear();
+  auto &R = MappingRecords[CurrentRecord];
+  RawCoverageMappingReader Reader(
+      R.FunctionName, R.CoverageMapping,
+      makeArrayRef(Filenames.data() + R.FilenamesBegin, R.FilenamesSize),
+      FunctionsFilenames, Expressions, MappingRegions);
+  if (auto Err = Reader.read(Record))
+    return Err;
+  Record.FunctionHash = R.FunctionHash;
+  ++CurrentRecord;
+  return success();
+}
diff --git a/lib/ProfileData/CoverageMappingWriter.cpp b/lib/ProfileData/CoverageMappingWriter.cpp
new file mode 100644
index 0000000..6969c2a
--- /dev/null
+++ b/lib/ProfileData/CoverageMappingWriter.cpp
@@ -0,0 +1,187 @@
+//=-- CoverageMappingWriter.cpp - Code coverage mapping writer -------------=//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file contains support for writing coverage mapping data for
+// instrumentation based coverage.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/ProfileData/CoverageMappingWriter.h"
+#include "llvm/Support/LEB128.h"
+
+using namespace llvm;
+using namespace coverage;
+
+void CoverageFilenamesSectionWriter::write(raw_ostream &OS) {
+  encodeULEB128(Filenames.size(), OS);
+  for (const auto &Filename : Filenames) {
+    encodeULEB128(Filename.size(), OS);
+    OS << Filename;
+  }
+}
+
+namespace {
+/// \brief Gather only the expressions that are used by the mapping
+/// regions in this function.
+class CounterExpressionsMinimizer {
+  ArrayRef<CounterExpression> Expressions;
+  llvm::SmallVector<CounterExpression, 16> UsedExpressions;
+  std::vector<unsigned> AdjustedExpressionIDs;
+
+public:
+  void mark(Counter C) {
+    if (!C.isExpression())
+      return;
+    unsigned ID = C.getExpressionID();
+    AdjustedExpressionIDs[ID] = 1;
+    mark(Expressions[ID].LHS);
+    mark(Expressions[ID].RHS);
+  }
+
+  void gatherUsed(Counter C) {
+    if (!C.isExpression() || !AdjustedExpressionIDs[C.getExpressionID()])
+      return;
+    AdjustedExpressionIDs[C.getExpressionID()] = UsedExpressions.size();
+    const auto &E = Expressions[C.getExpressionID()];
+    UsedExpressions.push_back(E);
+    gatherUsed(E.LHS);
+    gatherUsed(E.RHS);
+  }
+
+  CounterExpressionsMinimizer(ArrayRef<CounterExpression> Expressions,
+                              ArrayRef<CounterMappingRegion> MappingRegions)
+      : Expressions(Expressions) {
+    AdjustedExpressionIDs.resize(Expressions.size(), 0);
+    for (const auto &I : MappingRegions)
+      mark(I.Count);
+    for (const auto &I : MappingRegions)
+      gatherUsed(I.Count);
+  }
+
+  ArrayRef<CounterExpression> getExpressions() const { return UsedExpressions; }
+
+  /// \brief Adjust the given counter to correctly transition from the old
+  /// expression ids to the new expression ids.
+  Counter adjust(Counter C) const {
+    if (C.isExpression())
+      C = Counter::getExpression(AdjustedExpressionIDs[C.getExpressionID()]);
+    return C;
+  }
+};
+}
+
+/// \brief Encode the counter.
+///
+/// The encoding uses the following format:
+/// Low 2 bits - Tag:
+///   Counter::Zero(0) - A Counter with kind Counter::Zero
+///   Counter::CounterValueReference(1) - A counter with kind
+///     Counter::CounterValueReference
+///   Counter::Expression(2) + CounterExpression::Subtract(0) -
+///     A counter with kind Counter::Expression and an expression
+///     with kind CounterExpression::Subtract
+///   Counter::Expression(2) + CounterExpression::Add(1) -
+///     A counter with kind Counter::Expression and an expression
+///     with kind CounterExpression::Add
+/// Remaining bits - Counter/Expression ID.
+static unsigned encodeCounter(ArrayRef<CounterExpression> Expressions,
+                              Counter C) {
+  unsigned Tag = unsigned(C.getKind());
+  if (C.isExpression())
+    Tag += Expressions[C.getExpressionID()].Kind;
+  unsigned ID = C.getCounterID();
+  assert(ID <=
+         (std::numeric_limits<unsigned>::max() >> Counter::EncodingTagBits));
+  return Tag | (ID << Counter::EncodingTagBits);
+}
+
+static void writeCounter(ArrayRef<CounterExpression> Expressions, Counter C,
+                         raw_ostream &OS) {
+  encodeULEB128(encodeCounter(Expressions, C), OS);
+}
+
+void CoverageMappingWriter::write(raw_ostream &OS) {
+  // Sort the regions in an ascending order by the file id and the starting
+  // location.
+  std::sort(MappingRegions.begin(), MappingRegions.end());
+
+  // Write out the fileid -> filename mapping.
+  encodeULEB128(VirtualFileMapping.size(), OS);
+  for (const auto &FileID : VirtualFileMapping)
+    encodeULEB128(FileID, OS);
+
+  // Write out the expressions.
+  CounterExpressionsMinimizer Minimizer(Expressions, MappingRegions);
+  auto MinExpressions = Minimizer.getExpressions();
+  encodeULEB128(MinExpressions.size(), OS);
+  for (const auto &E : MinExpressions) {
+    writeCounter(MinExpressions, Minimizer.adjust(E.LHS), OS);
+    writeCounter(MinExpressions, Minimizer.adjust(E.RHS), OS);
+  }
+
+  // Write out the mapping regions.
+  // Split the regions into subarrays where each region in a
+  // subarray has a fileID which is the index of that subarray.
+  unsigned PrevLineStart = 0;
+  unsigned CurrentFileID = ~0U;
+  for (auto I = MappingRegions.begin(), E = MappingRegions.end(); I != E; ++I) {
+    if (I->FileID != CurrentFileID) {
+      // Ensure that all file ids have at least one mapping region.
+      assert(I->FileID == (CurrentFileID + 1));
+      // Find the number of regions with this file id.
+      unsigned RegionCount = 1;
+      for (auto J = I + 1; J != E && I->FileID == J->FileID; ++J)
+        ++RegionCount;
+      // Start a new region sub-array.
+      encodeULEB128(RegionCount, OS);
+
+      CurrentFileID = I->FileID;
+      PrevLineStart = 0;
+    }
+    Counter Count = Minimizer.adjust(I->Count);
+    switch (I->Kind) {
+    case CounterMappingRegion::CodeRegion:
+      writeCounter(MinExpressions, Count, OS);
+      break;
+    case CounterMappingRegion::ExpansionRegion: {
+      assert(Count.isZero());
+      assert(I->ExpandedFileID <=
+             (std::numeric_limits<unsigned>::max() >>
+              Counter::EncodingCounterTagAndExpansionRegionTagBits));
+      // Mark an expansion region with a set bit that follows the counter tag,
+      // and pack the expanded file id into the remaining bits.
+      unsigned EncodedTagExpandedFileID =
+          (1 << Counter::EncodingTagBits) |
+          (I->ExpandedFileID
+           << Counter::EncodingCounterTagAndExpansionRegionTagBits);
+      encodeULEB128(EncodedTagExpandedFileID, OS);
+      break;
+    }
+    case CounterMappingRegion::SkippedRegion:
+      assert(Count.isZero());
+      encodeULEB128(unsigned(I->Kind)
+                        << Counter::EncodingCounterTagAndExpansionRegionTagBits,
+                    OS);
+      break;
+    }
+    assert(I->LineStart >= PrevLineStart);
+    encodeULEB128(I->LineStart - PrevLineStart, OS);
+    uint64_t CodeBeforeColumnStart =
+        uint64_t(I->HasCodeBefore) |
+        (uint64_t(I->ColumnStart)
+         << CounterMappingRegion::EncodingHasCodeBeforeBits);
+    encodeULEB128(CodeBeforeColumnStart, OS);
+    assert(I->LineEnd >= I->LineStart);
+    encodeULEB128(I->LineEnd - I->LineStart, OS);
+    encodeULEB128(I->ColumnEnd, OS);
+    PrevLineStart = I->LineStart;
+  }
+  // Ensure that all file ids have at least one mapping region.
+  assert(CurrentFileID == (VirtualFileMapping.size() - 1));
+}
diff --git a/lib/ProfileData/InstrProf.cpp b/lib/ProfileData/InstrProf.cpp
index 0121222..900dff9 100644
--- a/lib/ProfileData/InstrProf.cpp
+++ b/lib/ProfileData/InstrProf.cpp
@@ -14,6 +14,7 @@
 
 #include "llvm/ProfileData/InstrProf.h"
 #include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/ManagedStatic.h"
 
 using namespace llvm;
 
@@ -55,7 +56,8 @@ class InstrProfErrorCategoryType : public std::error_category {
 };
 }
 
+static ManagedStatic<InstrProfErrorCategoryType> ErrorCategory;
+
 const std::error_category &llvm::instrprof_category() {
-  static InstrProfErrorCategoryType C;
-  return C;
+  return *ErrorCategory;
 }
diff --git a/lib/ProfileData/InstrProfIndexed.h b/lib/ProfileData/InstrProfIndexed.h
index 7761704..c2bc46c 100644
--- a/lib/ProfileData/InstrProfIndexed.h
+++ b/lib/ProfileData/InstrProfIndexed.h
@@ -11,8 +11,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_PROFILEDATA_INSTRPROF_INDEXED_H_
-#define LLVM_PROFILEDATA_INSTRPROF_INDEXED_H_
+#ifndef LLVM_LIB_PROFILEDATA_INSTRPROFINDEXED_H
+#define LLVM_LIB_PROFILEDATA_INSTRPROFINDEXED_H
 
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/MD5.h"
@@ -46,10 +46,10 @@ static inline uint64_t ComputeHash(HashT Type, StringRef K) {
 }
 
 const uint64_t Magic = 0x8169666f72706cff; // "\xfflprofi\x81"
-const uint64_t Version = 1;
+const uint64_t Version = 2;
 const HashT HashType = HashT::MD5;
 }
 
 } // end namespace llvm
 
-#endif // LLVM_PROFILEDATA_INSTRPROF_INDEXED_H_
+#endif
diff --git a/lib/ProfileData/InstrProfReader.cpp b/lib/ProfileData/InstrProfReader.cpp
index 0b36728..0160a64 100644
--- a/lib/ProfileData/InstrProfReader.cpp
+++ b/lib/ProfileData/InstrProfReader.cpp
@@ -21,32 +21,34 @@
 
 using namespace llvm;
 
-static std::error_code
-setupMemoryBuffer(std::string Path, std::unique_ptr<MemoryBuffer> &Buffer) {
+static ErrorOr<std::unique_ptr<MemoryBuffer>>
+setupMemoryBuffer(std::string Path) {
   ErrorOr<std::unique_ptr<MemoryBuffer>> BufferOrErr =
       MemoryBuffer::getFileOrSTDIN(Path);
   if (std::error_code EC = BufferOrErr.getError())
     return EC;
-  Buffer = std::move(BufferOrErr.get());
+  auto Buffer = std::move(BufferOrErr.get());
 
   // Sanity check the file.
   if (Buffer->getBufferSize() > std::numeric_limits<unsigned>::max())
     return instrprof_error::too_large;
-  return instrprof_error::success;
+  return std::move(Buffer);
 }
 
 static std::error_code initializeReader(InstrProfReader &Reader) {
   return Reader.readHeader();
 }
 
-std::error_code
-InstrProfReader::create(std::string Path,
-                        std::unique_ptr<InstrProfReader> &Result) {
+ErrorOr<std::unique_ptr<InstrProfReader>>
+InstrProfReader::create(std::string Path) {
   // Set up the buffer to read.
-  std::unique_ptr<MemoryBuffer> Buffer;
-  if (std::error_code EC = setupMemoryBuffer(Path, Buffer))
+  auto BufferOrError = setupMemoryBuffer(Path);
+  if (std::error_code EC = BufferOrError.getError())
     return EC;
 
+  auto Buffer = std::move(BufferOrError.get());
+  std::unique_ptr<InstrProfReader> Result;
+
   // Create the reader.
   if (IndexedInstrProfReader::hasFormat(*Buffer))
     Result.reset(new IndexedInstrProfReader(std::move(Buffer)));
@@ -58,16 +60,20 @@ InstrProfReader::create(std::string Path,
     Result.reset(new TextInstrProfReader(std::move(Buffer)));
 
   // Initialize the reader and return the result.
-  return initializeReader(*Result);
+  if (std::error_code EC = initializeReader(*Result))
+    return EC;
+
+  return std::move(Result);
 }
 
 std::error_code IndexedInstrProfReader::create(
     std::string Path, std::unique_ptr<IndexedInstrProfReader> &Result) {
   // Set up the buffer to read.
-  std::unique_ptr<MemoryBuffer> Buffer;
-  if (std::error_code EC = setupMemoryBuffer(Path, Buffer))
+  auto BufferOrError = setupMemoryBuffer(Path);
+  if (std::error_code EC = BufferOrError.getError())
     return EC;
 
+  auto Buffer = std::move(BufferOrError.get());
   // Create the reader.
   if (!IndexedInstrProfReader::hasFormat(*Buffer))
     return instrprof_error::bad_magic;
@@ -83,8 +89,8 @@ void InstrProfIterator::Increment() {
 }
 
 std::error_code TextInstrProfReader::readNextRecord(InstrProfRecord &Record) {
-  // Skip empty lines.
-  while (!Line.is_at_end() && Line->empty())
+  // Skip empty lines and comments.
+  while (!Line.is_at_end() && (Line->empty() || Line->startswith("#")))
     ++Line;
   // If we hit EOF while looking for a name, we're done.
   if (Line.is_at_end())
@@ -190,6 +196,9 @@ RawInstrProfReader<IntPtrT>::readNextHeader(const char *CurrentPos) {
   // garbage at the end of the file.
   if (CurrentPos + sizeof(RawHeader) > End)
     return instrprof_error::malformed;
+  // The writer ensures each profile is padded to start at an aligned address.
+  if (reinterpret_cast<size_t>(CurrentPos) % alignOf<uint64_t>())
+    return instrprof_error::malformed;
   // The magic should have the same byte order as in the previous header.
   uint64_t Magic = *reinterpret_cast<const uint64_t *>(CurrentPos);
   if (Magic != swap(getRawMagic<IntPtrT>()))
@@ -307,8 +316,8 @@ std::error_code IndexedInstrProfReader::readHeader() {
     return error(instrprof_error::bad_magic);
 
   // Read the version.
-  uint64_t Version = endian::readNext<uint64_t, little, unaligned>(Cur);
-  if (Version != IndexedInstrProf::Version)
+  FormatVersion = endian::readNext<uint64_t, little, unaligned>(Cur);
+  if (FormatVersion > IndexedInstrProf::Version)
     return error(instrprof_error::unsupported_version);
 
   // Read the maximal function count.
@@ -331,18 +340,31 @@ std::error_code IndexedInstrProfReader::readHeader() {
 }
 
 std::error_code IndexedInstrProfReader::getFunctionCounts(
-    StringRef FuncName, uint64_t &FuncHash, std::vector<uint64_t> &Counts) {
-  const auto &Iter = Index->find(FuncName);
+    StringRef FuncName, uint64_t FuncHash, std::vector<uint64_t> &Counts) {
+  auto Iter = Index->find(FuncName);
   if (Iter == Index->end())
     return error(instrprof_error::unknown_function);
 
-  // Found it. Make sure it's valid before giving back a result.
-  const InstrProfRecord &Record = *Iter;
-  if (Record.Name.empty())
-    return error(instrprof_error::malformed);
-  FuncHash = Record.Hash;
-  Counts = Record.Counts;
-  return success();
+  // Found it. Look for counters with the right hash.
+  ArrayRef<uint64_t> Data = (*Iter).Data;
+  uint64_t NumCounts;
+  for (uint64_t I = 0, E = Data.size(); I != E; I += NumCounts) {
+    // The function hash comes first.
+    uint64_t FoundHash = Data[I++];
+    // In v1, we have at least one count. Later, we have the number of counts.
+    if (I == E)
+      return error(instrprof_error::malformed);
+    NumCounts = FormatVersion == 1 ? E - I : Data[I++];
+    // If we have more counts than data, this is bogus.
+    if (I + NumCounts > E)
+      return error(instrprof_error::malformed);
+    // Check for a match and fill the vector if there is one.
+    if (FoundHash == FuncHash) {
+      Counts = Data.slice(I, NumCounts);
+      return success();
+    }
+  }
+  return error(instrprof_error::hash_mismatch);
 }
 
 std::error_code
@@ -351,10 +373,30 @@ IndexedInstrProfReader::readNextRecord(InstrProfRecord &Record) {
   if (RecordIterator == Index->data_end())
     return error(instrprof_error::eof);
 
-  // Read the next one.
-  Record = *RecordIterator;
-  ++RecordIterator;
-  if (Record.Name.empty())
+  // Record the current function name.
+  Record.Name = (*RecordIterator).Name;
+
+  ArrayRef<uint64_t> Data = (*RecordIterator).Data;
+  // Valid data starts with a hash and either a count or the number of counts.
+  if (CurrentOffset + 1 > Data.size())
     return error(instrprof_error::malformed);
+  // First we have a function hash.
+  Record.Hash = Data[CurrentOffset++];
+  // In version 1 we knew the number of counters implicitly, but in newer
+  // versions we store the number of counters next.
+  uint64_t NumCounts =
+      FormatVersion == 1 ? Data.size() - CurrentOffset : Data[CurrentOffset++];
+  if (CurrentOffset + NumCounts > Data.size())
+    return error(instrprof_error::malformed);
+  // And finally the counts themselves.
+  Record.Counts = Data.slice(CurrentOffset, NumCounts);
+
+  // If we've exhausted this function's data, increment the record.
+  CurrentOffset += NumCounts;
+  if (CurrentOffset == Data.size()) {
+    ++RecordIterator;
+    CurrentOffset = 0;
+  }
+
   return success();
 }
diff --git a/lib/ProfileData/InstrProfWriter.cpp b/lib/ProfileData/InstrProfWriter.cpp
index e55c299..ad1b876 100644
--- a/lib/ProfileData/InstrProfWriter.cpp
+++ b/lib/ProfileData/InstrProfWriter.cpp
@@ -45,7 +45,9 @@ public:
     offset_type N = K.size();
     LE.write<offset_type>(N);
 
-    offset_type M = (1 + V->Counts.size()) * sizeof(uint64_t);
+    offset_type M = 0;
+    for (const auto &Counts : *V)
+      M += (2 + Counts.second.size()) * sizeof(uint64_t);
     LE.write<offset_type>(M);
 
     return std::make_pair(N, M);
@@ -59,9 +61,13 @@ public:
                        offset_type) {
     using namespace llvm::support;
     endian::Writer<little> LE(Out);
-    LE.write<uint64_t>(V->Hash);
-    for (uint64_t I : V->Counts)
-      LE.write<uint64_t>(I);
+
+    for (const auto &Counts : *V) {
+      LE.write<uint64_t>(Counts.first);
+      LE.write<uint64_t>(Counts.second.size());
+      for (uint64_t I : Counts.second)
+        LE.write<uint64_t>(I);
+    }
   }
 };
 }
@@ -70,41 +76,43 @@ std::error_code
 InstrProfWriter::addFunctionCounts(StringRef FunctionName,
                                    uint64_t FunctionHash,
                                    ArrayRef<uint64_t> Counters) {
-  auto Where = FunctionData.find(FunctionName);
-  if (Where == FunctionData.end()) {
-    // If this is the first time we've seen this function, just add it.
-    auto &Data = FunctionData[FunctionName];
-    Data.Hash = FunctionHash;
-    Data.Counts = Counters;
+  auto &CounterData = FunctionData[FunctionName];
+
+  auto Where = CounterData.find(FunctionHash);
+  if (Where == CounterData.end()) {
+    // We've never seen a function with this name and hash, add it.
+    CounterData[FunctionHash] = Counters;
+    // We keep track of the max function count as we go for simplicity.
+    if (Counters[0] > MaxFunctionCount)
+      MaxFunctionCount = Counters[0];
     return instrprof_error::success;
   }
 
-  auto &Data = Where->getValue();
-  // We can only add to existing functions if they match, so we check the hash
-  // and number of counters.
-  if (Data.Hash != FunctionHash)
-    return instrprof_error::hash_mismatch;
-  if (Data.Counts.size() != Counters.size())
+  // We're updating a function we've seen before.
+  auto &FoundCounters = Where->second;
+  // If the number of counters doesn't match we either have bad data or a hash
+  // collision.
+  if (FoundCounters.size() != Counters.size())
     return instrprof_error::count_mismatch;
-  // These match, add up the counters.
+
   for (size_t I = 0, E = Counters.size(); I < E; ++I) {
-    if (Data.Counts[I] + Counters[I] < Data.Counts[I])
+    if (FoundCounters[I] + Counters[I] < FoundCounters[I])
       return instrprof_error::counter_overflow;
-    Data.Counts[I] += Counters[I];
+    FoundCounters[I] += Counters[I];
   }
+  // We keep track of the max function count as we go for simplicity.
+  if (FoundCounters[0] > MaxFunctionCount)
+    MaxFunctionCount = FoundCounters[0];
+
   return instrprof_error::success;
 }
 
 void InstrProfWriter::write(raw_fd_ostream &OS) {
   OnDiskChainedHashTableGenerator<InstrProfRecordTrait> Generator;
-  uint64_t MaxFunctionCount = 0;
 
   // Populate the hash table generator.
-  for (const auto &I : FunctionData) {
+  for (const auto &I : FunctionData)
     Generator.insert(I.getKey(), &I.getValue());
-    if (I.getValue().Counts[0] > MaxFunctionCount)
-      MaxFunctionCount = I.getValue().Counts[0];
-  }
 
   using namespace llvm::support;
   endian::Writer<little> LE(OS);
diff --git a/lib/ProfileData/LLVMBuild.txt b/lib/ProfileData/LLVMBuild.txt
index 0a8cbe3..a7f471f 100644
--- a/lib/ProfileData/LLVMBuild.txt
+++ b/lib/ProfileData/LLVMBuild.txt
@@ -19,4 +19,4 @@
 type = Library
 name = ProfileData
 parent = Libraries
-required_libraries = Support
+required_libraries = Core Support Object
diff --git a/lib/ProfileData/SampleProf.cpp b/lib/ProfileData/SampleProf.cpp
new file mode 100644
index 0000000..920c48a
--- /dev/null
+++ b/lib/ProfileData/SampleProf.cpp
@@ -0,0 +1,51 @@
+//=-- SampleProf.cpp - Sample profiling format support --------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file contains common definitions used in the reading and writing of
+// sample profile data.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/ProfileData/SampleProf.h"
+#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/ManagedStatic.h"
+
+using namespace llvm;
+
+namespace {
+class SampleProfErrorCategoryType : public std::error_category {
+  const char *name() const LLVM_NOEXCEPT override { return "llvm.sampleprof"; }
+  std::string message(int IE) const override {
+    sampleprof_error E = static_cast<sampleprof_error>(IE);
+    switch (E) {
+    case sampleprof_error::success:
+      return "Success";
+    case sampleprof_error::bad_magic:
+      return "Invalid file format (bad magic)";
+    case sampleprof_error::unsupported_version:
+      return "Unsupported format version";
+    case sampleprof_error::too_large:
+      return "Too much profile data";
+    case sampleprof_error::truncated:
+      return "Truncated profile data";
+    case sampleprof_error::malformed:
+      return "Malformed profile data";
+    case sampleprof_error::unrecognized_format:
+      return "Unrecognized profile encoding format";
+    }
+    llvm_unreachable("A value of sampleprof_error has no message.");
+  }
+};
+}
+
+static ManagedStatic<SampleProfErrorCategoryType> ErrorCategory;
+
+const std::error_category &llvm::sampleprof_category() {
+  return *ErrorCategory;
+}
diff --git a/lib/ProfileData/SampleProfReader.cpp b/lib/ProfileData/SampleProfReader.cpp
new file mode 100644
index 0000000..b39bfd6
--- /dev/null
+++ b/lib/ProfileData/SampleProfReader.cpp
@@ -0,0 +1,399 @@
+//===- SampleProfReader.cpp - Read LLVM sample profile data ---------------===//
+//
+//                      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 class that reads LLVM sample profiles. It
+// supports two file formats: text and binary. The textual representation
+// is useful for debugging and testing purposes. The binary representation
+// is more compact, resulting in smaller file sizes. However, they can
+// both be used interchangeably.
+//
+// NOTE: If you are making changes to the file format, please remember
+//       to document them in the Clang documentation at
+//       tools/clang/docs/UsersManual.rst.
+//
+// Text format
+// -----------
+//
+// Sample profiles are written as ASCII text. The file is divided into
+// sections, which correspond to each of the functions executed at runtime.
+// Each section has the following format
+//
+//     function1:total_samples:total_head_samples
+//     offset1[.discriminator]: number_of_samples [fn1:num fn2:num ... ]
+//     offset2[.discriminator]: number_of_samples [fn3:num fn4:num ... ]
+//     ...
+//     offsetN[.discriminator]: number_of_samples [fn5:num fn6:num ... ]
+//
+// The file may contain blank lines between sections and within a
+// section. However, the spacing within a single line is fixed. Additional
+// spaces will result in an error while reading the file.
+//
+// Function names must be mangled in order for the profile loader to
+// match them in the current translation unit. The two numbers in the
+// function header specify how many total samples were accumulated in the
+// function (first number), and the total number of samples accumulated
+// in the prologue of the function (second number). This head sample
+// count provides an indicator of how frequently the function is invoked.
+//
+// Each sampled line may contain several items. Some are optional (marked
+// below):
+//
+// a. Source line offset. This number represents the line number
+//    in the function where the sample was collected. The line number is
+//    always relative to the line where symbol of the function is
+//    defined. So, if the function has its header at line 280, the offset
+//    13 is at line 293 in the file.
+//
+//    Note that this offset should never be a negative number. This could
+//    happen in cases like macros. The debug machinery will register the
+//    line number at the point of macro expansion. So, if the macro was
+//    expanded in a line before the start of the function, the profile
+//    converter should emit a 0 as the offset (this means that the optimizers
+//    will not be able to associate a meaningful weight to the instructions
+//    in the macro).
+//
+// b. [OPTIONAL] Discriminator. This is used if the sampled program
+//    was compiled with DWARF discriminator support
+//    (http://wiki.dwarfstd.org/index.php?title=Path_Discriminators).
+//    DWARF discriminators are unsigned integer values that allow the
+//    compiler to distinguish between multiple execution paths on the
+//    same source line location.
+//
+//    For example, consider the line of code ``if (cond) foo(); else bar();``.
+//    If the predicate ``cond`` is true 80% of the time, then the edge
+//    into function ``foo`` should be considered to be taken most of the
+//    time. But both calls to ``foo`` and ``bar`` are at the same source
+//    line, so a sample count at that line is not sufficient. The
+//    compiler needs to know which part of that line is taken more
+//    frequently.
+//
+//    This is what discriminators provide. In this case, the calls to
+//    ``foo`` and ``bar`` will be at the same line, but will have
+//    different discriminator values. This allows the compiler to correctly
+//    set edge weights into ``foo`` and ``bar``.
+//
+// c. Number of samples. This is an integer quantity representing the
+//    number of samples collected by the profiler at this source
+//    location.
+//
+// d. [OPTIONAL] Potential call targets and samples. If present, this
+//    line contains a call instruction. This models both direct and
+//    number of samples. For example,
+//
+//      130: 7  foo:3  bar:2  baz:7
+//
+//    The above means that at relative line offset 130 there is a call
+//    instruction that calls one of ``foo()``, ``bar()`` and ``baz()``,
+//    with ``baz()`` being the relatively more frequently called target.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/ProfileData/SampleProfReader.h"
+#include "llvm/Support/Debug.h"
+#include "llvm/Support/ErrorOr.h"
+#include "llvm/Support/LEB128.h"
+#include "llvm/Support/LineIterator.h"
+#include "llvm/Support/MemoryBuffer.h"
+#include "llvm/Support/Regex.h"
+
+using namespace llvm::sampleprof;
+using namespace llvm;
+
+/// \brief Print the samples collected for a function on stream \p OS.
+///
+/// \param OS Stream to emit the output to.
+void FunctionSamples::print(raw_ostream &OS) {
+  OS << TotalSamples << ", " << TotalHeadSamples << ", " << BodySamples.size()
+     << " sampled lines\n";
+  for (const auto &SI : BodySamples) {
+    LineLocation Loc = SI.first;
+    const SampleRecord &Sample = SI.second;
+    OS << "\tline offset: " << Loc.LineOffset
+       << ", discriminator: " << Loc.Discriminator
+       << ", number of samples: " << Sample.getSamples();
+    if (Sample.hasCalls()) {
+      OS << ", calls:";
+      for (const auto &I : Sample.getCallTargets())
+        OS << " " << I.first() << ":" << I.second;
+    }
+    OS << "\n";
+  }
+  OS << "\n";
+}
+
+/// \brief Dump the function profile for \p FName.
+///
+/// \param FName Name of the function to print.
+/// \param OS Stream to emit the output to.
+void SampleProfileReader::dumpFunctionProfile(StringRef FName,
+                                              raw_ostream &OS) {
+  OS << "Function: " << FName << ": ";
+  Profiles[FName].print(OS);
+}
+
+/// \brief Dump all the function profiles found on stream \p OS.
+void SampleProfileReader::dump(raw_ostream &OS) {
+  for (const auto &I : Profiles)
+    dumpFunctionProfile(I.getKey(), OS);
+}
+
+/// \brief Load samples from a text file.
+///
+/// See the documentation at the top of the file for an explanation of
+/// the expected format.
+///
+/// \returns true if the file was loaded successfully, false otherwise.
+std::error_code SampleProfileReaderText::read() {
+  line_iterator LineIt(*Buffer, /*SkipBlanks=*/true, '#');
+
+  // Read the profile of each function. Since each function may be
+  // mentioned more than once, and we are collecting flat profiles,
+  // accumulate samples as we parse them.
+  Regex HeadRE("^([^0-9].*):([0-9]+):([0-9]+)$");
+  Regex LineSampleRE("^([0-9]+)\\.?([0-9]+)?: ([0-9]+)(.*)$");
+  Regex CallSampleRE(" +([^0-9 ][^ ]*):([0-9]+)");
+  while (!LineIt.is_at_eof()) {
+    // Read the header of each function.
+    //
+    // Note that for function identifiers we are actually expecting
+    // mangled names, but we may not always get them. This happens when
+    // the compiler decides not to emit the function (e.g., it was inlined
+    // and removed). In this case, the binary will not have the linkage
+    // name for the function, so the profiler will emit the function's
+    // unmangled name, which may contain characters like ':' and '>' in its
+    // name (member functions, templates, etc).
+    //
+    // The only requirement we place on the identifier, then, is that it
+    // should not begin with a number.
+    SmallVector<StringRef, 4> Matches;
+    if (!HeadRE.match(*LineIt, &Matches)) {
+      reportParseError(LineIt.line_number(),
+                       "Expected 'mangled_name:NUM:NUM', found " + *LineIt);
+      return sampleprof_error::malformed;
+    }
+    assert(Matches.size() == 4);
+    StringRef FName = Matches[1];
+    unsigned NumSamples, NumHeadSamples;
+    Matches[2].getAsInteger(10, NumSamples);
+    Matches[3].getAsInteger(10, NumHeadSamples);
+    Profiles[FName] = FunctionSamples();
+    FunctionSamples &FProfile = Profiles[FName];
+    FProfile.addTotalSamples(NumSamples);
+    FProfile.addHeadSamples(NumHeadSamples);
+    ++LineIt;
+
+    // Now read the body. The body of the function ends when we reach
+    // EOF or when we see the start of the next function.
+    while (!LineIt.is_at_eof() && isdigit((*LineIt)[0])) {
+      if (!LineSampleRE.match(*LineIt, &Matches)) {
+        reportParseError(
+            LineIt.line_number(),
+            "Expected 'NUM[.NUM]: NUM[ mangled_name:NUM]*', found " + *LineIt);
+        return sampleprof_error::malformed;
+      }
+      assert(Matches.size() == 5);
+      unsigned LineOffset, NumSamples, Discriminator = 0;
+      Matches[1].getAsInteger(10, LineOffset);
+      if (Matches[2] != "")
+        Matches[2].getAsInteger(10, Discriminator);
+      Matches[3].getAsInteger(10, NumSamples);
+
+      // If there are function calls in this line, generate a call sample
+      // entry for each call.
+      std::string CallsLine(Matches[4]);
+      while (CallsLine != "") {
+        SmallVector<StringRef, 3> CallSample;
+        if (!CallSampleRE.match(CallsLine, &CallSample)) {
+          reportParseError(LineIt.line_number(),
+                           "Expected 'mangled_name:NUM', found " + CallsLine);
+          return sampleprof_error::malformed;
+        }
+        StringRef CalledFunction = CallSample[1];
+        unsigned CalledFunctionSamples;
+        CallSample[2].getAsInteger(10, CalledFunctionSamples);
+        FProfile.addCalledTargetSamples(LineOffset, Discriminator,
+                                        CalledFunction, CalledFunctionSamples);
+        CallsLine = CallSampleRE.sub("", CallsLine);
+      }
+
+      FProfile.addBodySamples(LineOffset, Discriminator, NumSamples);
+      ++LineIt;
+    }
+  }
+
+  return sampleprof_error::success;
+}
+
+template <typename T> ErrorOr<T> SampleProfileReaderBinary::readNumber() {
+  unsigned NumBytesRead = 0;
+  std::error_code EC;
+  uint64_t Val = decodeULEB128(Data, &NumBytesRead);
+
+  if (Val > std::numeric_limits<T>::max())
+    EC = sampleprof_error::malformed;
+  else if (Data + NumBytesRead > End)
+    EC = sampleprof_error::truncated;
+  else
+    EC = sampleprof_error::success;
+
+  if (EC) {
+    reportParseError(0, EC.message());
+    return EC;
+  }
+
+  Data += NumBytesRead;
+  return static_cast<T>(Val);
+}
+
+ErrorOr<StringRef> SampleProfileReaderBinary::readString() {
+  std::error_code EC;
+  StringRef Str(reinterpret_cast<const char *>(Data));
+  if (Data + Str.size() + 1 > End) {
+    EC = sampleprof_error::truncated;
+    reportParseError(0, EC.message());
+    return EC;
+  }
+
+  Data += Str.size() + 1;
+  return Str;
+}
+
+std::error_code SampleProfileReaderBinary::read() {
+  while (!at_eof()) {
+    auto FName(readString());
+    if (std::error_code EC = FName.getError())
+      return EC;
+
+    Profiles[*FName] = FunctionSamples();
+    FunctionSamples &FProfile = Profiles[*FName];
+
+    auto Val = readNumber<unsigned>();
+    if (std::error_code EC = Val.getError())
+      return EC;
+    FProfile.addTotalSamples(*Val);
+
+    Val = readNumber<unsigned>();
+    if (std::error_code EC = Val.getError())
+      return EC;
+    FProfile.addHeadSamples(*Val);
+
+    // Read the samples in the body.
+    auto NumRecords = readNumber<unsigned>();
+    if (std::error_code EC = NumRecords.getError())
+      return EC;
+    for (unsigned I = 0; I < *NumRecords; ++I) {
+      auto LineOffset = readNumber<uint64_t>();
+      if (std::error_code EC = LineOffset.getError())
+        return EC;
+
+      auto Discriminator = readNumber<uint64_t>();
+      if (std::error_code EC = Discriminator.getError())
+        return EC;
+
+      auto NumSamples = readNumber<uint64_t>();
+      if (std::error_code EC = NumSamples.getError())
+        return EC;
+
+      auto NumCalls = readNumber<unsigned>();
+      if (std::error_code EC = NumCalls.getError())
+        return EC;
+
+      for (unsigned J = 0; J < *NumCalls; ++J) {
+        auto CalledFunction(readString());
+        if (std::error_code EC = CalledFunction.getError())
+          return EC;
+
+        auto CalledFunctionSamples = readNumber<uint64_t>();
+        if (std::error_code EC = CalledFunctionSamples.getError())
+          return EC;
+
+        FProfile.addCalledTargetSamples(*LineOffset, *Discriminator,
+                                        *CalledFunction,
+                                        *CalledFunctionSamples);
+      }
+
+      FProfile.addBodySamples(*LineOffset, *Discriminator, *NumSamples);
+    }
+  }
+
+  return sampleprof_error::success;
+}
+
+std::error_code SampleProfileReaderBinary::readHeader() {
+  Data = reinterpret_cast<const uint8_t *>(Buffer->getBufferStart());
+  End = Data + Buffer->getBufferSize();
+
+  // Read and check the magic identifier.
+  auto Magic = readNumber<uint64_t>();
+  if (std::error_code EC = Magic.getError())
+    return EC;
+  else if (*Magic != SPMagic())
+    return sampleprof_error::bad_magic;
+
+  // Read the version number.
+  auto Version = readNumber<uint64_t>();
+  if (std::error_code EC = Version.getError())
+    return EC;
+  else if (*Version != SPVersion())
+    return sampleprof_error::unsupported_version;
+
+  return sampleprof_error::success;
+}
+
+bool SampleProfileReaderBinary::hasFormat(const MemoryBuffer &Buffer) {
+  const uint8_t *Data =
+      reinterpret_cast<const uint8_t *>(Buffer.getBufferStart());
+  uint64_t Magic = decodeULEB128(Data);
+  return Magic == SPMagic();
+}
+
+/// \brief Prepare a memory buffer for the contents of \p Filename.
+///
+/// \returns an error code indicating the status of the buffer.
+static ErrorOr<std::unique_ptr<MemoryBuffer>>
+setupMemoryBuffer(std::string Filename) {
+  auto BufferOrErr = MemoryBuffer::getFileOrSTDIN(Filename);
+  if (std::error_code EC = BufferOrErr.getError())
+    return EC;
+  auto Buffer = std::move(BufferOrErr.get());
+
+  // Sanity check the file.
+  if (Buffer->getBufferSize() > std::numeric_limits<unsigned>::max())
+    return sampleprof_error::too_large;
+
+  return std::move(Buffer);
+}
+
+/// \brief Create a sample profile reader based on the format of the input file.
+///
+/// \param Filename The file to open.
+///
+/// \param Reader The reader to instantiate according to \p Filename's format.
+///
+/// \param C The LLVM context to use to emit diagnostics.
+///
+/// \returns an error code indicating the status of the created reader.
+ErrorOr<std::unique_ptr<SampleProfileReader>>
+SampleProfileReader::create(StringRef Filename, LLVMContext &C) {
+  auto BufferOrError = setupMemoryBuffer(Filename);
+  if (std::error_code EC = BufferOrError.getError())
+    return EC;
+
+  auto Buffer = std::move(BufferOrError.get());
+  std::unique_ptr<SampleProfileReader> Reader;
+  if (SampleProfileReaderBinary::hasFormat(*Buffer))
+    Reader.reset(new SampleProfileReaderBinary(std::move(Buffer), C));
+  else
+    Reader.reset(new SampleProfileReaderText(std::move(Buffer), C));
+
+  if (std::error_code EC = Reader->readHeader())
+    return EC;
+
+  return std::move(Reader);
+}
diff --git a/lib/ProfileData/SampleProfWriter.cpp b/lib/ProfileData/SampleProfWriter.cpp
new file mode 100644
index 0000000..8525045
--- /dev/null
+++ b/lib/ProfileData/SampleProfWriter.cpp
@@ -0,0 +1,126 @@
+//===- SampleProfWriter.cpp - Write LLVM sample profile data --------------===//
+//
+//                      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 class that writes LLVM sample profiles. It
+// supports two file formats: text and binary. The textual representation
+// is useful for debugging and testing purposes. The binary representation
+// is more compact, resulting in smaller file sizes. However, they can
+// both be used interchangeably.
+//
+// See lib/ProfileData/SampleProfReader.cpp for documentation on each of the
+// supported formats.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/ProfileData/SampleProfWriter.h"
+#include "llvm/Support/Debug.h"
+#include "llvm/Support/ErrorOr.h"
+#include "llvm/Support/MemoryBuffer.h"
+#include "llvm/Support/LEB128.h"
+#include "llvm/Support/LineIterator.h"
+#include "llvm/Support/Regex.h"
+
+using namespace llvm::sampleprof;
+using namespace llvm;
+
+/// \brief Write samples to a text file.
+bool SampleProfileWriterText::write(StringRef FName, const FunctionSamples &S) {
+  if (S.empty())
+    return true;
+
+  OS << FName << ":" << S.getTotalSamples() << ":" << S.getHeadSamples()
+     << "\n";
+
+  for (const auto &I : S.getBodySamples()) {
+    LineLocation Loc = I.first;
+    const SampleRecord &Sample = I.second;
+    if (Loc.Discriminator == 0)
+      OS << Loc.LineOffset << ": ";
+    else
+      OS << Loc.LineOffset << "." << Loc.Discriminator << ": ";
+
+    OS << Sample.getSamples();
+
+    for (const auto &J : Sample.getCallTargets())
+      OS << " " << J.first() << ":" << J.second;
+    OS << "\n";
+  }
+
+  return true;
+}
+
+SampleProfileWriterBinary::SampleProfileWriterBinary(StringRef F,
+                                                     std::error_code &EC)
+    : SampleProfileWriter(F, EC, sys::fs::F_None) {
+  if (EC)
+    return;
+
+  // Write the file header.
+  encodeULEB128(SPMagic(), OS);
+  encodeULEB128(SPVersion(), OS);
+}
+
+/// \brief Write samples to a binary file.
+///
+/// \returns true if the samples were written successfully, false otherwise.
+bool SampleProfileWriterBinary::write(StringRef FName,
+                                      const FunctionSamples &S) {
+  if (S.empty())
+    return true;
+
+  OS << FName;
+  encodeULEB128(0, OS);
+  encodeULEB128(S.getTotalSamples(), OS);
+  encodeULEB128(S.getHeadSamples(), OS);
+  encodeULEB128(S.getBodySamples().size(), OS);
+  for (const auto &I : S.getBodySamples()) {
+    LineLocation Loc = I.first;
+    const SampleRecord &Sample = I.second;
+    encodeULEB128(Loc.LineOffset, OS);
+    encodeULEB128(Loc.Discriminator, OS);
+    encodeULEB128(Sample.getSamples(), OS);
+    encodeULEB128(Sample.getCallTargets().size(), OS);
+    for (const auto &J : Sample.getCallTargets()) {
+      std::string Callee = J.first();
+      unsigned CalleeSamples = J.second;
+      OS << Callee;
+      encodeULEB128(0, OS);
+      encodeULEB128(CalleeSamples, OS);
+    }
+  }
+
+  return true;
+}
+
+/// \brief Create a sample profile writer based on the specified format.
+///
+/// \param Filename The file to create.
+///
+/// \param Writer The writer to instantiate according to the specified format.
+///
+/// \param Format Encoding format for the profile file.
+///
+/// \returns an error code indicating the status of the created writer.
+ErrorOr<std::unique_ptr<SampleProfileWriter>>
+SampleProfileWriter::create(StringRef Filename, SampleProfileFormat Format) {
+  std::error_code EC;
+  std::unique_ptr<SampleProfileWriter> Writer;
+
+  if (Format == SPF_Binary)
+    Writer.reset(new SampleProfileWriterBinary(Filename, EC));
+  else if (Format == SPF_Text)
+    Writer.reset(new SampleProfileWriterText(Filename, EC));
+  else
+    EC = sampleprof_error::unrecognized_format;
+
+  if (EC)
+    return EC;
+
+  return std::move(Writer);
+}