[𝘀𝗽𝗿] initial version

Created using spr 1.3.4

Author: aaupov

bolt/include/bolt/Core/BinaryFunction.h

@@ -386,8 +386,8 @@ class BinaryFunction {
   /// Profile match ratio.
   float ProfileMatchRatio{0.0f};
 
-  /// Raw branch count for this function in the profile.
-  uint64_t RawBranchCount{0};
+  /// Raw sample/branch count for this function in the profile.
+  uint64_t RawSampleCount{0};
 
   /// Dynamically executed function bytes, used for density computation.
   uint64_t SampleCountInBytes{0};
@@ -1880,13 +1880,12 @@ class BinaryFunction {
   /// Return COUNT_NO_PROFILE if there's no profile info.
   uint64_t getExecutionCount() const { return ExecutionCount; }
 
-  /// Return the raw profile information about the number of branch
-  /// executions corresponding to this function.
-  uint64_t getRawBranchCount() const { return RawBranchCount; }
+  /// Return the raw profile information about the number of samples (basic
+  /// profile) or branch executions (branch profile) recorded in this function.
+  uint64_t getRawSampleCount() const { return RawSampleCount; }
 
-  /// Set the profile data about the number of branch executions corresponding
-  /// to this function.
-  void setRawBranchCount(uint64_t Count) { RawBranchCount = Count; }
+  /// Set raw count of samples or branches recorded in this function.
+  void setRawSampleCount(uint64_t Count) { RawSampleCount = Count; }
 
   /// Return the number of dynamically executed bytes, from raw perf data.
   uint64_t getSampleCountInBytes() const { return SampleCountInBytes; }

bolt/include/bolt/Profile/DataAggregator.h

@@ -92,16 +92,6 @@ class DataAggregator : public DataReader {
     uint64_t Addr;
   };
 
-  /// Used for parsing specific pre-aggregated input files.
-  struct AggregatedLBREntry {
-    enum Type : char { BRANCH = 0, FT, FT_EXTERNAL_ORIGIN, TRACE };
-    Location From;
-    Location To;
-    uint64_t Count;
-    uint64_t Mispreds;
-    Type EntryType;
-  };
-
   struct Trace {
     uint64_t From;
     uint64_t To;
@@ -131,7 +121,6 @@ class DataAggregator : public DataReader {
   /// and use them later for processing and assigning profile.
   std::unordered_map<Trace, TakenBranchInfo, TraceHash> BranchLBRs;
   std::unordered_map<Trace, FTInfo, TraceHash> FallthroughLBRs;
-  std::vector<AggregatedLBREntry> AggregatedLBRs;
   std::unordered_map<uint64_t, uint64_t> BasicSamples;
   std::vector<PerfMemSample> MemSamples;
 
@@ -223,11 +212,6 @@ class DataAggregator : public DataReader {
   uint64_t NumTraces{0};
   uint64_t NumInvalidTraces{0};
   uint64_t NumLongRangeTraces{0};
-  /// Specifies how many samples were recorded in cold areas if we are dealing
-  /// with profiling data collected in a bolted binary. For LBRs, incremented
-  /// for the source of the branch to avoid counting cold activity twice (one
-  /// for source and another for destination).
-  uint64_t NumColdSamples{0};
   uint64_t NumTotalSamples{0};
 
   /// Looks into system PATH for Linux Perf and set up the aggregator to use it
@@ -416,14 +400,7 @@ class DataAggregator : public DataReader {
   /// F 41be90 41be90 4
   /// B 4b1942 39b57f0 3 0
   /// B 4b196f 4b19e0 2 0
-  void parsePreAggregated();
-
-  /// Parse the full output of pre-aggregated LBR samples generated by
-  /// an external tool.
-  std::error_code parsePreAggregatedLBRSamples();
-
-  /// Process parsed pre-aggregated data.
-  void processPreAggregated();
+  std::error_code parsePreAggregated();
 
   /// If \p Address falls into the binary address space based on memory
   /// mapping info \p MMI, then adjust it for further processing by subtracting
@@ -486,7 +463,6 @@ class DataAggregator : public DataReader {
   void dump(const PerfMemSample &Sample) const;
 
   /// Profile diagnostics print methods
-  void printColdSamplesDiagnostic() const;
   void printLongRangeTracesDiagnostic() const;
   void printBranchSamplesDiagnostics() const;
   void printBasicSamplesDiagnostics(uint64_t OutOfRangeSamples) const;

bolt/include/bolt/Profile/DataReader.h

@@ -252,6 +252,9 @@ struct FuncSampleData {
   /// Get the number of samples recorded in [Start, End)
   uint64_t getSamples(uint64_t Start, uint64_t End) const;
 
+  /// Returns the total number of samples recorded in this function.
+  uint64_t getSamples() const;
+
   /// Aggregation helper
   DenseMap<uint64_t, size_t> Index;
 

bolt/include/bolt/Profile/Heatmap.h

@@ -57,9 +57,9 @@ class Heatmap {
   }
 
   /// Register a single sample at \p Address.
-  void registerAddress(uint64_t Address) {
+  void registerAddress(uint64_t Address, uint64_t Count) {
     if (!ignoreAddress(Address))
-      ++Map[Address / BucketSize];
+      Map[Address / BucketSize] += Count;
   }
 
   /// Register \p Count samples at [\p StartAddress, \p EndAddress ].

bolt/include/bolt/Utils/CommandLineOpts.h

@@ -44,6 +44,7 @@ extern llvm::cl::opt<unsigned> HeatmapBlock;
 extern llvm::cl::opt<unsigned long long> HeatmapMaxAddress;
 extern llvm::cl::opt<unsigned long long> HeatmapMinAddress;
 extern llvm::cl::opt<bool> HeatmapPrintMappings;
+extern llvm::cl::opt<bool> HeatmapStats;
 extern llvm::cl::opt<bool> HotData;
 extern llvm::cl::opt<bool> HotFunctionsAtEnd;
 extern llvm::cl::opt<bool> HotText;

bolt/lib/Core/BinaryFunction.cpp

@@ -473,7 +473,7 @@ void BinaryFunction::print(raw_ostream &OS, std::string Annotation) {
     OS << "\n  Image       : 0x" << Twine::utohexstr(getImageAddress());
   if (ExecutionCount != COUNT_NO_PROFILE) {
     OS << "\n  Exec Count  : " << ExecutionCount;
-    OS << "\n  Branch Count: " << RawBranchCount;
+    OS << "\n  Branch Count: " << RawSampleCount;
     OS << "\n  Profile Acc : " << format("%.1f%%", ProfileMatchRatio * 100.0f);
   }
 

bolt/lib/Passes/BinaryPasses.cpp

@@ -1445,7 +1445,7 @@ Error PrintProgramStats::runOnFunctions(BinaryContext &BC) {
     if (!Function.hasProfile())
       continue;
 
-    uint64_t SampleCount = Function.getRawBranchCount();
+    uint64_t SampleCount = Function.getRawSampleCount();
     TotalSampleCount += SampleCount;
 
     if (Function.hasValidProfile()) {