[llvm-profdata] Add merge() to InstrProfRecord
[oota-llvm.git] / lib / ProfileData / InstrProfWriter.cpp
index d4cde2e195de24ccfd6d6b15aeb9fecb8f8a1ec8..465a4ea9e00a4dce0d1d9b753866555e2dd38468 100644 (file)
@@ -13,7 +13,6 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/ProfileData/InstrProfWriter.h"
-#include "InstrProfIndexed.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/Support/EndianStream.h"
 #include "llvm/Support/OnDiskHashTable.h"
 using namespace llvm;
 
 namespace {
+static support::endianness ValueProfDataEndianness = support::little;
+
 class InstrProfRecordTrait {
 public:
   typedef StringRef key_type;
   typedef StringRef key_type_ref;
 
-  typedef const InstrProfWriter::CounterData *const data_type;
-  typedef const InstrProfWriter::CounterData *const data_type_ref;
+  typedef const InstrProfWriter::ProfilingData *const data_type;
+  typedef const InstrProfWriter::ProfilingData *const data_type_ref;
 
   typedef uint64_t hash_value_type;
   typedef uint64_t offset_type;
@@ -45,8 +46,15 @@ public:
     LE.write<offset_type>(N);
 
     offset_type M = 0;
-    for (const auto &Counts : *V)
-      M += (2 + Counts.second.size()) * sizeof(uint64_t);
+    for (const auto &ProfileData : *V) {
+      const InstrProfRecord &ProfRecord = ProfileData.second;
+      M += sizeof(uint64_t); // The function hash
+      M += sizeof(uint64_t); // The size of the Counts vector
+      M += ProfRecord.Counts.size() * sizeof(uint64_t);
+
+      // Value data
+      M += ValueProfData::getSize(ProfileData.second);
+    }
     LE.write<offset_type>(M);
 
     return std::make_pair(N, M);
@@ -60,53 +68,63 @@ public:
                        offset_type) {
     using namespace llvm::support;
     endian::Writer<little> LE(Out);
+    for (const auto &ProfileData : *V) {
+      const InstrProfRecord &ProfRecord = ProfileData.second;
 
-    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>(ProfileData.first); // Function hash
+      LE.write<uint64_t>(ProfRecord.Counts.size());
+      for (uint64_t I : ProfRecord.Counts)
         LE.write<uint64_t>(I);
+
+      // Write value data
+      std::unique_ptr<ValueProfData> VDataPtr =
+          ValueProfData::serializeFrom(ProfileData.second);
+      uint32_t S = VDataPtr->getSize();
+      VDataPtr->swapBytesFromHost(ValueProfDataEndianness);
+      Out.write((const char *)VDataPtr.get(), S);
     }
   }
 };
 }
 
-std::error_code
-InstrProfWriter::addFunctionCounts(StringRef FunctionName,
-                                   uint64_t FunctionHash,
-                                   ArrayRef<uint64_t> Counters) {
-  auto &CounterData = FunctionData[FunctionName];
+// Internal interface for testing purpose only.
+void InstrProfWriter::setValueProfDataEndianness(
+    support::endianness Endianness) {
+  ValueProfDataEndianness = Endianness;
+}
+
+void InstrProfWriter::updateStringTableReferences(InstrProfRecord &I) {
+  I.updateStrings(&StringTable);
+}
 
-  auto Where = CounterData.find(FunctionHash);
-  if (Where == CounterData.end()) {
+std::error_code InstrProfWriter::addRecord(InstrProfRecord &&I) {
+  updateStringTableReferences(I);
+  auto &ProfileDataMap = FunctionData[I.Name];
+
+  bool NewFunc;
+  ProfilingData::iterator Where;
+  std::tie(Where, NewFunc) =
+      ProfileDataMap.insert(std::make_pair(I.Hash, InstrProfRecord()));
+  InstrProfRecord &Dest = Where->second;
+  if (NewFunc) {
     // 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;
+    Dest = std::move(I);
+  } else {
+    // We're updating a function we've seen before.
+    instrprof_error MergeResult = Dest.merge(I);
+    if (MergeResult != instrprof_error::success) {
+      return MergeResult;
+    }
   }
 
-  // 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;
-
-  for (size_t I = 0, E = Counters.size(); I < E; ++I) {
-    if (FoundCounters[I] + Counters[I] < FoundCounters[I])
-      return instrprof_error::counter_overflow;
-    FoundCounters[I] += Counters[I];
-  }
   // We keep track of the max function count as we go for simplicity.
-  if (FoundCounters[0] > MaxFunctionCount)
-    MaxFunctionCount = FoundCounters[0];
+  if (Dest.Counts[0] > MaxFunctionCount)
+    MaxFunctionCount = Dest.Counts[0];
 
   return instrprof_error::success;
 }
 
-void InstrProfWriter::write(raw_fd_ostream &OS) {
+std::pair<uint64_t, uint64_t> InstrProfWriter::writeImpl(raw_ostream &OS) {
   OnDiskChainedHashTableGenerator<InstrProfRecordTrait> Generator;
 
   // Populate the hash table generator.
@@ -117,18 +135,54 @@ void InstrProfWriter::write(raw_fd_ostream &OS) {
   endian::Writer<little> LE(OS);
 
   // Write the header.
-  LE.write<uint64_t>(IndexedInstrProf::Magic);
-  LE.write<uint64_t>(IndexedInstrProf::Version);
-  LE.write<uint64_t>(MaxFunctionCount);
-  LE.write<uint64_t>(static_cast<uint64_t>(IndexedInstrProf::HashType));
+  IndexedInstrProf::Header Header;
+  Header.Magic = IndexedInstrProf::Magic;
+  Header.Version = IndexedInstrProf::Version;
+  Header.MaxFunctionCount = MaxFunctionCount;
+  Header.HashType = static_cast<uint64_t>(IndexedInstrProf::HashType);
+  Header.HashOffset = 0;
+  int N = sizeof(IndexedInstrProf::Header) / sizeof(uint64_t);
+
+  // Only write out all the fields execpt 'HashOffset'. We need
+  // to remember the offset of that field to allow back patching
+  // later.
+  for (int I = 0; I < N - 1; I++)
+    LE.write<uint64_t>(reinterpret_cast<uint64_t *>(&Header)[I]);
 
   // Save a space to write the hash table start location.
   uint64_t HashTableStartLoc = OS.tell();
+  // Reserve the space for HashOffset field.
   LE.write<uint64_t>(0);
   // Write the hash table.
   uint64_t HashTableStart = Generator.Emit(OS);
 
+  return std::make_pair(HashTableStartLoc, HashTableStart);
+}
+
+void InstrProfWriter::write(raw_fd_ostream &OS) {
+  // Write the hash table.
+  auto TableStart = writeImpl(OS);
+
+  // Go back and fill in the hash table start.
+  using namespace support;
+  OS.seek(TableStart.first);
+  // Now patch the HashOffset field previously reserved.
+  endian::Writer<little>(OS).write<uint64_t>(TableStart.second);
+}
+
+std::unique_ptr<MemoryBuffer> InstrProfWriter::writeBuffer() {
+  std::string Data;
+  llvm::raw_string_ostream OS(Data);
+  // Write the hash table.
+  auto TableStart = writeImpl(OS);
+  OS.flush();
+
   // Go back and fill in the hash table start.
-  OS.seek(HashTableStartLoc);
-  LE.write<uint64_t>(HashTableStart);
+  using namespace support;
+  uint64_t Bytes = endian::byte_swap<uint64_t, little>(TableStart.second);
+  Data.replace(TableStart.first, sizeof(uint64_t), (const char *)&Bytes,
+               sizeof(uint64_t));
+
+  // Return this in an aligned memory buffer.
+  return MemoryBuffer::getMemBufferCopy(Data);
 }