X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FProfileData%2FInstrProfWriter.cpp;h=1c4a4fede2868f1fc655b6cf57528bb81e590bfa;hb=12af22e8cc217827cf4f118b0f5e4ebbda9925ae;hp=3024f9676e329883c8c5f841ef98caadd1079a71;hpb=c61e1f53d97e35b9f75dabb7099f35eabffcce5b;p=oota-llvm.git

diff --git a/lib/ProfileData/InstrProfWriter.cpp b/lib/ProfileData/InstrProfWriter.cpp
index 3024f9676e3..1c4a4fede28 100644
--- a/lib/ProfileData/InstrProfWriter.cpp
+++ b/lib/ProfileData/InstrProfWriter.cpp
@@ -13,48 +13,124 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/ProfileData/InstrProfWriter.h"
-#include "llvm/Support/Endian.h"
+#include "llvm/ADT/StringExtras.h"
+#include "llvm/Support/EndianStream.h"
+#include "llvm/Support/OnDiskHashTable.h"
+
+#include "InstrProfIndexed.h"
 
 using namespace llvm;
 
-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;
-    return instrprof_error::success;;
+namespace {
+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 uint64_t hash_value_type;
+  typedef uint64_t offset_type;
+
+  static hash_value_type ComputeHash(key_type_ref K) {
+    return IndexedInstrProf::ComputeHash(IndexedInstrProf::HashType, K);
+  }
+
+  static std::pair<offset_type, offset_type>
+  EmitKeyDataLength(raw_ostream &Out, key_type_ref K, data_type_ref V) {
+    using namespace llvm::support;
+    endian::Writer<little> LE(Out);
+
+    offset_type N = K.size();
+    LE.write<offset_type>(N);
+
+    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);
+  }
+
+  static void EmitKey(raw_ostream &Out, key_type_ref K, offset_type N){
+    Out.write(K.data(), N);
+  }
+
+  static void EmitData(raw_ostream &Out, key_type_ref, data_type_ref V,
+                       offset_type) {
+    using namespace llvm::support;
+    endian::Writer<little> LE(Out);
+
+    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);
+    }
   }
+};
+}
 
-  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())
+std::error_code
+InstrProfWriter::addFunctionCounts(StringRef FunctionName,
+                                   uint64_t FunctionHash,
+                                   ArrayRef<uint64_t> 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;
+  }
+
+  // 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_ostream &OS) {
-  // Write out the counts for each function.
-  for (const auto &I : FunctionData) {
-    StringRef Name = I.getKey();
-    uint64_t Hash = I.getValue().Hash;
-    const std::vector<uint64_t> &Counts = I.getValue().Counts;
-
-    OS << Name << "\n" << Hash << "\n" << Counts.size() << "\n";
-    for (uint64_t Count : Counts)
-      OS << Count << "\n";
-    OS << "\n";
-  }
+void InstrProfWriter::write(raw_fd_ostream &OS) {
+  OnDiskChainedHashTableGenerator<InstrProfRecordTrait> Generator;
+
+  // Populate the hash table generator.
+  std::vector<uint64_t> CounterBuffer;
+  for (const auto &I : FunctionData)
+    Generator.insert(I.getKey(), &I.getValue());
+
+  using namespace llvm::support;
+  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));
+
+  // Save a space to write the hash table start location.
+  uint64_t HashTableStartLoc = OS.tell();
+  LE.write<uint64_t>(0);
+  // Write the hash table.
+  uint64_t HashTableStart = Generator.Emit(OS);
+
+  // Go back and fill in the hash table start.
+  OS.seek(HashTableStartLoc);
+  LE.write<uint64_t>(HashTableStart);
 }