Merge branch 'main' into ide_folders_libclc

Author: Meinersbur

.ci/generate-buildkite-pipeline-premerge

@@ -68,7 +68,7 @@ function compute-projects-to-test() {
       done
     ;;
     clang)
-      for p in clang-tools-extra compiler-rt flang lldb cross-project-tests; do
+      for p in clang-tools-extra compiler-rt lldb cross-project-tests; do
         echo $p
       done
     ;;

bolt/docs/BAT.md

@@ -106,9 +106,14 @@ equals output offset.
 `BRANCHENTRY` bit denotes whether a given offset pair is a control flow source
 (branch or call instruction). If not set, it signifies a control flow target
 (basic block offset).
+
 `InputAddr` is omitted for equal offsets in input and output function. In this
 case, `BRANCHENTRY` bits are encoded separately in a `BranchEntries` bitvector.
 
+Deleted basic blocks are emitted as having `OutputOffset` equal to the size of
+the function. They don't affect address translation and only participate in
+input basic block mapping.
+
 ### Secondary Entry Points table
 The table is emitted for hot fragments only. It contains `NumSecEntryPoints`
 offsets denoting secondary entry points, delta encoded, implicitly starting at zero.

bolt/include/bolt/Core/BinaryContext.h

@@ -17,6 +17,7 @@
 #include "bolt/Core/BinaryData.h"
 #include "bolt/Core/BinarySection.h"
 #include "bolt/Core/DebugData.h"
+#include "bolt/Core/DynoStats.h"
 #include "bolt/Core/JumpTable.h"
 #include "bolt/Core/MCPlusBuilder.h"
 #include "bolt/RuntimeLibs/RuntimeLibrary.h"
@@ -359,7 +360,7 @@ class BinaryContext {
   void setFileBuildID(StringRef ID) { FileBuildID = std::string(ID); }
 
   bool hasSymbolsWithFileName() const { return HasSymbolsWithFileName; }
-  void setHasSymbolsWithFileName(bool Value) { HasSymbolsWithFileName = true; }
+  void setHasSymbolsWithFileName(bool Value) { HasSymbolsWithFileName = Value; }
 
   /// Return true if relocations against symbol with a given name
   /// must be created.
@@ -677,6 +678,9 @@ class BinaryContext {
   /// have an origin file name available.
   bool HasSymbolsWithFileName{false};
 
+  /// Does the binary have BAT section.
+  bool HasBATSection{false};
+
   /// Sum of execution count of all functions
   uint64_t SumExecutionCount{0};
 
@@ -714,6 +718,9 @@ class BinaryContext {
     uint64_t NumStaleBlocksWithEqualIcount{0};
   } Stats;
 
+  // Original binary execution count stats.
+  DynoStats InitialDynoStats;
+
   // Address of the first allocated segment.
   uint64_t FirstAllocAddress{std::numeric_limits<uint64_t>::max()};
 
@@ -1217,8 +1224,7 @@ class BinaryContext {
 
   /// Return a signed value of \p Size stored at \p Address. The address has
   /// to be a valid statically allocated address for the binary.
-  ErrorOr<uint64_t> getSignedValueAtAddress(uint64_t Address,
-                                            size_t Size) const;
+  ErrorOr<int64_t> getSignedValueAtAddress(uint64_t Address, size_t Size) const;
 
   /// Special case of getUnsignedValueAtAddress() that uses a pointer size.
   ErrorOr<uint64_t> getPointerAtAddress(uint64_t Address) const {

bolt/include/bolt/Passes/BinaryPasses.h

@@ -16,6 +16,7 @@
 #include "bolt/Core/BinaryContext.h"
 #include "bolt/Core/BinaryFunction.h"
 #include "bolt/Core/DynoStats.h"
+#include "bolt/Profile/BoltAddressTranslation.h"
 #include "llvm/Support/CommandLine.h"
 #include <atomic>
 #include <set>
@@ -52,15 +53,31 @@ class BinaryFunctionPass {
   virtual Error runOnFunctions(BinaryContext &BC) = 0;
 };
 
+/// A pass to set initial program-wide dynostats.
+class DynoStatsSetPass : public BinaryFunctionPass {
+public:
+  DynoStatsSetPass() : BinaryFunctionPass(false) {}
+
+  const char *getName() const override {
+    return "set dyno-stats before optimizations";
+  }
+
+  bool shouldPrint(const BinaryFunction &BF) const override { return false; }
+
+  Error runOnFunctions(BinaryContext &BC) override {
+    BC.InitialDynoStats = getDynoStats(BC.getBinaryFunctions(), BC.isAArch64());
+    return Error::success();
+  }
+};
+
 /// A pass to print program-wide dynostats.
 class DynoStatsPrintPass : public BinaryFunctionPass {
 protected:
-  DynoStats PrevDynoStats;
   std::string Title;
 
 public:
-  DynoStatsPrintPass(const DynoStats &PrevDynoStats, const char *Title)
-      : BinaryFunctionPass(false), PrevDynoStats(PrevDynoStats), Title(Title) {}
+  DynoStatsPrintPass(const char *Title)
+      : BinaryFunctionPass(false), Title(Title) {}
 
   const char *getName() const override {
     return "print dyno-stats after optimizations";
@@ -69,6 +86,7 @@ class DynoStatsPrintPass : public BinaryFunctionPass {
   bool shouldPrint(const BinaryFunction &BF) const override { return false; }
 
   Error runOnFunctions(BinaryContext &BC) override {
+    const DynoStats PrevDynoStats = BC.InitialDynoStats;
     const DynoStats NewDynoStats =
         getDynoStats(BC.getBinaryFunctions(), BC.isAArch64());
     const bool Changed = (NewDynoStats != PrevDynoStats);
@@ -399,8 +417,11 @@ class PrintProfileStats : public BinaryFunctionPass {
 /// Prints a list of the top 100 functions sorted by a set of
 /// dyno stats categories.
 class PrintProgramStats : public BinaryFunctionPass {
+  BoltAddressTranslation *BAT = nullptr;
+
 public:
-  explicit PrintProgramStats() : BinaryFunctionPass(false) {}
+  explicit PrintProgramStats(BoltAddressTranslation *BAT = nullptr)
+      : BinaryFunctionPass(false), BAT(BAT) {}
 
   const char *getName() const override { return "print-stats"; }
   bool shouldPrint(const BinaryFunction &) const override { return false; }

bolt/include/bolt/Passes/MCF.h

@@ -9,20 +9,14 @@
 #ifndef BOLT_PASSES_MCF_H
 #define BOLT_PASSES_MCF_H
 
+#include "bolt/Passes/BinaryPasses.h"
+#include "llvm/Support/CommandLine.h"
+
 namespace llvm {
 namespace bolt {
 
-class BinaryFunction;
 class DataflowInfoManager;
 
-enum MCFCostFunction : char {
-  MCF_DISABLE = 0,
-  MCF_LINEAR,
-  MCF_QUADRATIC,
-  MCF_LOG,
-  MCF_BLAMEFTS
-};
-
 /// Implement the idea in "SamplePGO - The Power of Profile Guided Optimizations
 /// without the Usability Burden" by Diego Novillo to make basic block counts
 /// equal if we show that A dominates B, B post-dominates A and they are in the
@@ -31,23 +25,18 @@ void equalizeBBCounts(DataflowInfoManager &Info, BinaryFunction &BF);
 
 /// Fill edge counts based on the basic block count. Used in nonLBR mode when
 /// we only have bb count.
-void estimateEdgeCounts(BinaryFunction &BF);
-
-/// Entry point for computing a min-cost flow for the CFG with the goal
-/// of fixing the flow of the CFG edges, that is, making sure it obeys the
-/// flow-conservation equation  SumInEdges = SumOutEdges.
-///
-/// To do this, we create an instance of the min-cost flow problem in a
-/// similar way as the one discussed in the work of Roy Levin "Completing
-/// Incomplete Edge Profile by Applying Minimum Cost Circulation Algorithms".
-/// We do a few things differently, though. We don't populate edge counts using
-/// weights coming from a static branch prediction technique and we don't
-/// use the same cost function.
-///
-/// If cost function BlameFTs is used, assign all remaining flow to
-/// fall-throughs. This is used when the sampling is based on taken branches
-/// that do not account for them.
-void solveMCF(BinaryFunction &BF, MCFCostFunction CostFunction);
+class EstimateEdgeCounts : public BinaryFunctionPass {
+  void runOnFunction(BinaryFunction &BF);
+
+public:
+  explicit EstimateEdgeCounts(const cl::opt<bool> &PrintPass)
+      : BinaryFunctionPass(PrintPass) {}
+
+  const char *getName() const override { return "estimate-edge-counts"; }
+
+  /// Pass entry point
+  Error runOnFunctions(BinaryContext &BC) override;
+};
 
 } // end namespace bolt
 } // end namespace llvm

bolt/include/bolt/Passes/StokeInfo.h

@@ -87,10 +87,10 @@ struct StokeFuncInfo {
               << "," << NumBlocks << "," << IsLoopFree << "," << NumLoops << ","
               << MaxLoopDepth << "," << HotSize << "," << TotalSize << ","
               << Score << "," << HasCall << ",\"{ ";
-      for (std::string S : DefIn)
+      for (const std::string &S : DefIn)
         Outfile << "%" << S << " ";
       Outfile << "}\",\"{ ";
-      for (std::string S : LiveOut)
+      for (const std::string &S : LiveOut)
         Outfile << "%" << S << " ";
       Outfile << "}\"," << HeapOut << "," << StackOut << "," << HasRipAddr
               << "," << Omitted << "\n";

bolt/include/bolt/Profile/BoltAddressTranslation.h

@@ -70,7 +70,7 @@ class BinaryFunction;
 class BoltAddressTranslation {
 public:
   // In-memory representation of the address translation table
-  using MapTy = std::map<uint32_t, uint32_t>;
+  using MapTy = std::multimap<uint32_t, uint32_t>;
 
   // List of taken fall-throughs
   using FallthroughListTy = SmallVector<std::pair<uint64_t, uint64_t>, 16>;
@@ -90,7 +90,7 @@ class BoltAddressTranslation {
   std::error_code parse(raw_ostream &OS, StringRef Buf);
 
   /// Dump the parsed address translation tables
-  void dump(raw_ostream &OS);
+  void dump(raw_ostream &OS) const;
 
   /// If the maps are loaded in memory, perform the lookup to translate LBR
   /// addresses in function located at \p FuncAddress.
@@ -107,7 +107,12 @@ class BoltAddressTranslation {
 
   /// If available, fetch the address of the hot part linked to the cold part
   /// at \p Address. Return 0 otherwise.
-  uint64_t fetchParentAddress(uint64_t Address) const;
+  uint64_t fetchParentAddress(uint64_t Address) const {
+    auto Iter = ColdPartSource.find(Address);
+    if (Iter == ColdPartSource.end())
+      return 0;
+    return Iter->second;
+  }
 
   /// True if the input binary has a translation table we can use to convert
   /// addresses when aggregating profile
@@ -132,7 +137,8 @@ class BoltAddressTranslation {
   /// emitted for the start of the BB. More entries may be emitted to cover
   /// the location of calls or any instruction that may change control flow.
   void writeEntriesForBB(MapTy &Map, const BinaryBasicBlock &BB,
-                         uint64_t FuncInputAddress, uint64_t FuncOutputAddress);
+                         uint64_t FuncInputAddress,
+                         uint64_t FuncOutputAddress) const;
 
   /// Write the serialized address translation table for a function.
   template <bool Cold>
@@ -147,7 +153,7 @@ class BoltAddressTranslation {
 
   /// Returns the bitmask with set bits corresponding to indices of BRANCHENTRY
   /// entries in function address translation map.
-  APInt calculateBranchEntriesBitMask(MapTy &Map, size_t EqualElems);
+  APInt calculateBranchEntriesBitMask(MapTy &Map, size_t EqualElems) const;
 
   /// Calculate the number of equal offsets (output = input - skew) in the
   /// beginning of the function.
@@ -178,14 +184,9 @@ class BoltAddressTranslation {
 public:
   /// Map basic block input offset to a basic block index and hash pair.
   class BBHashMapTy {
-    class EntryTy {
+    struct EntryTy {
       unsigned Index;
       size_t Hash;
-
-    public:
-      unsigned getBBIndex() const { return Index; }
-      size_t getBBHash() const { return Hash; }
-      EntryTy(unsigned Index, size_t Hash) : Index(Index), Hash(Hash) {}
     };
 
     std::map<uint32_t, EntryTy> Map;
@@ -201,34 +202,30 @@ class BoltAddressTranslation {
     }
 
     unsigned getBBIndex(uint32_t BBInputOffset) const {
-      return getEntry(BBInputOffset).getBBIndex();
+      return getEntry(BBInputOffset).Index;
     }
 
     size_t getBBHash(uint32_t BBInputOffset) const {
-      return getEntry(BBInputOffset).getBBHash();
+      return getEntry(BBInputOffset).Hash;
     }
 
     void addEntry(uint32_t BBInputOffset, unsigned BBIndex, size_t BBHash) {
-      Map.emplace(BBInputOffset, EntryTy(BBIndex, BBHash));
+      Map.emplace(BBInputOffset, EntryTy{BBIndex, BBHash});
     }
 
     size_t getNumBasicBlocks() const { return Map.size(); }
 
     auto begin() const { return Map.begin(); }
     auto end() const { return Map.end(); }
     auto upper_bound(uint32_t Offset) const { return Map.upper_bound(Offset); }
+    auto size() const { return Map.size(); }
   };
 
   /// Map function output address to its hash and basic blocks hash map.
   class FuncHashesTy {
-    class EntryTy {
+    struct EntryTy {
       size_t Hash;
       BBHashMapTy BBHashMap;
-
-    public:
-      size_t getBFHash() const { return Hash; }
-      const BBHashMapTy &getBBHashMap() const { return BBHashMap; }
-      EntryTy(size_t Hash) : Hash(Hash) {}
     };
 
     std::unordered_map<uint64_t, EntryTy> Map;
@@ -240,23 +237,23 @@ class BoltAddressTranslation {
 
   public:
     size_t getBFHash(uint64_t FuncOutputAddress) const {
-      return getEntry(FuncOutputAddress).getBFHash();
+      return getEntry(FuncOutputAddress).Hash;
     }
 
     const BBHashMapTy &getBBHashMap(uint64_t FuncOutputAddress) const {
-      return getEntry(FuncOutputAddress).getBBHashMap();
+      return getEntry(FuncOutputAddress).BBHashMap;
     }
 
     void addEntry(uint64_t FuncOutputAddress, size_t BFHash) {
-      Map.emplace(FuncOutputAddress, EntryTy(BFHash));
+      Map.emplace(FuncOutputAddress, EntryTy{BFHash, BBHashMapTy()});
     }
 
     size_t getNumFunctions() const { return Map.size(); };
 
     size_t getNumBasicBlocks() const {
       size_t NumBasicBlocks{0};
       for (auto &I : Map)
-        NumBasicBlocks += I.second.getBBHashMap().getNumBasicBlocks();
+        NumBasicBlocks += I.second.BBHashMap.getNumBasicBlocks();
       return NumBasicBlocks;
     }
   };
@@ -278,7 +275,9 @@ class BoltAddressTranslation {
 
   /// Returns the number of basic blocks in a function.
   size_t getNumBasicBlocks(uint64_t OutputAddress) const {
-    return NumBasicBlocksMap.at(OutputAddress);
+    auto It = NumBasicBlocksMap.find(OutputAddress);
+    assert(It != NumBasicBlocksMap.end());
+    return It->second;
   }
 
 private:

bolt/include/bolt/Profile/DataAggregator.h

@@ -15,6 +15,7 @@
 #define BOLT_PROFILE_DATA_AGGREGATOR_H
 
 #include "bolt/Profile/DataReader.h"
+#include "bolt/Profile/YAMLProfileWriter.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/Support/Error.h"
 #include "llvm/Support/Program.h"
@@ -248,7 +249,7 @@ class DataAggregator : public DataReader {
   BinaryFunction *getBATParentFunction(const BinaryFunction &Func) const;
 
   /// Retrieve the location name to be used for samples recorded in \p Func.
-  StringRef getLocationName(const BinaryFunction &Func) const;
+  static StringRef getLocationName(const BinaryFunction &Func, bool BAT);
 
   /// Semantic actions - parser hooks to interpret parsed perf samples
   /// Register a sample (non-LBR mode), i.e. a new hit at \p Address
@@ -490,6 +491,8 @@ class DataAggregator : public DataReader {
   /// Parse the output generated by "perf buildid-list" to extract build-ids
   /// and return a file name matching a given \p FileBuildID.
   std::optional<StringRef> getFileNameForBuildID(StringRef FileBuildID);
+
+  friend class YAMLProfileWriter;
 };
 } // namespace bolt
 } // namespace llvm