Merge branch 'main' into users/vikramRH/enable_opt

Author: vikramRH

.github/new-prs-labeler.yml

@@ -775,6 +775,29 @@ backend:AArch64:
   - clang/include/clang/Sema/SemaARM.h
   - clang/lib/Sema/SemaARM.cpp
 
+backend:Hexagon:
+  - clang/include/clang/Basic/BuiltinsHexagon*.def
+  - clang/include/clang/Sema/SemaHexagon.h
+  - clang/lib/Basic/Targets/Hexagon.*
+  - clang/lib/CodeGen/Targets/Hexagon.cpp
+  - clang/lib/Driver/ToolChains/Hexagon.*
+  - clang/lib/Sema/SemaHexagon.cpp
+  - lld/ELF/Arch/Hexagon.cpp
+  - lldb/source/Plugins/ABI/Hexagon/**
+  - lldb/source/Plugins/DynamicLoader/Hexagon-DYLD/**
+  - llvm/include/llvm/BinaryFormat/ELFRelocs/Hexagon.def
+  - llvm/include/llvm/IR/IntrinsicsHexagon*
+  - llvm/include/llvm/Support/Hexagon*
+  - llvm/lib/Support/Hexagon*
+  - llvm/lib/Target/Hexagon/**
+  - llvm/test/CodeGen/Hexagon/**
+  - llvm/test/CodeGen/*/Hexagon/**
+  - llvm/test/DebugInfo/*/Hexagon/**
+  - llvm/test/Transforms/*/Hexagon
+  - llvm/test/MC/Disassembler/Hexagon/**
+  - llvm/test/MC/Hexagon/**
+  - llvm/test/tools/llvm-objdump/ELF/Hexagon/**
+
 backend:loongarch:
   - llvm/include/llvm/IR/IntrinsicsLoongArch.td
   - llvm/test/MC/LoongArch/**

bolt/include/bolt/Core/DebugData.h

@@ -226,17 +226,16 @@ class DebugRangesSectionWriter {
   /// Needs to be invoked before each \p CU is processed.
   void virtual initSection(DWARFUnit &CU){};
 
+  /// Initializes Ranges section with empty list.
+  void initSection();
+
 protected:
   std::unique_ptr<DebugBufferVector> RangesBuffer;
 
   std::unique_ptr<raw_svector_ostream> RangesStream;
 
   std::mutex WriterMutex;
 
-  /// Current offset in the section (updated as new entries are written).
-  /// Starts with 16 since the first 16 bytes are reserved for an empty range.
-  uint32_t SectionOffset{0};
-
   /// Offset of an empty address ranges list.
   static constexpr uint64_t EmptyRangesOffset{0};
 

bolt/include/bolt/Core/HashUtilities.h

@@ -16,6 +16,7 @@
 
 #include "bolt/Core/BinaryBasicBlock.h"
 #include "bolt/Core/BinaryContext.h"
+#include "bolt/Profile/ProfileYAMLMapping.h"
 
 namespace llvm {
 namespace bolt {
@@ -35,6 +36,13 @@ std::string hashBlock(BinaryContext &BC, const BinaryBasicBlock &BB,
 
 std::string hashBlockLoose(BinaryContext &BC, const BinaryBasicBlock &BB);
 
+std::string hashBlockCalls(BinaryContext &BC, const BinaryBasicBlock &BB);
+
+std::string
+hashBlockCalls(const DenseMap<uint32_t, yaml::bolt::BinaryFunctionProfile *>
+                   &IdToYamlFunction,
+               const yaml::bolt::BinaryBasicBlockProfile &YamlBB);
+
 } // namespace bolt
 } // namespace llvm
 

bolt/include/bolt/Profile/YAMLProfileReader.h

@@ -40,6 +40,9 @@ class YAMLProfileReader : public ProfileReaderBase {
   /// Check if the file contains YAML.
   static bool isYAML(StringRef Filename);
 
+  using ProfileLookupMap =
+      DenseMap<uint32_t, yaml::bolt::BinaryFunctionProfile *>;
+
 private:
   /// Adjustments for basic samples profiles (without LBR).
   bool NormalizeByInsnCount{false};
@@ -56,6 +59,10 @@ class YAMLProfileReader : public ProfileReaderBase {
   /// is attributed.
   FunctionSet ProfiledFunctions;
 
+  /// Maps profiled function id to function, for function matching with calls as
+  /// anchors.
+  ProfileLookupMap IdToYamLBF;
+
   /// For LTO symbol resolution.
   /// Map a common LTO prefix to a list of YAML profiles matching the prefix.
   StringMap<std::vector<yaml::bolt::BinaryFunctionProfile *>> LTOCommonNameMap;

bolt/include/bolt/Utils/NameResolver.h

@@ -61,6 +61,12 @@ class NameResolver {
     std::tie(LHS, RHS) = UniqueName.split(Sep);
     return (LHS + Suffix + Twine(Sep) + RHS).str();
   }
+
+  // Drops the suffix that describes the function's number of names.
+  static StringRef dropNumNames(StringRef Name) {
+    const size_t Pos = Name.find("(*");
+    return Pos != StringRef::npos ? Name.substr(0, Pos) : Name;
+  }
 };
 
 } // namespace bolt

bolt/lib/Core/DebugData.cpp

@@ -137,12 +137,14 @@ DebugRangesSectionWriter::DebugRangesSectionWriter() {
   RangesBuffer = std::make_unique<DebugBufferVector>();
   RangesStream = std::make_unique<raw_svector_ostream>(*RangesBuffer);
 
-  // Add an empty range as the first entry;
-  SectionOffset +=
-      writeAddressRanges(*RangesStream.get(), DebugAddressRangesVector{});
   Kind = RangesWriterKind::DebugRangesWriter;
 }
 
+void DebugRangesSectionWriter::initSection() {
+  // Adds an empty range to the buffer.
+  writeAddressRanges(*RangesStream.get(), DebugAddressRangesVector{});
+}
+
 uint64_t DebugRangesSectionWriter::addRanges(
     DebugAddressRangesVector &&Ranges,
     std::map<DebugAddressRangesVector, uint64_t> &CachedRanges) {
@@ -166,21 +168,20 @@ uint64_t DebugRangesSectionWriter::addRanges(DebugAddressRangesVector &Ranges) {
   // Reading the SectionOffset and updating it should be atomic to guarantee
   // unique and correct offsets in patches.
   std::lock_guard<std::mutex> Lock(WriterMutex);
-  const uint32_t EntryOffset = SectionOffset;
-  SectionOffset += writeAddressRanges(*RangesStream.get(), Ranges);
+  const uint32_t EntryOffset = RangesBuffer->size();
+  writeAddressRanges(*RangesStream.get(), Ranges);
 
   return EntryOffset;
 }
 
 uint64_t DebugRangesSectionWriter::getSectionOffset() {
   std::lock_guard<std::mutex> Lock(WriterMutex);
-  return SectionOffset;
+  return RangesBuffer->size();
 }
 
 void DebugRangesSectionWriter::appendToRangeBuffer(
     const DebugBufferVector &CUBuffer) {
   *RangesStream << CUBuffer;
-  SectionOffset = RangesBuffer->size();
 }
 
 DebugAddrWriter *DebugRangeListsSectionWriter::AddrWriter = nullptr;
@@ -327,7 +328,6 @@ void DebugRangeListsSectionWriter::finalizeSection() {
   *RangesStream << *Header;
   *RangesStream << *CUArrayBuffer;
   *RangesStream << *CUBodyBuffer;
-  SectionOffset = RangesBuffer->size();
 }
 
 void DebugRangeListsSectionWriter::initSection(DWARFUnit &Unit) {

bolt/lib/Core/HashUtilities.cpp

@@ -12,6 +12,7 @@
 
 #include "bolt/Core/HashUtilities.h"
 #include "bolt/Core/BinaryContext.h"
+#include "bolt/Utils/NameResolver.h"
 #include "llvm/MC/MCInstPrinter.h"
 
 namespace llvm {
@@ -155,5 +156,50 @@ std::string hashBlockLoose(BinaryContext &BC, const BinaryBasicBlock &BB) {
   return HashString;
 }
 
+/// An even looser hash level relative to $ hashBlockLoose to use with stale
+/// profile matching, composed of the names of a block's called functions in
+/// lexicographic order.
+std::string hashBlockCalls(BinaryContext &BC, const BinaryBasicBlock &BB) {
+  // The hash is computed by creating a string of all lexicographically ordered
+  // called function names.
+  std::vector<std::string> FunctionNames;
+  for (const MCInst &Instr : BB) {
+    // Skip non-call instructions.
+    if (!BC.MIB->isCall(Instr))
+      continue;
+    const MCSymbol *CallSymbol = BC.MIB->getTargetSymbol(Instr);
+    if (!CallSymbol)
+      continue;
+    FunctionNames.push_back(std::string(CallSymbol->getName()));
+  }
+  std::sort(FunctionNames.begin(), FunctionNames.end());
+  std::string HashString;
+  for (const std::string &FunctionName : FunctionNames)
+    HashString.append(FunctionName);
+
+  return HashString;
+}
+
+/// The same as the $hashBlockCalls function, but for profiled functions.
+std::string
+hashBlockCalls(const DenseMap<uint32_t, yaml::bolt::BinaryFunctionProfile *>
+                   &IdToYamlFunction,
+               const yaml::bolt::BinaryBasicBlockProfile &YamlBB) {
+  std::vector<std::string> FunctionNames;
+  for (const yaml::bolt::CallSiteInfo &CallSiteInfo : YamlBB.CallSites) {
+    auto It = IdToYamlFunction.find(CallSiteInfo.DestId);
+    if (It == IdToYamlFunction.end())
+      continue;
+    StringRef Name = NameResolver::dropNumNames(It->second->Name);
+    FunctionNames.push_back(std::string(Name));
+  }
+  std::sort(FunctionNames.begin(), FunctionNames.end());
+  std::string HashString;
+  for (const std::string &FunctionName : FunctionNames)
+    HashString.append(FunctionName);
+
+  return HashString;
+}
+
 } // namespace bolt
 } // namespace llvm

bolt/lib/Profile/StaleProfileMatching.cpp

@@ -193,18 +193,43 @@ class StaleMatcher {
 public:
   /// Initialize stale matcher.
   void init(const std::vector<FlowBlock *> &Blocks,
-            const std::vector<BlendedBlockHash> &Hashes) {
+            const std::vector<BlendedBlockHash> &Hashes,
+            const std::vector<uint64_t> &CallHashes) {
     assert(Blocks.size() == Hashes.size() &&
+           Hashes.size() == CallHashes.size() &&
            "incorrect matcher initialization");
     for (size_t I = 0; I < Blocks.size(); I++) {
       FlowBlock *Block = Blocks[I];
       uint16_t OpHash = Hashes[I].OpcodeHash;
       OpHashToBlocks[OpHash].push_back(std::make_pair(Hashes[I], Block));
+      if (CallHashes[I])
+        CallHashToBlocks[CallHashes[I]].push_back(
+            std::make_pair(Hashes[I], Block));
     }
   }
 
   /// Find the most similar block for a given hash.
-  const FlowBlock *matchBlock(BlendedBlockHash BlendedHash) const {
+  const FlowBlock *matchBlock(BlendedBlockHash BlendedHash,
+                              uint64_t CallHash) const {
+    const FlowBlock *BestBlock = matchWithOpcodes(BlendedHash);
+    return BestBlock ? BestBlock : matchWithCalls(BlendedHash, CallHash);
+  }
+
+  /// Returns true if the two basic blocks (in the binary and in the profile)
+  /// corresponding to the given hashes are matched to each other with a high
+  /// confidence.
+  static bool isHighConfidenceMatch(BlendedBlockHash Hash1,
+                                    BlendedBlockHash Hash2) {
+    return Hash1.InstrHash == Hash2.InstrHash;
+  }
+
+private:
+  using HashBlockPairType = std::pair<BlendedBlockHash, FlowBlock *>;
+  std::unordered_map<uint16_t, std::vector<HashBlockPairType>> OpHashToBlocks;
+  std::unordered_map<uint64_t, std::vector<HashBlockPairType>> CallHashToBlocks;
+
+  // Uses OpcodeHash to find the most similar block for a given hash.
+  const FlowBlock *matchWithOpcodes(BlendedBlockHash BlendedHash) const {
     auto BlockIt = OpHashToBlocks.find(BlendedHash.OpcodeHash);
     if (BlockIt == OpHashToBlocks.end())
       return nullptr;
@@ -220,17 +245,27 @@ class StaleMatcher {
     return BestBlock;
   }
 
-  /// Returns true if the two basic blocks (in the binary and in the profile)
-  /// corresponding to the given hashes are matched to each other with a high
-  /// confidence.
-  static bool isHighConfidenceMatch(BlendedBlockHash Hash1,
-                                    BlendedBlockHash Hash2) {
-    return Hash1.InstrHash == Hash2.InstrHash;
+  // Uses CallHash to find the most similar block for a given hash.
+  const FlowBlock *matchWithCalls(BlendedBlockHash BlendedHash,
+                                  uint64_t CallHash) const {
+    if (!CallHash)
+      return nullptr;
+    auto BlockIt = CallHashToBlocks.find(CallHash);
+    if (BlockIt == CallHashToBlocks.end())
+      return nullptr;
+    FlowBlock *BestBlock = nullptr;
+    uint64_t BestDist = std::numeric_limits<uint64_t>::max();
+    for (const auto &[Hash, Block] : BlockIt->second) {
+      uint64_t Dist = Hash.OpcodeHash > BlendedHash.OpcodeHash
+                          ? Hash.OpcodeHash - BlendedHash.OpcodeHash
+                          : BlendedHash.OpcodeHash - Hash.OpcodeHash;
+      if (BestBlock == nullptr || Dist < BestDist) {
+        BestDist = Dist;
+        BestBlock = Block;
+      }
+    }
+    return BestBlock;
   }
-
-private:
-  using HashBlockPairType = std::pair<BlendedBlockHash, FlowBlock *>;
-  std::unordered_map<uint16_t, std::vector<HashBlockPairType>> OpHashToBlocks;
 };
 
 void BinaryFunction::computeBlockHashes(HashFunction HashFunction) const {
@@ -412,33 +447,62 @@ createFlowFunction(const BinaryFunction::BasicBlockOrderType &BlockOrder) {
 /// of the basic blocks in the binary, the count is "matched" to the block.
 /// Similarly, if both the source and the target of a count in the profile are
 /// matched to a jump in the binary, the count is recorded in CFG.
-size_t matchWeightsByHashes(
-    BinaryContext &BC, const BinaryFunction::BasicBlockOrderType &BlockOrder,
-    const yaml::bolt::BinaryFunctionProfile &YamlBF, FlowFunction &Func) {
+size_t
+matchWeightsByHashes(BinaryContext &BC,
+                     const BinaryFunction::BasicBlockOrderType &BlockOrder,
+                     const yaml::bolt::BinaryFunctionProfile &YamlBF,
+                     FlowFunction &Func, HashFunction HashFunction,
+                     YAMLProfileReader::ProfileLookupMap &IdToYamlBF) {
+
   assert(Func.Blocks.size() == BlockOrder.size() + 2);
 
+  std::vector<uint64_t> CallHashes;
   std::vector<FlowBlock *> Blocks;
   std::vector<BlendedBlockHash> BlendedHashes;
   for (uint64_t I = 0; I < BlockOrder.size(); I++) {
     const BinaryBasicBlock *BB = BlockOrder[I];
     assert(BB->getHash() != 0 && "empty hash of BinaryBasicBlock");
+
+    std::string CallHashStr = hashBlockCalls(BC, *BB);
+    if (CallHashStr.empty()) {
+      CallHashes.push_back(0);
+    } else {
+      if (HashFunction == HashFunction::StdHash)
+        CallHashes.push_back(std::hash<std::string>{}(CallHashStr));
+      else if (HashFunction == HashFunction::XXH3)
+        CallHashes.push_back(llvm::xxh3_64bits(CallHashStr));
+      else
+        llvm_unreachable("Unhandled HashFunction");
+    }
+
     Blocks.push_back(&Func.Blocks[I + 1]);
     BlendedBlockHash BlendedHash(BB->getHash());
     BlendedHashes.push_back(BlendedHash);
     LLVM_DEBUG(dbgs() << "BB with index " << I << " has hash = "
                       << Twine::utohexstr(BB->getHash()) << "\n");
   }
   StaleMatcher Matcher;
-  Matcher.init(Blocks, BlendedHashes);
+  Matcher.init(Blocks, BlendedHashes, CallHashes);
 
   // Index in yaml profile => corresponding (matched) block
   DenseMap<uint64_t, const FlowBlock *> MatchedBlocks;
   // Match blocks from the profile to the blocks in CFG
   for (const yaml::bolt::BinaryBasicBlockProfile &YamlBB : YamlBF.Blocks) {
     assert(YamlBB.Hash != 0 && "empty hash of BinaryBasicBlockProfile");
     BlendedBlockHash YamlHash(YamlBB.Hash);
-    const FlowBlock *MatchedBlock = Matcher.matchBlock(YamlHash);
-    // Always match the entry block.
+
+    const FlowBlock *MatchedBlock = nullptr;
+    std::string CallHashStr = hashBlockCalls(IdToYamlBF, YamlBB);
+    uint64_t CallHash = 0;
+    if (!CallHashStr.empty()) {
+      if (HashFunction == HashFunction::StdHash)
+        CallHash = std::hash<std::string>{}(CallHashStr);
+      else if (HashFunction == HashFunction::XXH3)
+        CallHash = llvm::xxh3_64bits(CallHashStr);
+      else
+        llvm_unreachable("Unhandled HashFunction");
+    }
+    MatchedBlock = Matcher.matchBlock(YamlHash, CallHash);
     if (MatchedBlock == nullptr && YamlBB.Index == 0)
       MatchedBlock = Blocks[0];
     if (MatchedBlock != nullptr) {
@@ -763,7 +827,8 @@ bool YAMLProfileReader::inferStaleProfile(
 
   // Match as many block/jump counts from the stale profile as possible
   size_t MatchedBlocks =
-      matchWeightsByHashes(BF.getBinaryContext(), BlockOrder, YamlBF, Func);
+      matchWeightsByHashes(BF.getBinaryContext(), BlockOrder, YamlBF, Func,
+                           YamlBP.Header.HashFunction, IdToYamLBF);
 
   // Adjust the flow function by marking unreachable blocks Unlikely so that
   // they don't get any counts assigned.

bolt/lib/Profile/YAMLProfileReader.cpp

@@ -611,6 +611,10 @@ Error YAMLProfileReader::readProfile(BinaryContext &BC) {
   NormalizeByInsnCount = usesEvent("cycles") || usesEvent("instructions");
   NormalizeByCalls = usesEvent("branches");
 
+  // Map profiled function ids to names.
+  for (yaml::bolt::BinaryFunctionProfile &YamlBF : YamlBP.Functions)
+    IdToYamLBF[YamlBF.Id] = &YamlBF;
+
   uint64_t NumUnused = 0;
   for (yaml::bolt::BinaryFunctionProfile &YamlBF : YamlBP.Functions) {
     if (YamlBF.Id >= YamlProfileToFunction.size()) {