diff --git a/bolt/include/bolt/Passes/MCF.h b/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

diff --git a/bolt/include/bolt/Passes/StokeInfo.h b/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";

diff --git a/bolt/include/bolt/Profile/BoltAddressTranslation.h b/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:

diff --git a/bolt/include/bolt/Profile/DataAggregator.h b/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

diff --git a/bolt/lib/Core/BinaryContext.cpp b/bolt/lib/Core/BinaryContext.cpp
@@ -142,7 +142,7 @@ BinaryContext::BinaryContext(std::unique_ptr<MCContext> Ctx,
       AsmInfo(std::move(AsmInfo)), MII(std::move(MII)), STI(std::move(STI)),
       InstPrinter(std::move(InstPrinter)), MIA(std::move(MIA)),
       MIB(std::move(MIB)), MRI(std::move(MRI)), DisAsm(std::move(DisAsm)),
-      Logger(Logger) {
+      Logger(Logger), InitialDynoStats(isAArch64()) {
   Relocation::Arch = this->TheTriple->getArch();
   RegularPageSize = isAArch64() ? RegularPageSizeAArch64 : RegularPageSizeX86;
   PageAlign = opts::NoHugePages ? RegularPageSize : HugePageSize;
@@ -934,10 +934,13 @@ std::string BinaryContext::generateJumpTableName(const BinaryFunction &BF,
   uint64_t Offset = 0;
   if (const JumpTable *JT = BF.getJumpTableContainingAddress(Address)) {
     Offset = Address - JT->getAddress();
-    auto Itr = JT->Labels.find(Offset);
-    if (Itr != JT->Labels.end())
-      return std::string(Itr->second->getName());
-    Id = JumpTableIds.at(JT->getAddress());
+    auto JTLabelsIt = JT->Labels.find(Offset);
+    if (JTLabelsIt != JT->Labels.end())
+      return std::string(JTLabelsIt->second->getName());
+
+    auto JTIdsIt = JumpTableIds.find(JT->getAddress());
+    assert(JTIdsIt != JumpTableIds.end());
+    Id = JTIdsIt->second;
   } else {
     Id = JumpTableIds[Address] = BF.JumpTables.size();
   }
@@ -1322,7 +1325,9 @@ void BinaryContext::processInterproceduralReferences() {
        InterproceduralReferences) {
     BinaryFunction &Function = *It.first;
     uint64_t Address = It.second;
-    if (!Address || Function.isIgnored())
+    // Process interprocedural references from ignored functions in BAT mode
+    // (non-simple in non-relocation mode) to properly register entry points
+    if (!Address || (Function.isIgnored() && !HasBATSection))
       continue;
 
     BinaryFunction *TargetFunction =
@@ -2212,8 +2217,8 @@ ErrorOr<uint64_t> BinaryContext::getUnsignedValueAtAddress(uint64_t Address,
   return DE.getUnsigned(&ValueOffset, Size);
 }
 
-ErrorOr<uint64_t> BinaryContext::getSignedValueAtAddress(uint64_t Address,
-                                                         size_t Size) const {
+ErrorOr<int64_t> BinaryContext::getSignedValueAtAddress(uint64_t Address,
+                                                        size_t Size) const {
   const ErrorOr<const BinarySection &> Section = getSectionForAddress(Address);
   if (!Section)
     return std::make_error_code(std::errc::bad_address);

diff --git a/bolt/lib/Core/BinaryEmitter.cpp b/bolt/lib/Core/BinaryEmitter.cpp
@@ -813,7 +813,9 @@ void BinaryEmitter::emitJumpTable(const JumpTable &JT, MCSection *HotSection,
   // determining its destination.
   std::map<MCSymbol *, uint64_t> LabelCounts;
   if (opts::JumpTables > JTS_SPLIT && !JT.Counts.empty()) {
-    MCSymbol *CurrentLabel = JT.Labels.at(0);
+    auto It = JT.Labels.find(0);
+    assert(It != JT.Labels.end());
+    MCSymbol *CurrentLabel = It->second;
     uint64_t CurrentLabelCount = 0;
     for (unsigned Index = 0; Index < JT.Entries.size(); ++Index) {
       auto LI = JT.Labels.find(Index * JT.EntrySize);

diff --git a/bolt/lib/Core/BinaryFunction.cpp b/bolt/lib/Core/BinaryFunction.cpp
@@ -851,15 +851,19 @@ BinaryFunction::processIndirectBranch(MCInst &Instruction, unsigned Size,
     return IndirectBranchType::UNKNOWN;
   }
 
-  // RIP-relative addressing should be converted to symbol form by now
-  // in processed instructions (but not in jump).
-  if (DispExpr) {
+  auto getExprValue = [&](const MCExpr *Expr) {
     const MCSymbol *TargetSym;
     uint64_t TargetOffset;
-    std::tie(TargetSym, TargetOffset) = BC.MIB->getTargetSymbolInfo(DispExpr);
+    std::tie(TargetSym, TargetOffset) = BC.MIB->getTargetSymbolInfo(Expr);
     ErrorOr<uint64_t> SymValueOrError = BC.getSymbolValue(*TargetSym);
-    assert(SymValueOrError && "global symbol needs a value");
-    ArrayStart = *SymValueOrError + TargetOffset;
+    assert(SymValueOrError && "Global symbol needs a value");
+    return *SymValueOrError + TargetOffset;
+  };
+
+  // RIP-relative addressing should be converted to symbol form by now
+  // in processed instructions (but not in jump).
+  if (DispExpr) {
+    ArrayStart = getExprValue(DispExpr);
     BaseRegNum = BC.MIB->getNoRegister();
     if (BC.isAArch64()) {
       ArrayStart &= ~0xFFFULL;
@@ -1666,7 +1670,8 @@ void BinaryFunction::postProcessEntryPoints() {
     // In non-relocation mode there's potentially an external undetectable
     // reference to the entry point and hence we cannot move this entry
     // point. Optimizing without moving could be difficult.
-    if (!BC.HasRelocations)
+    // In BAT mode, register any known entry points for CFG construction.
+    if (!BC.HasRelocations && !BC.HasBATSection)
       setSimple(false);
 
     const uint32_t Offset = KV.first;
@@ -3697,6 +3702,13 @@ BinaryFunction::BasicBlockListType BinaryFunction::dfs() const {
 
 size_t BinaryFunction::computeHash(bool UseDFS, HashFunction HashFunction,
                                    OperandHashFuncTy OperandHashFunc) const {
+  LLVM_DEBUG({
+    dbgs() << "BOLT-DEBUG: computeHash " << getPrintName() << ' '
+           << (UseDFS ? "dfs" : "bin") << " order "
+           << (HashFunction == HashFunction::StdHash ? "std::hash" : "xxh3")
+           << '\n';
+  });
+
   if (size() == 0)
     return 0;
 

diff --git a/bolt/lib/Core/DebugNames.cpp b/bolt/lib/Core/DebugNames.cpp
@@ -112,8 +112,6 @@ void DWARF5AcceleratorTable::addUnit(DWARFUnit &Unit,
 // Returns true if DW_TAG_variable should be included in .debug-names based on
 // section 6.1.1.1 for DWARF5 spec.
 static bool shouldIncludeVariable(const DWARFUnit &Unit, const DIE &Die) {
-  if (Die.findAttribute(dwarf::Attribute::DW_AT_declaration))
-    return false;
   const DIEValue LocAttrInfo =
       Die.findAttribute(dwarf::Attribute::DW_AT_location);
   if (!LocAttrInfo)
@@ -148,6 +146,8 @@ static bool shouldIncludeVariable(const DWARFUnit &Unit, const DIE &Die) {
 
 bool static canProcess(const DWARFUnit &Unit, const DIE &Die,
                        std::string &NameToUse, const bool TagsOnly) {
+  if (Die.findAttribute(dwarf::Attribute::DW_AT_declaration))
+    return false;
   switch (Die.getTag()) {
   case dwarf::DW_TAG_base_type:
   case dwarf::DW_TAG_class_type:

diff --git a/bolt/lib/Core/DynoStats.cpp b/bolt/lib/Core/DynoStats.cpp
@@ -114,8 +114,9 @@ void DynoStats::print(raw_ostream &OS, const DynoStats *Other,
     for (auto &Stat : llvm::reverse(SortedHistogram)) {
       OS << format("%20s,%'18lld", Printer->getOpcodeName(Stat.second).data(),
                    Stat.first * opts::DynoStatsScale);
-
-      MaxOpcodeHistogramTy MaxMultiMap = OpcodeHistogram.at(Stat.second).second;
+      auto It = OpcodeHistogram.find(Stat.second);
+      assert(It != OpcodeHistogram.end());
+      MaxOpcodeHistogramTy MaxMultiMap = It->second.second;
       // Start with function name:BB offset with highest execution count.
       for (auto &Max : llvm::reverse(MaxMultiMap)) {
         OS << format(", %'18lld, ", Max.first * opts::DynoStatsScale)

diff --git a/bolt/lib/Passes/BinaryFunctionCallGraph.cpp b/bolt/lib/Passes/BinaryFunctionCallGraph.cpp
@@ -56,7 +56,9 @@ std::deque<BinaryFunction *> BinaryFunctionCallGraph::buildTraversalOrder() {
   std::stack<NodeId> Worklist;
 
   for (BinaryFunction *Func : Funcs) {
-    const NodeId Id = FuncToNodeId.at(Func);
+    auto It = FuncToNodeId.find(Func);
+    assert(It != FuncToNodeId.end());
+    const NodeId Id = It->second;
     Worklist.push(Id);
     NodeStatus[Id] = NEW;
   }

diff --git a/bolt/lib/Passes/BinaryPasses.cpp b/bolt/lib/Passes/BinaryPasses.cpp
@@ -674,7 +674,8 @@ static uint64_t fixDoubleJumps(BinaryFunction &Function, bool MarkInvalid) {
   MCPlusBuilder *MIB = Function.getBinaryContext().MIB.get();
   for (BinaryBasicBlock &BB : Function) {
     auto checkAndPatch = [&](BinaryBasicBlock *Pred, BinaryBasicBlock *Succ,
-                             const MCSymbol *SuccSym) {
+                             const MCSymbol *SuccSym,
+                             std::optional<uint32_t> Offset) {
       // Ignore infinite loop jumps or fallthrough tail jumps.
       if (Pred == Succ || Succ == &BB)
         return false;
@@ -715,9 +716,11 @@ static uint64_t fixDoubleJumps(BinaryFunction &Function, bool MarkInvalid) {
           Pred->removeSuccessor(&BB);
           Pred->eraseInstruction(Pred->findInstruction(Branch));
           Pred->addTailCallInstruction(SuccSym);
-          MCInst *TailCall = Pred->getLastNonPseudoInstr();
-          assert(TailCall);
-          MIB->setOffset(*TailCall, BB.getOffset());
+          if (Offset) {
+            MCInst *TailCall = Pred->getLastNonPseudoInstr();
+            assert(TailCall);
+            MIB->setOffset(*TailCall, *Offset);
+          }
         } else {
           return false;
         }
@@ -760,7 +763,8 @@ static uint64_t fixDoubleJumps(BinaryFunction &Function, bool MarkInvalid) {
       if (Pred->getSuccessor() == &BB ||
           (Pred->getConditionalSuccessor(true) == &BB && !IsTailCall) ||
           Pred->getConditionalSuccessor(false) == &BB)
-        if (checkAndPatch(Pred, Succ, SuccSym) && MarkInvalid)
+        if (checkAndPatch(Pred, Succ, SuccSym, MIB->getOffset(*Inst)) &&
+            MarkInvalid)
           BB.markValid(BB.pred_size() != 0 || BB.isLandingPad() ||
                        BB.isEntryPoint());
     }
@@ -1386,9 +1390,19 @@ Error PrintProgramStats::runOnFunctions(BinaryContext &BC) {
     if (Function.isPLTFunction())
       continue;
 
+    // Adjustment for BAT mode: the profile for BOLT split fragments is combined
+    // so only count the hot fragment.
+    const uint64_t Address = Function.getAddress();
+    bool IsHotParentOfBOLTSplitFunction = !Function.getFragments().empty() &&
+                                          BAT && BAT->isBATFunction(Address) &&
+                                          !BAT->fetchParentAddress(Address);
+
     ++NumRegularFunctions;
 
-    if (!Function.isSimple()) {
+    // In BOLTed binaries split functions are non-simple (due to non-relocation
+    // mode), but the original function is known to be simple and we have a
+    // valid profile for it.
+    if (!Function.isSimple() && !IsHotParentOfBOLTSplitFunction) {
       if (Function.hasProfile())
         ++NumNonSimpleProfiledFunctions;
       continue;
@@ -1549,23 +1563,28 @@ Error PrintProgramStats::runOnFunctions(BinaryContext &BC) {
     const bool Ascending =
         opts::DynoStatsSortOrderOpt == opts::DynoStatsSortOrder::Ascending;
 
-    if (SortAll) {
-      llvm::stable_sort(Functions,
-                        [Ascending, &Stats](const BinaryFunction *A,
-                                            const BinaryFunction *B) {
-                          return Ascending ? Stats.at(A) < Stats.at(B)
-                                           : Stats.at(B) < Stats.at(A);
-                        });
-    } else {
-      llvm::stable_sort(
-          Functions, [Ascending, &Stats](const BinaryFunction *A,
-                                         const BinaryFunction *B) {
-            const DynoStats &StatsA = Stats.at(A);
-            const DynoStats &StatsB = Stats.at(B);
-            return Ascending ? StatsA.lessThan(StatsB, opts::PrintSortedBy)
-                             : StatsB.lessThan(StatsA, opts::PrintSortedBy);
-          });
-    }
+    std::function<bool(const DynoStats &, const DynoStats &)>
+        DynoStatsComparator =
+            SortAll ? [](const DynoStats &StatsA,
+                         const DynoStats &StatsB) { return StatsA < StatsB; }
+                    : [](const DynoStats &StatsA, const DynoStats &StatsB) {
+                        return StatsA.lessThan(StatsB, opts::PrintSortedBy);
+                      };
+
+    llvm::stable_sort(Functions,
+                      [Ascending, &Stats, DynoStatsComparator](
+                          const BinaryFunction *A, const BinaryFunction *B) {
+                        auto StatsItr = Stats.find(A);
+                        assert(StatsItr != Stats.end());
+                        const DynoStats &StatsA = StatsItr->second;
+
+                        StatsItr = Stats.find(B);
+                        assert(StatsItr != Stats.end());
+                        const DynoStats &StatsB = StatsItr->second;
+
+                        return Ascending ? DynoStatsComparator(StatsA, StatsB)
+                                         : DynoStatsComparator(StatsB, StatsA);
+                      });
 
     BC.outs() << "BOLT-INFO: top functions sorted by ";
     if (SortAll) {

diff --git a/bolt/lib/Passes/CacheMetrics.cpp b/bolt/lib/Passes/CacheMetrics.cpp
@@ -67,7 +67,20 @@ calcTSPScore(const std::vector<BinaryFunction *> &BinaryFunctions,
       for (BinaryBasicBlock *DstBB : SrcBB->successors()) {
         if (SrcBB != DstBB && BI->Count != BinaryBasicBlock::COUNT_NO_PROFILE) {
           JumpCount += BI->Count;
-          if (BBAddr.at(SrcBB) + BBSize.at(SrcBB) == BBAddr.at(DstBB))
+
+          auto BBAddrIt = BBAddr.find(SrcBB);
+          assert(BBAddrIt != BBAddr.end());
+          uint64_t SrcBBAddr = BBAddrIt->second;
+
+          auto BBSizeIt = BBSize.find(SrcBB);
+          assert(BBSizeIt != BBSize.end());
+          uint64_t SrcBBSize = BBSizeIt->second;
+
+          BBAddrIt = BBAddr.find(DstBB);
+          assert(BBAddrIt != BBAddr.end());
+          uint64_t DstBBAddr = BBAddrIt->second;
+
+          if (SrcBBAddr + SrcBBSize == DstBBAddr)
             Score += BI->Count;
         }
         ++BI;
@@ -149,29 +162,39 @@ double expectedCacheHitRatio(
   for (BinaryFunction *BF : BinaryFunctions) {
     if (BF->getLayout().block_empty())
       continue;
-    const uint64_t Page =
-        BBAddr.at(BF->getLayout().block_front()) / ITLBPageSize;
-    PageSamples[Page] += FunctionSamples.at(BF);
+    auto BBAddrIt = BBAddr.find(BF->getLayout().block_front());
+    assert(BBAddrIt != BBAddr.end());
+    const uint64_t Page = BBAddrIt->second / ITLBPageSize;
+
+    auto FunctionSamplesIt = FunctionSamples.find(BF);
+    assert(FunctionSamplesIt != FunctionSamples.end());
+    PageSamples[Page] += FunctionSamplesIt->second;
   }
 
   // Computing the expected number of misses for every function
   double Misses = 0;
   for (BinaryFunction *BF : BinaryFunctions) {
     // Skip the function if it has no samples
-    if (BF->getLayout().block_empty() || FunctionSamples.at(BF) == 0.0)
+    auto FunctionSamplesIt = FunctionSamples.find(BF);
+    assert(FunctionSamplesIt != FunctionSamples.end());
+    double Samples = FunctionSamplesIt->second;
+    if (BF->getLayout().block_empty() || Samples == 0.0)
       continue;
-    double Samples = FunctionSamples.at(BF);
-    const uint64_t Page =
-        BBAddr.at(BF->getLayout().block_front()) / ITLBPageSize;
+
+    auto BBAddrIt = BBAddr.find(BF->getLayout().block_front());
+    assert(BBAddrIt != BBAddr.end());
+    const uint64_t Page = BBAddrIt->second / ITLBPageSize;
     // The probability that the page is not present in the cache
     const double MissProb =
         pow(1.0 - PageSamples[Page] / TotalSamples, ITLBEntries);
 
     // Processing all callers of the function
     for (std::pair<BinaryFunction *, uint64_t> Pair : Calls[BF]) {
       BinaryFunction *SrcFunction = Pair.first;
-      const uint64_t SrcPage =
-          BBAddr.at(SrcFunction->getLayout().block_front()) / ITLBPageSize;
+
+      BBAddrIt = BBAddr.find(SrcFunction->getLayout().block_front());
+      assert(BBAddrIt != BBAddr.end());
+      const uint64_t SrcPage = BBAddrIt->second / ITLBPageSize;
       // Is this a 'long' or a 'short' call?
       if (Page != SrcPage) {
         // This is a miss

diff --git a/bolt/lib/Passes/Inliner.cpp b/bolt/lib/Passes/Inliner.cpp
@@ -355,7 +355,9 @@ Inliner::inlineCall(BinaryBasicBlock &CallerBB,
     std::vector<BinaryBasicBlock *> Successors(BB.succ_size());
     llvm::transform(BB.successors(), Successors.begin(),
                     [&InlinedBBMap](const BinaryBasicBlock *BB) {
-                      return InlinedBBMap.at(BB);
+                      auto It = InlinedBBMap.find(BB);
+                      assert(It != InlinedBBMap.end());
+                      return It->second;
                     });
 
     if (CallerFunction.hasValidProfile() && Callee.hasValidProfile())

diff --git a/bolt/lib/Passes/MCF.cpp b/bolt/lib/Passes/MCF.cpp
@@ -12,9 +12,11 @@
 
 #include "bolt/Passes/MCF.h"
 #include "bolt/Core/BinaryFunction.h"
+#include "bolt/Core/ParallelUtilities.h"
 #include "bolt/Passes/DataflowInfoManager.h"
 #include "bolt/Utils/CommandLineOpts.h"
 #include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/STLExtras.h"
 #include "llvm/Support/CommandLine.h"
 #include <algorithm>
 #include <vector>
@@ -29,19 +31,10 @@ namespace opts {
 
 extern cl::OptionCategory BoltOptCategory;
 
-extern cl::opt<bool> TimeOpts;
-
 static cl::opt<bool> IterativeGuess(
     "iterative-guess",
     cl::desc("in non-LBR mode, guess edge counts using iterative technique"),
     cl::Hidden, cl::cat(BoltOptCategory));
-
-static cl::opt<bool> UseRArcs(
-    "mcf-use-rarcs",
-    cl::desc("in MCF, consider the possibility of cancelling flow to balance "
-             "edges"),
-    cl::Hidden, cl::cat(BoltOptCategory));
-
 } // namespace opts
 
 namespace llvm {
@@ -441,7 +434,7 @@ void equalizeBBCounts(DataflowInfoManager &Info, BinaryFunction &BF) {
   }
 }
 
-void estimateEdgeCounts(BinaryFunction &BF) {
+void EstimateEdgeCounts::runOnFunction(BinaryFunction &BF) {
   EdgeWeightMap PredEdgeWeights;
   EdgeWeightMap SuccEdgeWeights;
   if (!opts::IterativeGuess) {
@@ -462,8 +455,24 @@ void estimateEdgeCounts(BinaryFunction &BF) {
   recalculateBBCounts(BF, /*AllEdges=*/false);
 }
 
-void solveMCF(BinaryFunction &BF, MCFCostFunction CostFunction) {
-  llvm_unreachable("not implemented");
+Error EstimateEdgeCounts::runOnFunctions(BinaryContext &BC) {
+  if (llvm::none_of(llvm::make_second_range(BC.getBinaryFunctions()),
+                    [](const BinaryFunction &BF) {
+                      return BF.getProfileFlags() == BinaryFunction::PF_SAMPLE;
+                    }))
+    return Error::success();
+
+  ParallelUtilities::WorkFuncTy WorkFun = [&](BinaryFunction &BF) {
+    runOnFunction(BF);
+  };
+  ParallelUtilities::PredicateTy SkipFunc = [&](const BinaryFunction &BF) {
+    return BF.getProfileFlags() != BinaryFunction::PF_SAMPLE;
+  };
+
+  ParallelUtilities::runOnEachFunction(
+      BC, ParallelUtilities::SchedulingPolicy::SP_BB_QUADRATIC, WorkFun,
+      SkipFunc, "EstimateEdgeCounts");
+  return Error::success();
 }
 
 } // namespace bolt

diff --git a/bolt/lib/Profile/BoltAddressTranslation.cpp b/bolt/lib/Profile/BoltAddressTranslation.cpp
@@ -20,10 +20,9 @@ namespace bolt {
 
 const char *BoltAddressTranslation::SECTION_NAME = ".note.bolt_bat";
 
-void BoltAddressTranslation::writeEntriesForBB(MapTy &Map,
-                                               const BinaryBasicBlock &BB,
-                                               uint64_t FuncInputAddress,
-                                               uint64_t FuncOutputAddress) {
+void BoltAddressTranslation::writeEntriesForBB(
+    MapTy &Map, const BinaryBasicBlock &BB, uint64_t FuncInputAddress,
+    uint64_t FuncOutputAddress) const {
   const uint64_t BBOutputOffset =
       BB.getOutputAddressRange().first - FuncOutputAddress;
   const uint32_t BBInputOffset = BB.getInputOffset();
@@ -55,7 +54,7 @@ void BoltAddressTranslation::writeEntriesForBB(MapTy &Map,
   // and this deleted block will both share the same output address (the same
   // key), and we need to map back. We choose here to privilege the successor by
   // allowing it to overwrite the previously inserted key in the map.
-  Map[BBOutputOffset] = BBInputOffset << 1;
+  Map.emplace(BBOutputOffset, BBInputOffset << 1);
 
   const auto &IOAddressMap =
       BB.getFunction()->getBinaryContext().getIOAddressMap();
@@ -72,8 +71,7 @@ void BoltAddressTranslation::writeEntriesForBB(MapTy &Map,
 
     LLVM_DEBUG(dbgs() << "  Key: " << Twine::utohexstr(OutputOffset) << " Val: "
                       << Twine::utohexstr(InputOffset) << " (branch)\n");
-    Map.insert(std::pair<uint32_t, uint32_t>(OutputOffset,
-                                             (InputOffset << 1) | BRANCHENTRY));
+    Map.emplace(OutputOffset, (InputOffset << 1) | BRANCHENTRY);
   }
 }
 
@@ -108,6 +106,19 @@ void BoltAddressTranslation::write(const BinaryContext &BC, raw_ostream &OS) {
     for (const BinaryBasicBlock *const BB :
          Function.getLayout().getMainFragment())
       writeEntriesForBB(Map, *BB, InputAddress, OutputAddress);
+    // Add entries for deleted blocks. They are still required for correct BB
+    // mapping of branches modified by SCTC. By convention, they would have the
+    // end of the function as output address.
+    const BBHashMapTy &BBHashMap = getBBHashMap(InputAddress);
+    if (BBHashMap.size() != Function.size()) {
+      const uint64_t EndOffset = Function.getOutputSize();
+      std::unordered_set<uint32_t> MappedInputOffsets;
+      for (const BinaryBasicBlock &BB : Function)
+        MappedInputOffsets.emplace(BB.getInputOffset());
+      for (const auto &[InputOffset, _] : BBHashMap)
+        if (!llvm::is_contained(MappedInputOffsets, InputOffset))
+          Map.emplace(EndOffset, InputOffset << 1);
+    }
     Maps.emplace(Function.getOutputAddress(), std::move(Map));
     ReverseMap.emplace(OutputAddress, InputAddress);
 
@@ -138,8 +149,8 @@ void BoltAddressTranslation::write(const BinaryContext &BC, raw_ostream &OS) {
             << " basic block hashes\n";
 }
 
-APInt BoltAddressTranslation::calculateBranchEntriesBitMask(MapTy &Map,
-                                                            size_t EqualElems) {
+APInt BoltAddressTranslation::calculateBranchEntriesBitMask(
+    MapTy &Map, size_t EqualElems) const {
   APInt BitMask(alignTo(EqualElems, 8), 0);
   size_t Index = 0;
   for (std::pair<const uint32_t, uint32_t> &KeyVal : Map) {
@@ -422,7 +433,7 @@ void BoltAddressTranslation::parseMaps(std::vector<uint64_t> &HotFuncs,
   }
 }
 
-void BoltAddressTranslation::dump(raw_ostream &OS) {
+void BoltAddressTranslation::dump(raw_ostream &OS) const {
   const size_t NumTables = Maps.size();
   OS << "BAT tables for " << NumTables << " functions:\n";
   for (const auto &MapEntry : Maps) {
@@ -447,11 +458,15 @@ void BoltAddressTranslation::dump(raw_ostream &OS) {
         OS << formatv(" hash: {0:x}", BBHashMap.getBBHash(Val));
       OS << "\n";
     }
-    if (IsHotFunction)
-      OS << "NumBlocks: " << NumBasicBlocksMap[Address] << '\n';
-    if (SecondaryEntryPointsMap.count(Address)) {
+    if (IsHotFunction) {
+      auto NumBasicBlocksIt = NumBasicBlocksMap.find(Address);
+      assert(NumBasicBlocksIt != NumBasicBlocksMap.end());
+      OS << "NumBlocks: " << NumBasicBlocksIt->second << '\n';
+    }
+    auto SecondaryEntryPointsIt = SecondaryEntryPointsMap.find(Address);
+    if (SecondaryEntryPointsIt != SecondaryEntryPointsMap.end()) {
       const std::vector<uint32_t> &SecondaryEntryPoints =
-          SecondaryEntryPointsMap[Address];
+          SecondaryEntryPointsIt->second;
       OS << SecondaryEntryPoints.size() << " secondary entry points:\n";
       for (uint32_t EntryPointOffset : SecondaryEntryPoints)
         OS << formatv("{0:x}\n", EntryPointOffset);
@@ -547,13 +562,6 @@ BoltAddressTranslation::getFallthroughsInTrace(uint64_t FuncAddress,
   return Res;
 }
 
-uint64_t BoltAddressTranslation::fetchParentAddress(uint64_t Address) const {
-  auto Iter = ColdPartSource.find(Address);
-  if (Iter == ColdPartSource.end())
-    return 0;
-  return Iter->second;
-}
-
 bool BoltAddressTranslation::enabledFor(
     llvm::object::ELFObjectFileBase *InputFile) const {
   for (const SectionRef &Section : InputFile->sections()) {

diff --git a/bolt/lib/Profile/CMakeLists.txt b/bolt/lib/Profile/CMakeLists.txt
@@ -17,6 +17,5 @@ add_llvm_library(LLVMBOLTProfile
 target_link_libraries(LLVMBOLTProfile
   PRIVATE
   LLVMBOLTCore
-  LLVMBOLTPasses
   LLVMBOLTUtils
   )
diff --git a/bolt/lib/Profile/DataAggregator.cpp b/bolt/lib/Profile/DataAggregator.cpp
@@ -613,7 +613,6 @@ Error DataAggregator::readProfile(BinaryContext &BC) {
         if (std::error_code EC = writeBATYAML(BC, opts::SaveProfile))
           report_error("cannot create output data file", EC);
     }
-    BC.logBOLTErrorsAndQuitOnFatal(PrintProgramStats().runOnFunctions(BC));
   }
 
   return Error::success();
@@ -673,7 +672,8 @@ DataAggregator::getBATParentFunction(const BinaryFunction &Func) const {
   return nullptr;
 }
 
-StringRef DataAggregator::getLocationName(const BinaryFunction &Func) const {
+StringRef DataAggregator::getLocationName(const BinaryFunction &Func,
+                                          bool BAT) {
   if (!BAT)
     return Func.getOneName();
 
@@ -702,7 +702,7 @@ bool DataAggregator::doSample(BinaryFunction &OrigFunc, uint64_t Address,
   auto I = NamesToSamples.find(Func.getOneName());
   if (I == NamesToSamples.end()) {
     bool Success;
-    StringRef LocName = getLocationName(Func);
+    StringRef LocName = getLocationName(Func, BAT);
     std::tie(I, Success) = NamesToSamples.insert(
         std::make_pair(Func.getOneName(),
                        FuncSampleData(LocName, FuncSampleData::ContainerTy())));
@@ -722,7 +722,7 @@ bool DataAggregator::doIntraBranch(BinaryFunction &Func, uint64_t From,
   FuncBranchData *AggrData = getBranchData(Func);
   if (!AggrData) {
     AggrData = &NamesToBranches[Func.getOneName()];
-    AggrData->Name = getLocationName(Func);
+    AggrData->Name = getLocationName(Func, BAT);
     setBranchData(Func, AggrData);
   }
 
@@ -741,7 +741,7 @@ bool DataAggregator::doInterBranch(BinaryFunction *FromFunc,
   StringRef SrcFunc;
   StringRef DstFunc;
   if (FromFunc) {
-    SrcFunc = getLocationName(*FromFunc);
+    SrcFunc = getLocationName(*FromFunc, BAT);
     FromAggrData = getBranchData(*FromFunc);
     if (!FromAggrData) {
       FromAggrData = &NamesToBranches[FromFunc->getOneName()];
@@ -752,7 +752,7 @@ bool DataAggregator::doInterBranch(BinaryFunction *FromFunc,
     recordExit(*FromFunc, From, Mispreds, Count);
   }
   if (ToFunc) {
-    DstFunc = getLocationName(*ToFunc);
+    DstFunc = getLocationName(*ToFunc, BAT);
     ToAggrData = getBranchData(*ToFunc);
     if (!ToAggrData) {
       ToAggrData = &NamesToBranches[ToFunc->getOneName()];
@@ -1227,7 +1227,7 @@ ErrorOr<Location> DataAggregator::parseLocationOrOffset() {
   if (Sep == StringRef::npos)
     return parseOffset();
   StringRef LookAhead = ParsingBuf.substr(0, Sep);
-  if (LookAhead.find_first_of(":") == StringRef::npos)
+  if (!LookAhead.contains(':'))
     return parseOffset();
 
   ErrorOr<StringRef> BuildID = parseString(':');
@@ -2340,7 +2340,7 @@ std::error_code DataAggregator::writeBATYAML(BinaryContext &BC,
         continue;
       BinaryFunction *BF = BC.getBinaryFunctionAtAddress(FuncAddress);
       assert(BF);
-      YamlBF.Name = getLocationName(*BF);
+      YamlBF.Name = getLocationName(*BF, BAT);
       YamlBF.Id = BF->getFunctionNumber();
       YamlBF.Hash = BAT->getBFHash(FuncAddress);
       YamlBF.ExecCount = BF->getKnownExecutionCount();
@@ -2349,11 +2349,11 @@ std::error_code DataAggregator::writeBATYAML(BinaryContext &BC,
           BAT->getBBHashMap(FuncAddress);
       YamlBF.Blocks.resize(YamlBF.NumBasicBlocks);
 
-      for (auto &&[Idx, YamlBB] : llvm::enumerate(YamlBF.Blocks))
-        YamlBB.Index = Idx;
-
-      for (auto BI = BlockMap.begin(), BE = BlockMap.end(); BI != BE; ++BI)
-        YamlBF.Blocks[BI->second.getBBIndex()].Hash = BI->second.getBBHash();
+      for (auto &&[Entry, YamlBB] : llvm::zip(BlockMap, YamlBF.Blocks)) {
+        const auto &Block = Entry.second;
+        YamlBB.Hash = Block.Hash;
+        YamlBB.Index = Block.Index;
+      }
 
       // Lookup containing basic block offset and index
       auto getBlock = [&BlockMap](uint32_t Offset) {
@@ -2363,7 +2363,7 @@ std::error_code DataAggregator::writeBATYAML(BinaryContext &BC,
           exit(1);
         }
         --BlockIt;
-        return std::pair(BlockIt->first, BlockIt->second.getBBIndex());
+        return std::pair(BlockIt->first, BlockIt->second.Index);
       };
 
       for (const BranchInfo &BI : Branches.Data) {

diff --git a/bolt/lib/Profile/DataReader.cpp b/bolt/lib/Profile/DataReader.cpp
@@ -598,8 +598,6 @@ void DataReader::readSampleData(BinaryFunction &BF) {
   }
 
   BF.ExecutionCount = TotalEntryCount;
-
-  estimateEdgeCounts(BF);
 }
 
 void DataReader::convertBranchData(BinaryFunction &BF) const {

diff --git a/bolt/lib/Profile/StaleProfileMatching.cpp b/bolt/lib/Profile/StaleProfileMatching.cpp
@@ -30,6 +30,7 @@
 #include "llvm/ADT/Bitfields.h"
 #include "llvm/ADT/Hashing.h"
 #include "llvm/Support/CommandLine.h"
+#include "llvm/Support/Timer.h"
 #include "llvm/Support/xxhash.h"
 #include "llvm/Transforms/Utils/SampleProfileInference.h"
 
@@ -42,6 +43,7 @@ using namespace llvm;
 
 namespace opts {
 
+extern cl::opt<bool> TimeRewrite;
 extern cl::OptionCategory BoltOptCategory;
 
 cl::opt<bool>
@@ -372,8 +374,10 @@ createFlowFunction(const BinaryFunction::BasicBlockOrderType &BlockOrder) {
 
   // Create necessary metadata for the flow function
   for (FlowJump &Jump : Func.Jumps) {
-    Func.Blocks.at(Jump.Source).SuccJumps.push_back(&Jump);
-    Func.Blocks.at(Jump.Target).PredJumps.push_back(&Jump);
+    assert(Jump.Source < Func.Blocks.size());
+    Func.Blocks[Jump.Source].SuccJumps.push_back(&Jump);
+    assert(Jump.Target < Func.Blocks.size());
+    Func.Blocks[Jump.Target].PredJumps.push_back(&Jump);
   }
   return Func;
 }
@@ -705,6 +709,10 @@ void assignProfile(BinaryFunction &BF,
 
 bool YAMLProfileReader::inferStaleProfile(
     BinaryFunction &BF, const yaml::bolt::BinaryFunctionProfile &YamlBF) {
+
+  NamedRegionTimer T("inferStaleProfile", "stale profile inference", "rewrite",
+                     "Rewrite passes", opts::TimeRewrite);
+
   if (!BF.hasCFG())
     return false;
 

diff --git a/bolt/lib/Profile/YAMLProfileReader.cpp b/bolt/lib/Profile/YAMLProfileReader.cpp
@@ -99,11 +99,17 @@ bool YAMLProfileReader::parseFunctionProfile(
       FuncRawBranchCount += YamlSI.Count;
   BF.setRawBranchCount(FuncRawBranchCount);
 
-  if (!opts::IgnoreHash &&
-      YamlBF.Hash != BF.computeHash(IsDFSOrder, HashFunction)) {
-    if (opts::Verbosity >= 1)
-      errs() << "BOLT-WARNING: function hash mismatch\n";
-    ProfileMatched = false;
+  if (BF.empty())
+    return true;
+
+  if (!opts::IgnoreHash) {
+    if (!BF.getHash())
+      BF.computeHash(IsDFSOrder, HashFunction);
+    if (YamlBF.Hash != BF.getHash()) {
+      if (opts::Verbosity >= 1)
+        errs() << "BOLT-WARNING: function hash mismatch\n";
+      ProfileMatched = false;
+    }
   }
 
   if (YamlBF.NumBasicBlocks != BF.size()) {
@@ -250,10 +256,8 @@ bool YAMLProfileReader::parseFunctionProfile(
     if (BB.getExecutionCount() == BinaryBasicBlock::COUNT_NO_PROFILE)
       BB.setExecutionCount(0);
 
-  if (YamlBP.Header.Flags & BinaryFunction::PF_SAMPLE) {
+  if (YamlBP.Header.Flags & BinaryFunction::PF_SAMPLE)
     BF.setExecutionCount(FunctionExecutionCount);
-    estimateEdgeCounts(BF);
-  }
 
   ProfileMatched &= !MismatchedBlocks && !MismatchedCalls && !MismatchedEdges;
 

diff --git a/bolt/lib/Profile/YAMLProfileWriter.cpp b/bolt/lib/Profile/YAMLProfileWriter.cpp
@@ -10,6 +10,7 @@
 #include "bolt/Core/BinaryBasicBlock.h"
 #include "bolt/Core/BinaryFunction.h"
 #include "bolt/Profile/BoltAddressTranslation.h"
+#include "bolt/Profile/DataAggregator.h"
 #include "bolt/Profile/ProfileReaderBase.h"
 #include "bolt/Rewrite/RewriteInstance.h"
 #include "llvm/Support/CommandLine.h"
@@ -39,6 +40,10 @@ const BinaryFunction *YAMLProfileWriter::setCSIDestination(
             BC.getFunctionForSymbol(Symbol, &EntryID)) {
       if (BAT && BAT->isBATFunction(Callee->getAddress()))
         std::tie(Callee, EntryID) = BAT->translateSymbol(BC, *Symbol, Offset);
+      else if (const BinaryBasicBlock *BB =
+                   Callee->getBasicBlockContainingOffset(Offset))
+        BC.getFunctionForSymbol(Callee->getSecondaryEntryPointSymbol(*BB),
+                                &EntryID);
       CSI.DestId = Callee->getFunctionNumber();
       CSI.EntryDiscriminator = EntryID;
       return Callee;
@@ -59,7 +64,7 @@ YAMLProfileWriter::convert(const BinaryFunction &BF, bool UseDFS,
   BF.computeHash(UseDFS);
   BF.computeBlockHashes();
 
-  YamlBF.Name = BF.getPrintName();
+  YamlBF.Name = DataAggregator::getLocationName(BF, BAT);
   YamlBF.Id = BF.getFunctionNumber();
   YamlBF.Hash = BF.getHash();
   YamlBF.NumBasicBlocks = BF.size();

diff --git a/bolt/lib/Rewrite/BinaryPassManager.cpp b/bolt/lib/Rewrite/BinaryPassManager.cpp
@@ -23,6 +23,7 @@
 #include "bolt/Passes/JTFootprintReduction.h"
 #include "bolt/Passes/LongJmp.h"
 #include "bolt/Passes/LoopInversionPass.h"
+#include "bolt/Passes/MCF.h"
 #include "bolt/Passes/PLTCall.h"
 #include "bolt/Passes/PatchEntries.h"
 #include "bolt/Passes/RegReAssign.h"
@@ -90,6 +91,11 @@ PrintAfterLowering("print-after-lowering",
   cl::desc("print function after instruction lowering"),
   cl::Hidden, cl::cat(BoltOptCategory));
 
+static cl::opt<bool> PrintEstimateEdgeCounts(
+    "print-estimate-edge-counts",
+    cl::desc("print function after edge counts are set for no-LBR profile"),
+    cl::Hidden, cl::cat(BoltOptCategory));
+
 cl::opt<bool>
 PrintFinalized("print-finalized",
   cl::desc("print function after CFG is finalized"),
@@ -334,8 +340,10 @@ Error BinaryFunctionPassManager::runPasses() {
 Error BinaryFunctionPassManager::runAllPasses(BinaryContext &BC) {
   BinaryFunctionPassManager Manager(BC);
 
-  const DynoStats InitialDynoStats =
-      getDynoStats(BC.getBinaryFunctions(), BC.isAArch64());
+  Manager.registerPass(
+      std::make_unique<EstimateEdgeCounts>(PrintEstimateEdgeCounts));
+
+  Manager.registerPass(std::make_unique<DynoStatsSetPass>());
 
   Manager.registerPass(std::make_unique<AsmDumpPass>(),
                        opts::AsmDump.getNumOccurrences());
@@ -447,10 +455,9 @@ Error BinaryFunctionPassManager::runAllPasses(BinaryContext &BC) {
   Manager.registerPass(std::make_unique<SplitFunctions>(PrintSplit));
 
   // Print final dyno stats right while CFG and instruction analysis are intact.
-  Manager.registerPass(
-      std::make_unique<DynoStatsPrintPass>(
-          InitialDynoStats, "after all optimizations before SCTC and FOP"),
-      opts::PrintDynoStats || opts::DynoStatsAll);
+  Manager.registerPass(std::make_unique<DynoStatsPrintPass>(
+                           "after all optimizations before SCTC and FOP"),
+                       opts::PrintDynoStats || opts::DynoStatsAll);
 
   // Add the StokeInfo pass, which extract functions for stoke optimization and
   // get the liveness information for them

diff --git a/bolt/lib/Rewrite/DWARFRewriter.cpp b/bolt/lib/Rewrite/DWARFRewriter.cpp
@@ -73,8 +73,7 @@ static void printDie(DWARFUnit &DU, uint64_t DIEOffset) {
   DWARFDataExtractor DebugInfoData = DU.getDebugInfoExtractor();
   DWARFDebugInfoEntry DIEEntry;
   if (DIEEntry.extractFast(DU, &DIEOffset, DebugInfoData, NextCUOffset, 0)) {
-    if (const DWARFAbbreviationDeclaration *AbbrDecl =
-            DIEEntry.getAbbreviationDeclarationPtr()) {
+    if (DIEEntry.getAbbreviationDeclarationPtr()) {
       DWARFDie DDie(&DU, &DIEEntry);
       printDie(DDie);
     } else {

diff --git a/bolt/lib/Rewrite/LinuxKernelRewriter.cpp b/bolt/lib/Rewrite/LinuxKernelRewriter.cpp
@@ -393,7 +393,7 @@ void LinuxKernelRewriter::processLKKSymtab(bool IsGPL) {
 
   for (uint64_t I = 0; I < SectionSize; I += 4) {
     const uint64_t EntryAddress = SectionAddress + I;
-    ErrorOr<uint64_t> Offset = BC.getSignedValueAtAddress(EntryAddress, 4);
+    ErrorOr<int64_t> Offset = BC.getSignedValueAtAddress(EntryAddress, 4);
     assert(Offset && "Reading valid PC-relative offset for a ksymtab entry");
     const int32_t SignedOffset = *Offset;
     const uint64_t RefAddress = EntryAddress + SignedOffset;

diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -17,6 +17,7 @@
 #include "bolt/Core/MCPlusBuilder.h"
 #include "bolt/Core/ParallelUtilities.h"
 #include "bolt/Core/Relocation.h"
+#include "bolt/Passes/BinaryPasses.h"
 #include "bolt/Passes/CacheMetrics.h"
 #include "bolt/Passes/ReorderFunctions.h"
 #include "bolt/Profile/BoltAddressTranslation.h"
@@ -86,6 +87,7 @@ extern cl::list<std::string> ReorderData;
 extern cl::opt<bolt::ReorderFunctions::ReorderType> ReorderFunctions;
 extern cl::opt<bool> TerminalTrap;
 extern cl::opt<bool> TimeBuild;
+extern cl::opt<bool> TimeRewrite;
 
 cl::opt<bool> AllowStripped("allow-stripped",
                             cl::desc("allow processing of stripped binaries"),
@@ -235,11 +237,6 @@ UseGnuStack("use-gnu-stack",
   cl::ZeroOrMore,
   cl::cat(BoltCategory));
 
-static cl::opt<bool>
-    TimeRewrite("time-rewrite",
-                cl::desc("print time spent in rewriting passes"), cl::Hidden,
-                cl::cat(BoltCategory));
-
 static cl::opt<bool>
 SequentialDisassembly("sequential-disassembly",
   cl::desc("performs disassembly sequentially"),
@@ -1500,7 +1497,7 @@ void RewriteInstance::registerFragments() {
   if (!BC->hasSymbolsWithFileName()) {
     BC->errs() << "BOLT-ERROR: input file has split functions but does not "
                   "have FILE symbols. If the binary was stripped, preserve "
-                  "FILE symbols with --keep-file-symbols strip option";
+                  "FILE symbols with --keep-file-symbols strip option\n";
     exit(1);
   }
 
@@ -1988,6 +1985,7 @@ Error RewriteInstance::readSpecialSections() {
 
   if (ErrorOr<BinarySection &> BATSec =
           BC->getUniqueSectionByName(BoltAddressTranslation::SECTION_NAME)) {
+    BC->HasBATSection = true;
     // Do not read BAT when plotting a heatmap
     if (!opts::HeatmapMode) {
       if (std::error_code EC = BAT->parse(BC->outs(), BATSec->getContents())) {
@@ -3208,12 +3206,14 @@ void RewriteInstance::preprocessProfileData() {
   if (Error E = ProfileReader->preprocessProfile(*BC.get()))
     report_error("cannot pre-process profile", std::move(E));
 
-  if (!BC->hasSymbolsWithFileName() && ProfileReader->hasLocalsWithFileName()) {
+  if (!BC->hasSymbolsWithFileName() && ProfileReader->hasLocalsWithFileName() &&
+      !opts::AllowStripped) {
     BC->errs()
         << "BOLT-ERROR: input binary does not have local file symbols "
            "but profile data includes function names with embedded file "
            "names. It appears that the input binary was stripped while a "
-           "profiled binary was not\n";
+           "profiled binary was not. If you know what you are doing and "
+           "wish to proceed, use -allow-stripped option.\n";
     exit(1);
   }
 }
@@ -3284,8 +3284,11 @@ void RewriteInstance::processProfileData() {
   // Release memory used by profile reader.
   ProfileReader.reset();
 
-  if (opts::AggregateOnly)
+  if (opts::AggregateOnly) {
+    PrintProgramStats PPS(&*BAT);
+    BC->logBOLTErrorsAndQuitOnFatal(PPS.runOnFunctions(*BC));
     exit(0);
+  }
 }
 
 void RewriteInstance::disassembleFunctions() {
@@ -4808,6 +4811,40 @@ void RewriteInstance::updateELFSymbolTable(
     // Create a new symbol based on the existing symbol.
     ELFSymTy NewSymbol = Symbol;
 
+    // Handle special symbols based on their name.
+    Expected<StringRef> SymbolName = Symbol.getName(StringSection);
+    assert(SymbolName && "cannot get symbol name");
+
+    auto updateSymbolValue = [&](const StringRef Name,
+                                 std::optional<uint64_t> Value = std::nullopt) {
+      NewSymbol.st_value = Value ? *Value : getNewValueForSymbol(Name);
+      NewSymbol.st_shndx = ELF::SHN_ABS;
+      BC->outs() << "BOLT-INFO: setting " << Name << " to 0x"
+                 << Twine::utohexstr(NewSymbol.st_value) << '\n';
+    };
+
+    if (*SymbolName == "__hot_start" || *SymbolName == "__hot_end") {
+      if (opts::HotText) {
+        updateSymbolValue(*SymbolName);
+        ++NumHotTextSymsUpdated;
+      }
+      goto registerSymbol;
+    }
+
+    if (*SymbolName == "__hot_data_start" || *SymbolName == "__hot_data_end") {
+      if (opts::HotData) {
+        updateSymbolValue(*SymbolName);
+        ++NumHotDataSymsUpdated;
+      }
+      goto registerSymbol;
+    }
+
+    if (*SymbolName == "_end") {
+      if (NextAvailableAddress > Symbol.st_value)
+        updateSymbolValue(*SymbolName, NextAvailableAddress);
+      goto registerSymbol;
+    }
+
     if (Function) {
       // If the symbol matched a function that was not emitted, update the
       // corresponding section index but otherwise leave it unchanged.
@@ -4904,33 +4941,7 @@ void RewriteInstance::updateELFSymbolTable(
       }
     }
 
-    // Handle special symbols based on their name.
-    Expected<StringRef> SymbolName = Symbol.getName(StringSection);
-    assert(SymbolName && "cannot get symbol name");
-
-    auto updateSymbolValue = [&](const StringRef Name,
-                                 std::optional<uint64_t> Value = std::nullopt) {
-      NewSymbol.st_value = Value ? *Value : getNewValueForSymbol(Name);
-      NewSymbol.st_shndx = ELF::SHN_ABS;
-      BC->outs() << "BOLT-INFO: setting " << Name << " to 0x"
-                 << Twine::utohexstr(NewSymbol.st_value) << '\n';
-    };
-
-    if (opts::HotText &&
-        (*SymbolName == "__hot_start" || *SymbolName == "__hot_end")) {
-      updateSymbolValue(*SymbolName);
-      ++NumHotTextSymsUpdated;
-    }
-
-    if (opts::HotData && (*SymbolName == "__hot_data_start" ||
-                          *SymbolName == "__hot_data_end")) {
-      updateSymbolValue(*SymbolName);
-      ++NumHotDataSymsUpdated;
-    }
-
-    if (*SymbolName == "_end" && NextAvailableAddress > Symbol.st_value)
-      updateSymbolValue(*SymbolName, NextAvailableAddress);
-
+  registerSymbol:
     if (IsDynSym)
       Write((&Symbol - cantFail(Obj.symbols(&SymTabSection)).begin()) *
                 sizeof(ELFSymTy),

diff --git a/bolt/lib/Target/X86/X86MCPlusBuilder.cpp b/bolt/lib/Target/X86/X86MCPlusBuilder.cpp
@@ -1932,6 +1932,19 @@ class X86MCPlusBuilder : public MCPlusBuilder {
     //    = R_X86_64_PC32(Ln) + En - JT
     //    = R_X86_64_PC32(Ln + offsetof(En))
     //
+    auto isRIPRel = [&](X86MemOperand &MO) {
+      // NB: DispExpr should be set
+      return MO.DispExpr != nullptr &&
+             MO.BaseRegNum == RegInfo->getProgramCounter() &&
+             MO.IndexRegNum == X86::NoRegister &&
+             MO.SegRegNum == X86::NoRegister;
+    };
+    auto isIndexed = [](X86MemOperand &MO, MCPhysReg R) {
+      // NB: IndexRegNum should be set.
+      return MO.IndexRegNum != X86::NoRegister && MO.BaseRegNum == R &&
+             MO.ScaleImm == 4 && MO.DispImm == 0 &&
+             MO.SegRegNum == X86::NoRegister;
+    };
     LLVM_DEBUG(dbgs() << "Checking for PIC jump table\n");
     MCInst *MemLocInstr = nullptr;
     const MCInst *MovInstr = nullptr;
@@ -1965,9 +1978,8 @@ class X86MCPlusBuilder : public MCPlusBuilder {
         std::optional<X86MemOperand> MO = evaluateX86MemoryOperand(Instr);
         if (!MO)
           break;
-        if (MO->BaseRegNum != R1 || MO->ScaleImm != 4 ||
-            MO->IndexRegNum == X86::NoRegister || MO->DispImm != 0 ||
-            MO->SegRegNum != X86::NoRegister)
+        if (!isIndexed(*MO, R1))
+          // POSSIBLE_PIC_JUMP_TABLE
           break;
         MovInstr = &Instr;
       } else {
@@ -1986,9 +1998,7 @@ class X86MCPlusBuilder : public MCPlusBuilder {
         std::optional<X86MemOperand> MO = evaluateX86MemoryOperand(Instr);
         if (!MO)
           break;
-        if (MO->BaseRegNum != RegInfo->getProgramCounter() ||
-            MO->IndexRegNum != X86::NoRegister ||
-            MO->SegRegNum != X86::NoRegister || MO->DispExpr == nullptr)
+        if (!isRIPRel(*MO))
           break;
         MemLocInstr = &Instr;
         break;
@@ -2105,13 +2115,15 @@ class X86MCPlusBuilder : public MCPlusBuilder {
       return IndirectBranchType::POSSIBLE_FIXED_BRANCH;
     }
 
-    if (Type == IndirectBranchType::POSSIBLE_PIC_JUMP_TABLE &&
-        (MO->ScaleImm != 1 || MO->BaseRegNum != RIPRegister))
-      return IndirectBranchType::UNKNOWN;
-
-    if (Type != IndirectBranchType::POSSIBLE_PIC_JUMP_TABLE &&
-        MO->ScaleImm != PtrSize)
-      return IndirectBranchType::UNKNOWN;
+    switch (Type) {
+    case IndirectBranchType::POSSIBLE_PIC_JUMP_TABLE:
+      if (MO->ScaleImm != 1 || MO->BaseRegNum != RIPRegister)
+        return IndirectBranchType::UNKNOWN;
+      break;
+    default:
+      if (MO->ScaleImm != PtrSize)
+        return IndirectBranchType::UNKNOWN;
+    }
 
     MemLocInstrOut = MemLocInstr;
 

diff --git a/bolt/lib/Utils/CommandLineOpts.cpp b/bolt/lib/Utils/CommandLineOpts.cpp
@@ -179,6 +179,10 @@ cl::opt<bool> TimeOpts("time-opts",
                        cl::desc("print time spent in each optimization"),
                        cl::cat(BoltOptCategory));
 
+cl::opt<bool> TimeRewrite("time-rewrite",
+                          cl::desc("print time spent in rewriting passes"),
+                          cl::Hidden, cl::cat(BoltCategory));
+
 cl::opt<bool> UseOldText(
     "use-old-text",
     cl::desc("re-use space in old .text if possible (relocation mode)"),

diff --git a/bolt/runtime/instr.cpp b/bolt/runtime/instr.cpp
@@ -1245,7 +1245,6 @@ void Graph::computeEdgeFrequencies(const uint64_t *Counters,
       continue;
 
     assert(SpanningTreeNodes[Cur].NumInEdges == 1, "must have 1 parent");
-    const uint32_t Parent = SpanningTreeNodes[Cur].InEdges[0].Node;
     const uint32_t ParentEdge = SpanningTreeNodes[Cur].InEdges[0].ID;
 
     // Calculate parent edge freq.
@@ -1464,9 +1463,8 @@ void visitCallFlowEntry(CallFlowHashTable::MapEntry &Entry, int FD,
 int openProfile() {
   // Build the profile name string by appending our PID
   char Buf[BufSize];
-  char *Ptr = Buf;
   uint64_t PID = __getpid();
-  Ptr = strCopy(Buf, __bolt_instr_filename, BufSize);
+  char *Ptr = strCopy(Buf, __bolt_instr_filename, BufSize);
   if (__bolt_instr_use_pid) {
     Ptr = strCopy(Ptr, ".", BufSize - (Ptr - Buf + 1));
     Ptr = intToStr(Ptr, PID, 10);

diff --git a/bolt/test/CMakeLists.txt b/bolt/test/CMakeLists.txt
@@ -56,15 +56,14 @@ list(APPEND BOLT_TEST_DEPS
   )
 
 add_custom_target(bolt-test-depends DEPENDS ${BOLT_TEST_DEPS})
-set_target_properties(bolt-test-depends PROPERTIES FOLDER "BOLT")
+set_target_properties(bolt-test-depends PROPERTIES FOLDER "BOLT/Tests")
 
 add_lit_testsuite(check-bolt "Running the BOLT regression tests"
   ${CMAKE_CURRENT_BINARY_DIR}
   PARAMS ${BOLT_TEST_PARAMS}
   DEPENDS ${BOLT_TEST_DEPS}
   ARGS ${BOLT_TEST_EXTRA_ARGS}
   )
-set_target_properties(check-bolt PROPERTIES FOLDER "BOLT")
 
 add_lit_testsuites(BOLT ${CMAKE_CURRENT_SOURCE_DIR}
   PARAMS ${BOLT_TEST_PARAMS}

diff --git a/bolt/test/X86/bb-with-two-tail-calls.s b/bolt/test/X86/bb-with-two-tail-calls.s
@@ -1,19 +1,27 @@
 # This reproduces a bug with dynostats when trying to compute branch stats
 # at a block with two tails calls (one conditional and one unconditional).
 
-# REQUIRES: system-linux
-
 # RUN: llvm-mc -filetype=obj -triple x86_64-unknown-unknown \
 # RUN:   %s -o %t.o
 # RUN: link_fdata %s %t.o %t.fdata
 # RUN: llvm-strip --strip-unneeded %t.o
 # RUN: %clang %cflags %t.o -o %t.exe -Wl,-q -nostdlib
 # RUN: llvm-bolt %t.exe -o %t.out --data %t.fdata --lite=0 --dyno-stats \
 # RUN:    --print-sctc --print-only=_start -enable-bat 2>&1 | FileCheck %s
+# RUN: llvm-objdump --syms %t.out > %t.log
+# RUN: llvm-bat-dump %t.out --dump-all >> %t.log
+# RUN: FileCheck %s --input-file %t.log --check-prefix=CHECK-BAT
+
 # CHECK-NOT: Assertion `BranchInfo.size() == 2 && "could only be called for blocks with 2 successors"' failed.
 # Two tail calls in the same basic block after SCTC:
 # CHECK:         {{.*}}:   ja      {{.*}} # TAILCALL # Offset: 7 # CTCTakenCount: 4
-# CHECK-NEXT:    {{.*}}:   jmp     {{.*}} # TAILCALL # Offset: 12
+# CHECK-NEXT:    {{.*}}:   jmp     {{.*}} # TAILCALL # Offset: 13
+
+# Confirm that a deleted basic block is emitted at function end offset (0xe)
+# CHECK-BAT: [[#%x,ADDR:]] g .text  [[#%x,SIZE:]] _start
+# CHECK-BAT: Function Address: 0x[[#%x,ADDR]]
+# CHECK-BAT: 0x[[#%x,SIZE]]
+# CHECK-BAT: NumBlocks: 5
 
   .globl _start
 _start:
@@ -23,7 +31,9 @@ a:  ja b
 x:  ret
 # FDATA: 1 _start #a# 1 _start #b# 2 4
 b:  jmp e
-c:  jmp f
+c:
+    .nops 1
+    jmp f
 
   .globl e
 e:

diff --git a/bolt/test/X86/bolt-address-translation-yaml.test b/bolt/test/X86/bolt-address-translation-yaml.test
@@ -31,7 +31,8 @@ RUN: perf2bolt %t.out --pa -p %p/Inputs/blarge_new_bat.preagg.txt -w %t.yaml -o
 RUN:   2>&1 | FileCheck --check-prefix READ-BAT-CHECK %s
 RUN: FileCheck --input-file %t.yaml --check-prefix YAML-BAT-CHECK %s
 # Check that YAML converted from fdata matches YAML created directly with BAT.
-RUN: llvm-bolt %t.exe -data %t.fdata -w %t.yaml-fdata -o /dev/null
+RUN: llvm-bolt %t.exe -data %t.fdata -w %t.yaml-fdata -o /dev/null \
+RUN:   2>&1 | FileCheck --check-prefix READ-BAT-FDATA-CHECK %s
 RUN: FileCheck --input-file %t.yaml-fdata --check-prefix YAML-BAT-CHECK %s
 
 # Test resulting YAML profile with the original binary (no-stale mode)
@@ -40,11 +41,13 @@ RUN:   | FileCheck --check-prefix CHECK-BOLT-YAML %s
 
 WRITE-BAT-CHECK: BOLT-INFO: Wrote 5 BAT maps
 WRITE-BAT-CHECK: BOLT-INFO: Wrote 4 function and 22 basic block hashes
-WRITE-BAT-CHECK: BOLT-INFO: BAT section size (bytes): 384
+WRITE-BAT-CHECK: BOLT-INFO: BAT section size (bytes): 404
 
 READ-BAT-CHECK-NOT: BOLT-ERROR: unable to save profile in YAML format for input file processed by BOLT
 READ-BAT-CHECK: BOLT-INFO: Parsed 5 BAT entries
 READ-BAT-CHECK: PERF2BOLT: read 79 aggregated LBR entries
+READ-BAT-CHECK: BOLT-INFO: 5 out of 21 functions in the binary (23.8%) have non-empty execution profile
+READ-BAT-FDATA-CHECK: BOLT-INFO: 5 out of 16 functions in the binary (31.2%) have non-empty execution profile
 
 YAML-BAT-CHECK:      functions:
 # Function not covered by BAT - has insns in basic block

diff --git a/bolt/test/X86/bolt-address-translation.test b/bolt/test/X86/bolt-address-translation.test
@@ -37,7 +37,7 @@
 # CHECK:      BOLT: 3 out of 7 functions were overwritten.
 # CHECK:      BOLT-INFO: Wrote 6 BAT maps
 # CHECK:      BOLT-INFO: Wrote 3 function and 58 basic block hashes
-# CHECK:      BOLT-INFO: BAT section size (bytes): 928
+# CHECK:      BOLT-INFO: BAT section size (bytes): 940
 #
 # usqrt mappings (hot part). We match against any key (left side containing
 # the bolted binary offsets) because BOLT may change where it puts instructions

diff --git a/bolt/test/X86/dwarf5-debug-names-class-type-decl.s b/bolt/test/X86/dwarf5-debug-names-class-type-decl.s
diff --git a/bolt/test/X86/dwarf5-debug-names-enumeration-type-decl.s b/bolt/test/X86/dwarf5-debug-names-enumeration-type-decl.s
diff --git a/bolt/test/X86/dwarf5-debug-names-structure-type-decl.s b/bolt/test/X86/dwarf5-debug-names-structure-type-decl.s
diff --git a/bolt/test/X86/ignored-interprocedural-reference.s b/bolt/test/X86/ignored-interprocedural-reference.s
@@ -0,0 +1,49 @@
+# This reproduces a bug with not processing interprocedural references from
+# ignored functions.
+
+# REQUIRES: system-linux
+
+# RUN: llvm-mc -filetype=obj -triple x86_64-unknown-unknown %s -o %t.o
+# RUN: %clang %cflags %t.o -o %t.exe -nostdlib -Wl,-q
+# RUN: llvm-bolt %t.exe -o %t.out --enable-bat -funcs=main
+# RUN: link_fdata %s %t.out %t.preagg PREAGG
+# RUN: perf2bolt %t.out -p %t.preagg --pa -o %t.fdata -w %t.yaml
+# RUN: FileCheck %s --input-file=%t.fdata --check-prefix=CHECK-FDATA
+# RUN: FileCheck %s --input-file=%t.yaml --check-prefix=CHECK-YAML
+
+# CHECK-FDATA: 1 main 0 1 foo a 1 1
+# CHECK-YAML: name: main
+# CHECK-YAML: calls: {{.*}} disc: 1
+
+# PREAGG: B #main# #foo_secondary# 1 1
+# main calls foo at valid instruction offset past nops that are to be stripped.
+  .globl main
+main:
+  .cfi_startproc
+  call foo_secondary
+  ret
+  .cfi_endproc
+.size main,.-main
+
+# Placeholder cold fragment to force main to be ignored in non-relocation mode.
+  .globl main.cold
+main.cold:
+  .cfi_startproc
+  ud2
+  .cfi_endproc
+.size main.cold,.-main.cold
+
+# foo is set up to contain a valid instruction at called offset, and trapping
+# instructions past that.
+  .globl foo
+foo:
+  .cfi_startproc
+  .nops 10
+  .globl foo_secondary
+foo_secondary:
+  ret
+  .rept 20
+  int3
+  .endr
+  .cfi_endproc
+.size foo,.-foo
diff --git a/bolt/test/X86/register-fragments-bolt-symbols.s b/bolt/test/X86/register-fragments-bolt-symbols.s
@@ -18,6 +18,11 @@
 # RUN: FileCheck --input-file %t.bat.fdata --check-prefix=CHECK-FDATA %s
 # RUN: FileCheck --input-file %t.bat.yaml --check-prefix=CHECK-YAML %s
 
+# RUN: link_fdata --no-redefine %s %t.bolt %t.preagg2 PREAGG2
+# PREAGG2: B X:0 #chain# 1 0
+# RUN: perf2bolt %t.bolt -p %t.preagg2 --pa -o %t.bat2.fdata -w %t.bat2.yaml
+# RUN: FileCheck %s --input-file %t.bat2.yaml --check-prefix=CHECK-YAML2
+
 # CHECK-SYMS: l df *ABS*          [[#]] chain.s
 # CHECK-SYMS: l  F .bolt.org.text [[#]] chain
 # CHECK-SYMS: l  F .text.cold     [[#]] chain.cold.0
@@ -28,6 +33,9 @@
 
 # CHECK-FDATA: 0 [unknown] 0 1 chain/chain.s/2 10 0 1
 # CHECK-YAML: - name: 'chain/chain.s/2'
+# CHECK-YAML2: - name: 'chain/chain.s/1'
+## non-BAT function has non-zero insns:
+# CHECK-YAML2: insns: 1
 
 .file "chain.s"
         .text

diff --git a/bolt/test/X86/yaml-non-simple.test b/bolt/test/X86/yaml-non-simple.test
@@ -0,0 +1,71 @@
+## Check that YAML profile for non-simple function is not reported as stale.
+
+# RUN: split-file %s %t
+# RUN: llvm-mc -filetype=obj -triple x86_64-unknown-unknown %t/main.s -o %t.o
+# RUN: %clang %cflags %t.o -o %t.exe -nostdlib
+# RUN: llvm-bolt %t.exe -o %t.out --data %t/yaml --profile-ignore-hash -v=1 \
+# RUN:   --report-stale 2>&1 | FileCheck %s
+
+# CHECK: BOLT-INFO: could not disassemble function main. Will ignore.
+# CHECK: BOLT-INFO: could not disassemble function main.cold. Will ignore.
+# CHECK: BOLT-INFO: 0 out of 2 functions in the binary (0.0%) have non-empty execution profile
+# CHECK: BOLT-INFO: 1 function with profile could not be optimized
+
+#--- main.s
+.globl main
+.type	main, @function
+main:
+  .cfi_startproc
+.LBB00:
+  pushq   %rbp
+  movq    %rsp, %rbp
+  subq    $16, %rsp
+  testq   %rax, %rax
+  js      .LBB03
+.LBB01:
+  jne     .LBB04
+.LBB02:
+  nop
+.LBB03:
+  xorl    %eax, %eax
+  addq    $16, %rsp
+  popq    %rbp
+  retq
+.LBB04:
+  xorl    %eax, %eax
+  addq    $16, %rsp
+  popq    %rbp
+  retq
+  .cfi_endproc
+  .size	main, .-main
+
+.globl main.cold
+.type	main.cold, @function
+main.cold:
+  .cfi_startproc
+  nop
+  .cfi_endproc
+  .size	main.cold, .-main.cold
+
+#--- yaml
+---
+header:
+  profile-version: 1
+  binary-name:     'yaml-non-simple.s.tmp.exe'
+  binary-build-id: '<unknown>'
+  profile-flags:   [ lbr ]
+  profile-origin:  branch profile reader
+  profile-events:  ''
+  dfs-order:       false
+  hash-func:       xxh3
+functions:
+  - name:            main
+    fid:             0
+    hash:            0x0000000000000000
+    exec:            1
+    nblocks:         5
+    blocks:
+      - bid:             1
+        insns:           1
+        succ:            [ { bid: 3, cnt: 1} ]
+...
diff --git a/bolt/test/link_fdata.py b/bolt/test/link_fdata.py
@@ -19,6 +19,7 @@
 parser.add_argument("prefix", nargs="?", default="FDATA", help="Custom FDATA prefix")
 parser.add_argument("--nmtool", default="nm", help="Path to nm tool")
 parser.add_argument("--no-lbr", action="store_true")
+parser.add_argument("--no-redefine", action="store_true")
 
 args = parser.parse_args()
 
@@ -90,6 +91,8 @@
 symbols = {}
 for symline in nm_output.splitlines():
     symval, _, symname = symline.split(maxsplit=2)
+    if symname in symbols and args.no_redefine:
+        continue
     symbols[symname] = symval
 
 

diff --git a/bolt/test/runtime/X86/hot-end-symbol.s b/bolt/test/runtime/X86/hot-end-symbol.s
@@ -12,6 +12,7 @@
 # RUN: %clang %cflags -no-pie %t.o -o %t.exe -Wl,-q
 
 # RUN: llvm-bolt %t.exe --relocs=1 --hot-text --reorder-functions=hfsort \
+# RUN:    --split-functions --split-strategy=all \
 # RUN:    --data %t.fdata -o %t.out | FileCheck %s
 
 # RUN: %t.out 1
@@ -30,12 +31,12 @@
 # CHECK-OUTPUT:       __hot_start
 # CHECK-OUTPUT-NEXT:  main
 # CHECK-OUTPUT-NEXT:  __hot_end
+# CHECK-OUTPUT-NOT:   __hot_start.cold
 
   .text
   .globl  main
   .type main, %function
   .globl  __hot_start
-  .type __hot_start, %object
   .p2align  4
 main:
 __hot_start:

diff --git a/bolt/unittests/CMakeLists.txt b/bolt/unittests/CMakeLists.txt
@@ -1,5 +1,5 @@
 add_custom_target(BoltUnitTests)
-set_target_properties(BoltUnitTests PROPERTIES FOLDER "BOLT tests")
+set_target_properties(BoltUnitTests PROPERTIES FOLDER "BOLT/Tests")
 
 function(add_bolt_unittest test_dirname)
   add_unittest(BoltUnitTests ${test_dirname} ${ARGN})

diff --git a/clang-tools-extra/CMakeLists.txt b/clang-tools-extra/CMakeLists.txt
@@ -1,3 +1,5 @@
+set(LLVM_SUBPROJECT_TITLE "Clang Tools Extra")
+
 include(CMakeDependentOption)
 include(GNUInstallDirs)
 

diff --git a/clang-tools-extra/clang-tidy/CMakeLists.txt b/clang-tools-extra/clang-tidy/CMakeLists.txt
@@ -121,7 +121,7 @@ if (NOT LLVM_INSTALL_TOOLCHAIN_ONLY)
     PATTERN "*.h"
     )
   add_custom_target(clang-tidy-headers)
-  set_target_properties(clang-tidy-headers PROPERTIES FOLDER "Misc")
+  set_target_properties(clang-tidy-headers PROPERTIES FOLDER "Clang Tools Extra/Resources")
   if(NOT LLVM_ENABLE_IDE)
     add_llvm_install_targets(install-clang-tidy-headers
                              DEPENDS clang-tidy-headers

diff --git a/clang-tools-extra/clang-tidy/bugprone/ForwardingReferenceOverloadCheck.cpp b/clang-tools-extra/clang-tidy/bugprone/ForwardingReferenceOverloadCheck.cpp
@@ -54,7 +54,9 @@ AST_MATCHER(QualType, isEnableIf) {
 AST_MATCHER_P(TemplateTypeParmDecl, hasDefaultArgument,
               clang::ast_matchers::internal::Matcher<QualType>, TypeMatcher) {
   return Node.hasDefaultArgument() &&
-         TypeMatcher.matches(Node.getDefaultArgument(), Finder, Builder);
+         TypeMatcher.matches(
+             Node.getDefaultArgument().getArgument().getAsType(), Finder,
+             Builder);
 }
 AST_MATCHER(TemplateDecl, hasAssociatedConstraints) {
   return Node.hasAssociatedConstraints();

diff --git a/clang-tools-extra/clang-tidy/bugprone/IncorrectEnableIfCheck.cpp b/clang-tools-extra/clang-tidy/bugprone/IncorrectEnableIfCheck.cpp
@@ -19,10 +19,11 @@ namespace {
 AST_MATCHER_P(TemplateTypeParmDecl, hasUnnamedDefaultArgument,
               ast_matchers::internal::Matcher<TypeLoc>, InnerMatcher) {
   if (Node.getIdentifier() != nullptr || !Node.hasDefaultArgument() ||
-      Node.getDefaultArgumentInfo() == nullptr)
+      Node.getDefaultArgument().getArgument().isNull())
     return false;
 
-  TypeLoc DefaultArgTypeLoc = Node.getDefaultArgumentInfo()->getTypeLoc();
+  TypeLoc DefaultArgTypeLoc =
+      Node.getDefaultArgument().getTypeSourceInfo()->getTypeLoc();
   return InnerMatcher.matches(DefaultArgTypeLoc, Finder, Builder);
 }
 

diff --git a/clang-tools-extra/clang-tidy/bugprone/SizeofExpressionCheck.cpp b/clang-tools-extra/clang-tidy/bugprone/SizeofExpressionCheck.cpp
@@ -144,16 +144,13 @@ void SizeofExpressionCheck::registerMatchers(MatchFinder *Finder) {
         unaryOperator(hasUnaryOperand(ArrayExpr), unless(hasOperatorName("*"))),
         binaryOperator(hasEitherOperand(ArrayExpr)),
         castExpr(hasSourceExpression(ArrayExpr))));
-    const auto PointerToArrayExpr = ignoringParenImpCasts(
-        hasType(hasCanonicalType(pointerType(pointee(arrayType())))));
+    const auto PointerToArrayExpr =
+        hasType(hasCanonicalType(pointerType(pointee(arrayType()))));
 
-    const auto StructAddrOfExpr = unaryOperator(
-        hasOperatorName("&"), hasUnaryOperand(ignoringParenImpCasts(
-                                  hasType(hasCanonicalType(recordType())))));
     const auto PointerToStructType =
         hasUnqualifiedDesugaredType(pointerType(pointee(recordType())));
-    const auto PointerToStructExpr = ignoringParenImpCasts(expr(
-        hasType(hasCanonicalType(PointerToStructType)), unless(cxxThisExpr())));
+    const auto PointerToStructExpr = expr(
+        hasType(hasCanonicalType(PointerToStructType)), unless(cxxThisExpr()));
 
     const auto ArrayOfPointersExpr = ignoringParenImpCasts(
         hasType(hasCanonicalType(arrayType(hasElementType(pointerType()))
@@ -166,18 +163,19 @@ void SizeofExpressionCheck::registerMatchers(MatchFinder *Finder) {
         ignoringParenImpCasts(arraySubscriptExpr(
             hasBase(ArrayOfSamePointersExpr), hasIndex(ZeroLiteral)));
     const auto ArrayLengthExprDenom =
-        expr(hasParent(expr(ignoringParenImpCasts(binaryOperator(
-                 hasOperatorName("/"), hasLHS(ignoringParenImpCasts(sizeOfExpr(
-                                           has(ArrayOfPointersExpr)))))))),
+        expr(hasParent(binaryOperator(hasOperatorName("/"),
+                                      hasLHS(ignoringParenImpCasts(sizeOfExpr(
+                                          has(ArrayOfPointersExpr)))))),
              sizeOfExpr(has(ArrayOfSamePointersZeroSubscriptExpr)));
 
-    Finder->addMatcher(expr(anyOf(sizeOfExpr(has(ignoringParenImpCasts(anyOf(
-                                      ArrayCastExpr, PointerToArrayExpr,
-                                      StructAddrOfExpr, PointerToStructExpr)))),
-                                  sizeOfExpr(has(PointerToStructType))),
-                            unless(ArrayLengthExprDenom))
-                           .bind("sizeof-pointer-to-aggregate"),
-                       this);
+    Finder->addMatcher(
+        expr(sizeOfExpr(anyOf(
+                 has(ignoringParenImpCasts(anyOf(
+                     ArrayCastExpr, PointerToArrayExpr, PointerToStructExpr))),
+                 has(PointerToStructType))),
+             unless(ArrayLengthExprDenom))
+            .bind("sizeof-pointer-to-aggregate"),
+        this);
   }
 
   // Detect expression like: sizeof(expr) <= k for a suspicious constant 'k'.

diff --git a/clang-tools-extra/clang-tidy/misc/CMakeLists.txt b/clang-tools-extra/clang-tidy/misc/CMakeLists.txt
@@ -15,6 +15,7 @@ add_custom_command(
     DEPENDS ${clang_tidy_confusable_chars_gen_target} ConfusableTable/confusables.txt)
 
 add_custom_target(genconfusable DEPENDS Confusables.inc)
+set_target_properties(genconfusable PROPERTIES FOLDER "Clang Tools Extra/Sourcegenning")
 
 add_clang_library(clangTidyMiscModule
   ConstCorrectnessCheck.cpp

diff --git a/clang-tools-extra/clang-tidy/modernize/UseConstraintsCheck.cpp b/clang-tools-extra/clang-tidy/modernize/UseConstraintsCheck.cpp
@@ -177,9 +177,11 @@ matchTrailingTemplateParam(const FunctionTemplateDecl *FunctionTemplate) {
           dyn_cast<TemplateTypeParmDecl>(LastParam)) {
     if (LastTemplateParam->hasDefaultArgument() &&
         LastTemplateParam->getIdentifier() == nullptr) {
-      return {matchEnableIfSpecialization(
-                  LastTemplateParam->getDefaultArgumentInfo()->getTypeLoc()),
-              LastTemplateParam};
+      return {
+          matchEnableIfSpecialization(LastTemplateParam->getDefaultArgument()
+                                          .getTypeSourceInfo()
+                                          ->getTypeLoc()),
+          LastTemplateParam};
     }
   }
   return {};

diff --git a/clang-tools-extra/clang-tidy/readability/ImplicitBoolConversionCheck.cpp b/clang-tools-extra/clang-tidy/readability/ImplicitBoolConversionCheck.cpp
@@ -50,7 +50,9 @@ StringRef getZeroLiteralToCompareWithForType(CastKind CastExprKind,
 
   case CK_PointerToBoolean:
   case CK_MemberPointerToBoolean: // Fall-through on purpose.
-    return Context.getLangOpts().CPlusPlus11 ? "nullptr" : "0";
+    return (Context.getLangOpts().CPlusPlus11 || Context.getLangOpts().C23)
+               ? "nullptr"
+               : "0";
 
   default:
     llvm_unreachable("Unexpected cast kind");
@@ -165,6 +167,12 @@ bool needsSpacePrefix(SourceLocation Loc, ASTContext &Context) {
 void fixGenericExprCastFromBool(DiagnosticBuilder &Diag,
                                 const ImplicitCastExpr *Cast,
                                 ASTContext &Context, StringRef OtherType) {
+  if (!Context.getLangOpts().CPlusPlus) {
+    Diag << FixItHint::CreateInsertion(Cast->getBeginLoc(),
+                                       (Twine("(") + OtherType + ")").str());
+    return;
+  }
+
   const Expr *SubExpr = Cast->getSubExpr();
   const bool NeedParens = !isa<ParenExpr>(SubExpr->IgnoreImplicit());
   const bool NeedSpace = needsSpacePrefix(Cast->getBeginLoc(), Context);
@@ -267,6 +275,10 @@ void ImplicitBoolConversionCheck::registerMatchers(MatchFinder *Finder) {
   auto BoolXor =
       binaryOperator(hasOperatorName("^"), hasLHS(ImplicitCastFromBool),
                      hasRHS(ImplicitCastFromBool));
+  auto ComparisonInCall = allOf(
+      hasParent(callExpr()),
+      hasSourceExpression(binaryOperator(hasAnyOperatorName("==", "!="))));
+
   Finder->addMatcher(
       traverse(TK_AsIs,
                implicitCastExpr(
@@ -281,6 +293,8 @@ void ImplicitBoolConversionCheck::registerMatchers(MatchFinder *Finder) {
                        stmt(anyOf(ifStmt(), whileStmt()), has(declStmt())))),
                    // Exclude cases common to implicit cast to and from bool.
                    unless(ExceptionCases), unless(has(BoolXor)),
+                   // Exclude C23 cases common to implicit cast to bool.
+                   unless(ComparisonInCall),
                    // Retrieve also parent statement, to check if we need
                    // additional parens in replacement.
                    optionally(hasParent(stmt().bind("parentStmt"))),

diff --git a/clang-tools-extra/clang-tidy/utils/RenamerClangTidyCheck.cpp b/clang-tools-extra/clang-tidy/utils/RenamerClangTidyCheck.cpp
@@ -123,6 +123,9 @@ static const NamedDecl *getFailureForNamedDecl(const NamedDecl *ND) {
   if (const auto *Method = dyn_cast<CXXMethodDecl>(ND)) {
     if (const CXXMethodDecl *Overridden = getOverrideMethod(Method))
       Canonical = cast<NamedDecl>(Overridden->getCanonicalDecl());
+    else if (const FunctionTemplateDecl *Primary = Method->getPrimaryTemplate())
+      if (const FunctionDecl *TemplatedDecl = Primary->getTemplatedDecl())
+        Canonical = cast<NamedDecl>(TemplatedDecl->getCanonicalDecl());
 
     if (Canonical != ND)
       return Canonical;

diff --git a/clang-tools-extra/clangd/Hover.cpp b/clang-tools-extra/clangd/Hover.cpp
@@ -247,8 +247,12 @@ fetchTemplateParameters(const TemplateParameterList *Params,
       if (!TTP->getName().empty())
         P.Name = TTP->getNameAsString();
 
-      if (TTP->hasDefaultArgument())
-        P.Default = TTP->getDefaultArgument().getAsString(PP);
+      if (TTP->hasDefaultArgument()) {
+        P.Default.emplace();
+        llvm::raw_string_ostream Out(*P.Default);
+        TTP->getDefaultArgument().getArgument().print(PP, Out,
+                                                      /*IncludeType=*/false);
+      }
     } else if (const auto *NTTP = dyn_cast<NonTypeTemplateParmDecl>(Param)) {
       P.Type = printType(NTTP, PP);
 
@@ -258,7 +262,8 @@ fetchTemplateParameters(const TemplateParameterList *Params,
       if (NTTP->hasDefaultArgument()) {
         P.Default.emplace();
         llvm::raw_string_ostream Out(*P.Default);
-        NTTP->getDefaultArgument()->printPretty(Out, nullptr, PP);
+        NTTP->getDefaultArgument().getArgument().print(PP, Out,
+                                                       /*IncludeType=*/false);
       }
     } else if (const auto *TTPD = dyn_cast<TemplateTemplateParmDecl>(Param)) {
       P.Type = printType(TTPD, PP);

diff --git a/clang-tools-extra/clangd/test/infinite-instantiation.test b/clang-tools-extra/clangd/test/infinite-instantiation.test
@@ -1,5 +1,6 @@
-// RUN: cp %s %t.cpp
-// RUN: not clangd -check=%t.cpp 2>&1 | FileCheck -strict-whitespace %s
+// RUN: rm -rf %t.dir && mkdir -p %t.dir
+// RUN: echo '[{"directory": "%/t.dir", "command": "clang -ftemplate-depth=100 -x c++ %/s", "file": "%/s"}]' > %t.dir/compile_commands.json
+// RUN: not clangd --compile-commands-dir=%t.dir -check=%s 2>&1 | FileCheck -strict-whitespace %s
 
 // CHECK: [template_recursion_depth_exceeded]
 

diff --git a/clang-tools-extra/clangd/unittests/CMakeLists.txt b/clang-tools-extra/clangd/unittests/CMakeLists.txt
@@ -29,6 +29,7 @@ include(${CMAKE_CURRENT_SOURCE_DIR}/../quality/CompletionModel.cmake)
 gen_decision_forest(${CMAKE_CURRENT_SOURCE_DIR}/decision_forest_model DecisionForestRuntimeTest ::ns1::ns2::test::Example)
 
 add_custom_target(ClangdUnitTests)
+set_target_properties(ClangdUnitTests PROPERTIES FOLDER "Clang Tools Extra/Tests")
 add_unittest(ClangdUnitTests ClangdTests
   Annotations.cpp
   ASTTests.cpp

diff --git a/clang-tools-extra/clangd/unittests/ClangdTests.cpp b/clang-tools-extra/clangd/unittests/ClangdTests.cpp
@@ -392,7 +392,7 @@ TEST(ClangdServerTest, SearchLibDir) {
   ErrorCheckingCallbacks DiagConsumer;
   MockCompilationDatabase CDB;
   CDB.ExtraClangFlags.insert(CDB.ExtraClangFlags.end(),
-                             {"-xc++", "-target", "x86_64-linux-unknown",
+                             {"-xc++", "--target=x86_64-unknown-linux-gnu",
                               "-m64", "--gcc-toolchain=/randomusr",
                               "-stdlib=libstdc++"});
   ClangdServer Server(CDB, FS, ClangdServer::optsForTest(), &DiagConsumer);

diff --git a/clang-tools-extra/clangd/unittests/FindTargetTests.cpp b/clang-tools-extra/clangd/unittests/FindTargetTests.cpp
@@ -836,7 +836,9 @@ TEST_F(TargetDeclTest, OverloadExpr) {
       [[delete]] x;
     }
   )cpp";
-  EXPECT_DECLS("CXXDeleteExpr", "void operator delete(void *) noexcept");
+  // Sized deallocation is enabled by default in C++14 onwards.
+  EXPECT_DECLS("CXXDeleteExpr",
+               "void operator delete(void *, unsigned long) noexcept");
 }
 
 TEST_F(TargetDeclTest, DependentExprs) {

diff --git a/clang-tools-extra/docs/CMakeLists.txt b/clang-tools-extra/docs/CMakeLists.txt
@@ -77,6 +77,7 @@ if (DOXYGEN_FOUND)
       COMMAND ${DOXYGEN_EXECUTABLE} ${CMAKE_CURRENT_BINARY_DIR}/doxygen.cfg
       WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
       COMMENT "Generating clang doxygen documentation." VERBATIM)
+    set_target_properties(doxygen-clang-tools PROPERTIES FOLDER "Clang Tools Extra/Docs")
 
     if (LLVM_BUILD_DOCS)
       add_dependencies(doxygen doxygen-clang-tools)

diff --git a/clang-tools-extra/docs/ReleaseNotes.rst b/clang-tools-extra/docs/ReleaseNotes.rst
@@ -375,12 +375,15 @@ Changes in existing checks
   <clang-tidy/checks/readability/identifier-naming>` check in `GetConfigPerFile`
   mode by resolving symbolic links to header files. Fixed handling of Hungarian
   Prefix when configured to `LowerCase`. Added support for renaming designated
-  initializers. Added support for renaming macro arguments.
+  initializers. Added support for renaming macro arguments. Fixed renaming
+  conflicts arising from out-of-line member function template definitions.
 
 - Improved :doc:`readability-implicit-bool-conversion
   <clang-tidy/checks/readability/implicit-bool-conversion>` check to provide
   valid fix suggestions for ``static_cast`` without a preceding space and
-  fixed problem with duplicate parentheses in double implicit casts.
+  fixed problem with duplicate parentheses in double implicit casts. Corrected
+  the fix suggestions for C23 and later by using C-style casts instead of
+  ``static_cast``.
 
 - Improved :doc:`readability-redundant-inline-specifier
   <clang-tidy/checks/readability/redundant-inline-specifier>` check to properly

diff --git a/clang-tools-extra/docs/clang-tidy/checks/readability/implicit-bool-conversion.rst b/clang-tools-extra/docs/clang-tidy/checks/readability/implicit-bool-conversion.rst
@@ -96,8 +96,8 @@ The rules for generating fix-it hints are:
   - ``if (!pointer)`` is changed to ``if (pointer == nullptr)``,
 
 - in case of conversions from bool to other built-in types, an explicit
-  ``static_cast`` is proposed to make it clear that a conversion is taking
-  place:
+  ``static_cast`` (or a C-style cast since C23) is proposed to make it clear
+  that a conversion is taking place:
 
   - ``int integer = boolean;`` is changed to
     ``int integer = static_cast<int>(boolean);``,

diff --git a/clang-tools-extra/include-cleaner/unittests/CMakeLists.txt b/clang-tools-extra/include-cleaner/unittests/CMakeLists.txt
@@ -4,6 +4,7 @@ set(LLVM_LINK_COMPONENTS
   )
 
 add_custom_target(ClangIncludeCleanerUnitTests)
+set_target_properties(ClangIncludeCleanerUnitTests PROPERTIES FOLDER "Clang Tools Extra/Tests")
 add_unittest(ClangIncludeCleanerUnitTests ClangIncludeCleanerTests
   AnalysisTest.cpp
   FindHeadersTest.cpp

diff --git a/clang-tools-extra/modularize/ModularizeUtilities.cpp b/clang-tools-extra/modularize/ModularizeUtilities.cpp
@@ -435,11 +435,9 @@ static std::string replaceDotDot(StringRef Path) {
   llvm::sys::path::const_iterator B = llvm::sys::path::begin(Path),
     E = llvm::sys::path::end(Path);
   while (B != E) {
-    if (B->compare(".") == 0) {
-    }
-    else if (B->compare("..") == 0)
+    if (*B == "..")
       llvm::sys::path::remove_filename(Buffer);
-    else
+    else if (*B != ".")
       llvm::sys::path::append(Buffer, *B);
     ++B;
   }

diff --git a/clang-tools-extra/pseudo/include/CMakeLists.txt b/clang-tools-extra/pseudo/include/CMakeLists.txt
@@ -29,3 +29,4 @@ add_custom_command(OUTPUT ${cxx_bnf_inc}
 add_custom_target(cxx_gen
     DEPENDS ${cxx_symbols_inc} ${cxx_bnf_inc}
     VERBATIM)
+set_target_properties(cxx_gen PROPERTIES FOLDER "Clang Tools Extra/Sourcegenning")
diff --git a/clang-tools-extra/pseudo/tool/CMakeLists.txt b/clang-tools-extra/pseudo/tool/CMakeLists.txt
@@ -26,4 +26,5 @@ add_custom_command(OUTPUT HTMLForestResources.inc
   DEPENDS ${CLANG_SOURCE_DIR}/utils/bundle_resources.py HTMLForest.css HTMLForest.js HTMLForest.html
   VERBATIM)
 add_custom_target(clang-pseudo-resources DEPENDS HTMLForestResources.inc)
+set_target_properties(clang-pseudo-resources PROPERTIES FOLDER "Clang Tools Extra/Resources")
 add_dependencies(clang-pseudo clang-pseudo-resources)
diff --git a/clang-tools-extra/pseudo/unittests/CMakeLists.txt b/clang-tools-extra/pseudo/unittests/CMakeLists.txt
@@ -3,6 +3,7 @@ set(LLVM_LINK_COMPONENTS
   )
 
 add_custom_target(ClangPseudoUnitTests)
+set_target_properties(ClangPseudoUnitTests PROPERTIES FOLDER "Clang Tools Extra/Tests")
 add_unittest(ClangPseudoUnitTests ClangPseudoTests
   BracketTest.cpp
   CXXTest.cpp

diff --git a/clang-tools-extra/test/CMakeLists.txt b/clang-tools-extra/test/CMakeLists.txt
@@ -97,7 +97,6 @@ add_lit_testsuite(check-clang-extra "Running clang-tools-extra/test"
    ${CMAKE_CURRENT_BINARY_DIR}
    DEPENDS ${CLANG_TOOLS_TEST_DEPS}
    )
-set_target_properties(check-clang-extra PROPERTIES FOLDER "Clang extra tools' tests")
 
 add_lit_testsuites(CLANG-EXTRA ${CMAKE_CURRENT_SOURCE_DIR}
   DEPENDS ${CLANG_TOOLS_TEST_DEPS}

diff --git a/...-tools-extra/test/clang-tidy/checkers/cppcoreguidelines/pro-type-member-init-no-crash.cpp b/...-tools-extra/test/clang-tidy/checkers/cppcoreguidelines/pro-type-member-init-no-crash.cpp
@@ -5,3 +5,11 @@ struct X {
   // CHECK-MESSAGES: :[[@LINE-1]]:5: error: field has incomplete type 'X' [clang-diagnostic-error]
   int a = 10;
 };
+
+template <typename T> class NoCrash {
+  // CHECK-MESSAGES: :[[@LINE+2]]:20: error: base class has incomplete type
+  // CHECK-MESSAGES: :[[@LINE-2]]:29: note: definition of 'NoCrash<T>' is not complete until the closing '}'
+  class B : public NoCrash {
+    template <typename U> B(U u) {}
+  };
+};
diff --git a/clang-tools-extra/test/clang-tidy/checkers/cppcoreguidelines/pro-type-member-init.cpp b/clang-tools-extra/test/clang-tidy/checkers/cppcoreguidelines/pro-type-member-init.cpp
@@ -463,12 +463,6 @@ struct NegativeIncompleteArrayMember {
   char e[];
 };
 
-template <typename T> class NoCrash {
-  class B : public NoCrash {
-    template <typename U> B(U u) {}
-  };
-};
-
 struct PositiveBitfieldMember {
   PositiveBitfieldMember() {}
   // CHECK-MESSAGES: :[[@LINE-1]]:3: warning: constructor does not initialize these fields: F

diff --git a/clang-tools-extra/test/clang-tidy/checkers/misc/new-delete-overloads.cpp b/clang-tools-extra/test/clang-tidy/checkers/misc/new-delete-overloads.cpp
@@ -12,16 +12,6 @@ struct S {
 // CHECK-MESSAGES: :[[@LINE+1]]:7: warning: declaration of 'operator new' has no matching declaration of 'operator delete' at the same scope
 void *operator new(size_t size) noexcept(false);
 
-struct T {
-  // Sized deallocations are not enabled by default, and so this new/delete pair
-  // does not match. However, we expect only one warning, for the new, because
-  // the operator delete is a placement delete and we do not warn on mismatching
-  // placement operations.
-  // CHECK-MESSAGES: :[[@LINE+1]]:9: warning: declaration of 'operator new' has no matching declaration of 'operator delete' at the same scope
-  void *operator new(size_t size) noexcept;
-  void operator delete(void *ptr, size_t) noexcept; // ok only if sized deallocation is enabled
-};
-
 struct U {
   void *operator new(size_t size) noexcept;
   void operator delete(void *ptr) noexcept;

diff --git a/clang-tools-extra/test/clang-tidy/checkers/readability/identifier-naming-outofline.cpp b/clang-tools-extra/test/clang-tidy/checkers/readability/identifier-naming-outofline.cpp
@@ -0,0 +1,30 @@
+// RUN: %check_clang_tidy %s readability-identifier-naming %t -std=c++20 \
+// RUN:   --config='{CheckOptions: { \
+// RUN:     readability-identifier-naming.MethodCase: CamelCase, \
+// RUN:  }}'
+
+namespace SomeNamespace {
+namespace Inner {
+
+class SomeClass {
+public:
+    template <typename T>
+    int someMethod();
+// CHECK-MESSAGES: :[[@LINE-1]]:9: warning: invalid case style for method 'someMethod' [readability-identifier-naming]
+// CHECK-FIXES: {{^}}    int SomeMethod();
+};
+template <typename T>
+int SomeClass::someMethod() {
+// CHECK-FIXES: {{^}}int SomeClass::SomeMethod() {
+    return 5;
+}
+
+} // namespace Inner
+
+void someFunc() {
+    Inner::SomeClass S;
+    S.someMethod<int>();
+// CHECK-FIXES: {{^}}    S.SomeMethod<int>();
+}
+
+} // namespace SomeNamespace
diff --git a/clang-tools-extra/test/clang-tidy/checkers/readability/implicit-bool-conversion.c b/clang-tools-extra/test/clang-tidy/checkers/readability/implicit-bool-conversion.c
@@ -0,0 +1,354 @@
+// RUN: %check_clang_tidy %s readability-implicit-bool-conversion %t -- -- -std=c23
+
+#undef NULL
+#define NULL 0L
+
+void functionTakingBool(bool);
+void functionTakingInt(int);
+void functionTakingUnsignedLong(unsigned long);
+void functionTakingChar(char);
+void functionTakingFloat(float);
+void functionTakingDouble(double);
+void functionTakingSignedChar(signed char);
+
+
+////////// Implicit conversion from bool.
+
+void implicitConversionFromBoolSimpleCases() {
+  bool boolean = true;
+
+  functionTakingBool(boolean);
+
+  functionTakingInt(boolean);
+  // CHECK-MESSAGES: :[[@LINE-1]]:21: warning: implicit conversion 'bool' -> 'int' [readability-implicit-bool-conversion]
+  // CHECK-FIXES: functionTakingInt((int)boolean);
+
+  functionTakingUnsignedLong(boolean);
+  // CHECK-MESSAGES: :[[@LINE-1]]:30: warning: implicit conversion 'bool' -> 'unsigned long'
+  // CHECK-FIXES: functionTakingUnsignedLong((unsigned long)boolean);
+
+  functionTakingChar(boolean);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'bool' -> 'char'
+  // CHECK-FIXES: functionTakingChar((char)boolean);
+
+  functionTakingFloat(boolean);
+  // CHECK-MESSAGES: :[[@LINE-1]]:23: warning: implicit conversion 'bool' -> 'float'
+  // CHECK-FIXES: functionTakingFloat((float)boolean);
+
+  functionTakingDouble(boolean);
+  // CHECK-MESSAGES: :[[@LINE-1]]:24: warning: implicit conversion 'bool' -> 'double'
+  // CHECK-FIXES: functionTakingDouble((double)boolean);
+}
+
+float implicitConversionFromBoolInReturnValue() {
+  bool boolean = false;
+  return boolean;
+  // CHECK-MESSAGES: :[[@LINE-1]]:10: warning: implicit conversion 'bool' -> 'float'
+  // CHECK-FIXES: return (float)boolean;
+}
+
+void implicitConversionFromBoolInSingleBoolExpressions(bool b1, bool b2) {
+  bool boolean = true;
+  boolean = b1 ^ b2;
+  boolean |= !b1 || !b2;
+  boolean &= b1;
+
+  int integer = boolean - 3;
+  // CHECK-MESSAGES: :[[@LINE-1]]:17: warning: implicit conversion 'bool' -> 'int'
+  // CHECK-FIXES: int integer = (int)boolean - 3;
+
+  float floating = boolean / 0.3f;
+  // CHECK-MESSAGES: :[[@LINE-1]]:20: warning: implicit conversion 'bool' -> 'float'
+  // CHECK-FIXES: float floating = (float)boolean / 0.3f;
+
+  char character = boolean;
+  // CHECK-MESSAGES: :[[@LINE-1]]:20: warning: implicit conversion 'bool' -> 'char'
+  // CHECK-FIXES: char character = (char)boolean;
+}
+
+void implicitConversionFromBoolInComplexBoolExpressions() {
+  bool boolean = true;
+  bool anotherBoolean = false;
+
+  int integer = boolean && anotherBoolean;
+  // CHECK-MESSAGES: :[[@LINE-1]]:17: warning: implicit conversion 'bool' -> 'int'
+  // CHECK-MESSAGES: :[[@LINE-2]]:28: warning: implicit conversion 'bool' -> 'int'
+  // CHECK-FIXES: int integer = (int)boolean && (int)anotherBoolean;
+
+  float floating = (boolean || anotherBoolean) * 0.3f;
+  // CHECK-MESSAGES: :[[@LINE-1]]:21: warning: implicit conversion 'bool' -> 'int'
+  // CHECK-MESSAGES: :[[@LINE-2]]:32: warning: implicit conversion 'bool' -> 'int'
+  // CHECK-FIXES: float floating = ((int)boolean || (int)anotherBoolean) * 0.3f;
+
+  double doubleFloating = (boolean && (anotherBoolean || boolean)) * 0.3;
+  // CHECK-MESSAGES: :[[@LINE-1]]:28: warning: implicit conversion 'bool' -> 'int'
+  // CHECK-MESSAGES: :[[@LINE-2]]:40: warning: implicit conversion 'bool' -> 'int'
+  // CHECK-MESSAGES: :[[@LINE-3]]:58: warning: implicit conversion 'bool' -> 'int'
+  // CHECK-FIXES: double doubleFloating = ((int)boolean && ((int)anotherBoolean || (int)boolean)) * 0.3;
+}
+
+void implicitConversionFromBoolLiterals() {
+  functionTakingInt(true);
+  // CHECK-MESSAGES: :[[@LINE-1]]:21: warning: implicit conversion 'bool' -> 'int'
+  // CHECK-FIXES: functionTakingInt(1);
+
+  functionTakingUnsignedLong(false);
+  // CHECK-MESSAGES: :[[@LINE-1]]:30: warning: implicit conversion 'bool' -> 'unsigned long'
+  // CHECK-FIXES: functionTakingUnsignedLong(0u);
+
+  functionTakingSignedChar(true);
+  // CHECK-MESSAGES: :[[@LINE-1]]:28: warning: implicit conversion 'bool' -> 'signed char'
+  // CHECK-FIXES: functionTakingSignedChar(1);
+
+  functionTakingFloat(false);
+  // CHECK-MESSAGES: :[[@LINE-1]]:23: warning: implicit conversion 'bool' -> 'float'
+  // CHECK-FIXES: functionTakingFloat(0.0f);
+
+  functionTakingDouble(true);
+  // CHECK-MESSAGES: :[[@LINE-1]]:24: warning: implicit conversion 'bool' -> 'double'
+  // CHECK-FIXES: functionTakingDouble(1.0);
+}
+
+void implicitConversionFromBoolInComparisons() {
+  bool boolean = true;
+  int integer = 0;
+
+  functionTakingBool(boolean == integer);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'bool' -> 'int'
+  // CHECK-FIXES: functionTakingBool((int)boolean == integer);
+
+  functionTakingBool(integer != boolean);
+  // CHECK-MESSAGES: :[[@LINE-1]]:33: warning: implicit conversion 'bool' -> 'int'
+  // CHECK-FIXES: functionTakingBool(integer != (int)boolean);
+}
+
+void ignoreBoolComparisons() {
+  bool boolean = true;
+  bool anotherBoolean = false;
+
+  functionTakingBool(boolean == anotherBoolean);
+  functionTakingBool(boolean != anotherBoolean);
+}
+
+void ignoreExplicitCastsFromBool() {
+  bool boolean = true;
+
+  int integer = (int)boolean + 3;
+  float floating = (float)boolean * 0.3f;
+  char character = (char)boolean;
+}
+
+void ignoreImplicitConversionFromBoolInMacroExpansions() {
+  bool boolean = true;
+
+  #define CAST_FROM_BOOL_IN_MACRO_BODY boolean + 3
+  int integerFromMacroBody = CAST_FROM_BOOL_IN_MACRO_BODY;
+
+  #define CAST_FROM_BOOL_IN_MACRO_ARGUMENT(x) x + 3
+  int integerFromMacroArgument = CAST_FROM_BOOL_IN_MACRO_ARGUMENT(boolean);
+}
+
+////////// Implicit conversions to bool.
+
+void implicitConversionToBoolSimpleCases() {
+  int integer = 10;
+  functionTakingBool(integer);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'int' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(integer != 0);
+
+  unsigned long unsignedLong = 10;
+  functionTakingBool(unsignedLong);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'unsigned long' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(unsignedLong != 0u);
+
+  float floating = 0.0f;
+  functionTakingBool(floating);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'float' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(floating != 0.0f);
+
+  double doubleFloating = 1.0f;
+  functionTakingBool(doubleFloating);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'double' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(doubleFloating != 0.0);
+
+  signed char character = 'a';
+  functionTakingBool(character);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'signed char' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(character != 0);
+
+  int* pointer = nullptr;
+  functionTakingBool(pointer);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'int *' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(pointer != nullptr);
+}
+
+void implicitConversionToBoolInSingleExpressions() {
+  int integer = 10;
+  bool boolComingFromInt;
+  boolComingFromInt = integer;
+  // CHECK-MESSAGES: :[[@LINE-1]]:23: warning: implicit conversion 'int' -> 'bool'
+  // CHECK-FIXES: boolComingFromInt = (integer != 0);
+
+  float floating = 10.0f;
+  bool boolComingFromFloat;
+  boolComingFromFloat = floating;
+  // CHECK-MESSAGES: :[[@LINE-1]]:25: warning: implicit conversion 'float' -> 'bool'
+  // CHECK-FIXES: boolComingFromFloat = (floating != 0.0f);
+
+  signed char character = 'a';
+  bool boolComingFromChar;
+  boolComingFromChar = character;
+  // CHECK-MESSAGES: :[[@LINE-1]]:24: warning: implicit conversion 'signed char' -> 'bool'
+  // CHECK-FIXES: boolComingFromChar = (character != 0);
+
+  int* pointer = nullptr;
+  bool boolComingFromPointer;
+  boolComingFromPointer = pointer;
+  // CHECK-MESSAGES: :[[@LINE-1]]:27: warning: implicit conversion 'int *' -> 'bool'
+  // CHECK-FIXES: boolComingFromPointer = (pointer != nullptr);
+}
+
+void implicitConversionToBoolInComplexExpressions() {
+  bool boolean = true;
+
+  int integer = 10;
+  int anotherInteger = 20;
+  bool boolComingFromInteger;
+  boolComingFromInteger = integer + anotherInteger;
+  // CHECK-MESSAGES: :[[@LINE-1]]:27: warning: implicit conversion 'int' -> 'bool'
+  // CHECK-FIXES: boolComingFromInteger = ((integer + anotherInteger) != 0);
+}
+
+void implicitConversionInNegationExpressions() {
+  int integer = 10;
+  bool boolComingFromNegatedInt;
+  boolComingFromNegatedInt = !integer;
+  // CHECK-MESSAGES: :[[@LINE-1]]:30: warning: implicit conversion 'int' -> 'bool'
+  // CHECK-FIXES: boolComingFromNegatedInt = ((!integer) != 0);
+}
+
+bool implicitConversionToBoolInReturnValue() {
+  float floating = 1.0f;
+  return floating;
+  // CHECK-MESSAGES: :[[@LINE-1]]:10: warning: implicit conversion 'float' -> 'bool'
+  // CHECK-FIXES: return floating != 0.0f;
+}
+
+void implicitConversionToBoolFromLiterals() {
+  functionTakingBool(0);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'int' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(false);
+
+  functionTakingBool(1);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'int' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(true);
+
+  functionTakingBool(2ul);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'unsigned long' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(true);
+
+  functionTakingBool(0.0f);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'float' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(false);
+
+  functionTakingBool(1.0f);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'float' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(true);
+
+  functionTakingBool(2.0);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'double' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(true);
+
+  functionTakingBool('\0');
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'int' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(false);
+
+  functionTakingBool('a');
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'int' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(true);
+
+  functionTakingBool("");
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'char *' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(true);
+
+  functionTakingBool("abc");
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'char *' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(true);
+
+  functionTakingBool(NULL);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'long' -> 'bool'
+  // CHECK-FIXES: functionTakingBool(false);
+}
+
+void implicitConversionToBoolFromUnaryMinusAndZeroLiterals() {
+  functionTakingBool(-0);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'int' -> 'bool'
+  // CHECK-FIXES: functionTakingBool((-0) != 0);
+
+  functionTakingBool(-0.0f);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'float' -> 'bool'
+  // CHECK-FIXES: functionTakingBool((-0.0f) != 0.0f);
+
+  functionTakingBool(-0.0);
+  // CHECK-MESSAGES: :[[@LINE-1]]:22: warning: implicit conversion 'double' -> 'bool'
+  // CHECK-FIXES: functionTakingBool((-0.0) != 0.0);
+}
+
+void ignoreExplicitCastsToBool() {
+  int integer = 10;
+  bool boolComingFromInt = (bool)integer;
+
+  float floating = 10.0f;
+  bool boolComingFromFloat = (bool)floating;
+
+  char character = 'a';
+  bool boolComingFromChar = (bool)character;
+
+  int* pointer = nullptr;
+  bool booleanComingFromPointer = (bool)pointer;
+}
+
+void ignoreImplicitConversionToBoolInMacroExpansions() {
+  int integer = 3;
+
+  #define CAST_TO_BOOL_IN_MACRO_BODY integer && false
+  bool boolFromMacroBody = CAST_TO_BOOL_IN_MACRO_BODY;
+
+  #define CAST_TO_BOOL_IN_MACRO_ARGUMENT(x) x || true
+  bool boolFromMacroArgument = CAST_TO_BOOL_IN_MACRO_ARGUMENT(integer);
+}
+
+int implicitConversionReturnInt()
+{
+    return true;
+    // CHECK-MESSAGES: :[[@LINE-1]]:12: warning: implicit conversion 'bool' -> 'int'
+    // CHECK-FIXES: return 1
+}
+
+int implicitConversionReturnIntWithParens()
+{
+    return (true);
+    // CHECK-MESSAGES: :[[@LINE-1]]:12: warning: implicit conversion 'bool' -> 'int'
+    // CHECK-FIXES: return 1
+}
+
+bool implicitConversionReturnBool()
+{
+    return 1;
+    // CHECK-MESSAGES: :[[@LINE-1]]:12: warning: implicit conversion 'int' -> 'bool'
+    // CHECK-FIXES: return true
+}
+
+bool implicitConversionReturnBoolWithParens()
+{
+    return (1);
+    // CHECK-MESSAGES: :[[@LINE-1]]:12: warning: implicit conversion 'int' -> 'bool'
+    // CHECK-FIXES: return true
+}
+
+int keepCompactReturnInC_PR71848() {
+  bool foo = false;
+  return( foo );
+// CHECK-MESSAGES: :[[@LINE-1]]:9: warning: implicit conversion 'bool' -> 'int' [readability-implicit-bool-conversion]
+// CHECK-FIXES: return(int)( foo );
+}
diff --git a/clang-tools-extra/unittests/CMakeLists.txt b/clang-tools-extra/unittests/CMakeLists.txt
@@ -1,5 +1,5 @@
 add_custom_target(ExtraToolsUnitTests)
-set_target_properties(ExtraToolsUnitTests PROPERTIES FOLDER "Extra Tools Unit Tests")
+set_target_properties(ExtraToolsUnitTests PROPERTIES FOLDER "Clang Tools Extra/Tests")
 
 function(add_extra_unittest test_dirname)
   add_unittest(ExtraToolsUnitTests ${test_dirname} ${ARGN})

diff --git a/clang/CMakeLists.txt b/clang/CMakeLists.txt
@@ -1,4 +1,5 @@
 cmake_minimum_required(VERSION 3.20.0)
+set(LLVM_SUBPROJECT_TITLE "Clang")
 
 if(NOT DEFINED LLVM_COMMON_CMAKE_UTILS)
   set(LLVM_COMMON_CMAKE_UTILS ${CMAKE_CURRENT_SOURCE_DIR}/../cmake)
@@ -349,10 +350,7 @@ if (LLVM_COMPILER_IS_GCC_COMPATIBLE)
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -pedantic -Wno-long-long")
   endif ()
 
-  check_cxx_compiler_flag("-Werror -Wnested-anon-types" CXX_SUPPORTS_NO_NESTED_ANON_TYPES_FLAG)
-  if( CXX_SUPPORTS_NO_NESTED_ANON_TYPES_FLAG )
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-nested-anon-types" )
-  endif()
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-nested-anon-types" )
 endif ()
 
 # Determine HOST_LINK_VERSION on Darwin.
@@ -394,14 +392,15 @@ if (NOT LLVM_INSTALL_TOOLCHAIN_ONLY)
   # Installing the headers needs to depend on generating any public
   # tablegen'd headers.
   add_custom_target(clang-headers DEPENDS clang-tablegen-targets)
-  set_target_properties(clang-headers PROPERTIES FOLDER "Misc")
+  set_target_properties(clang-headers PROPERTIES FOLDER "Clang/Resources")
   if(NOT LLVM_ENABLE_IDE)
     add_llvm_install_targets(install-clang-headers
                              DEPENDS clang-headers
                              COMPONENT clang-headers)
   endif()
 
   add_custom_target(bash-autocomplete DEPENDS utils/bash-autocomplete.sh)
+  set_target_properties(bash-autocomplete PROPERTIES FOLDER "Clang/Misc")
   install(FILES utils/bash-autocomplete.sh
           DESTINATION "${CMAKE_INSTALL_DATADIR}/clang"
           COMPONENT bash-autocomplete)
@@ -482,7 +481,7 @@ add_custom_target(clang-tablegen-targets
   omp_gen
   ClangDriverOptions
   ${CLANG_TABLEGEN_TARGETS})
-set_target_properties(clang-tablegen-targets PROPERTIES FOLDER "Misc")
+set_target_properties(clang-tablegen-targets PROPERTIES FOLDER "Clang/Tablegenning/Targets")
 list(APPEND LLVM_COMMON_DEPENDS clang-tablegen-targets)
 
 # Force target to be built as soon as possible. Clang modules builds depend
@@ -547,7 +546,7 @@ endif()
 
 # Custom target to install all clang libraries.
 add_custom_target(clang-libraries)
-set_target_properties(clang-libraries PROPERTIES FOLDER "Misc")
+set_target_properties(clang-libraries PROPERTIES FOLDER "Clang/Install")
 
 if(NOT LLVM_ENABLE_IDE)
   add_llvm_install_targets(install-clang-libraries

diff --git a/clang/bindings/python/tests/CMakeLists.txt b/clang/bindings/python/tests/CMakeLists.txt
@@ -11,7 +11,7 @@ add_custom_target(check-clang-python
     WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}/..)
 
 set(RUN_PYTHON_TESTS TRUE)
-set_target_properties(check-clang-python PROPERTIES FOLDER "Clang tests")
+set_target_properties(check-clang-python PROPERTIES FOLDER "Clang/Tests")
 
 # Tests require libclang.so which is only built with LLVM_ENABLE_PIC=ON
 if(NOT LLVM_ENABLE_PIC)

diff --git a/clang/cmake/caches/CrossWinToARMLinux.cmake b/clang/cmake/caches/CrossWinToARMLinux.cmake
@@ -89,6 +89,13 @@ endif()
 
 message(STATUS "Toolchain target to build: ${LLVM_TARGETS_TO_BUILD}")
 
+# Allow to override libc++ ABI version. Use 2 by default.
+if (NOT DEFINED LIBCXX_ABI_VERSION)
+  set(LIBCXX_ABI_VERSION 2)
+endif()
+
+message(STATUS "Toolchain's Libc++ ABI version: ${LIBCXX_ABI_VERSION}")
+
 if (NOT DEFINED CMAKE_BUILD_TYPE)
   set(CMAKE_BUILD_TYPE "Release" CACHE STRING "")
 endif()
@@ -109,8 +116,15 @@ set(CLANG_DEFAULT_OBJCOPY                   "llvm-objcopy" CACHE STRING "")
 set(CLANG_DEFAULT_RTLIB                     "compiler-rt" CACHE STRING "")
 set(CLANG_DEFAULT_UNWINDLIB                 "libunwind" CACHE STRING "")
 
-if(WIN32)
-  set(CMAKE_MSVC_RUNTIME_LIBRARY            "MultiThreaded" CACHE STRING "")
+if (NOT DEFINED CMAKE_MSVC_RUNTIME_LIBRARY AND WIN32)
+  #Note: Always specify MT DLL for the LLDB build configurations on Windows host.
+  if (CMAKE_BUILD_TYPE STREQUAL "Debug")
+    set(CMAKE_MSVC_RUNTIME_LIBRARY            "MultiThreadedDebugDLL" CACHE STRING "")
+  else()
+    set(CMAKE_MSVC_RUNTIME_LIBRARY            "MultiThreadedDLL" CACHE STRING "")
+  endif()
+  # Grab all ucrt/vcruntime related DLLs into the binary installation folder.
+  set(CMAKE_INSTALL_UCRT_LIBRARIES          ON CACHE BOOL "")
 endif()
 
 # Set up RPATH for the target runtime/builtin libraries.
@@ -127,6 +141,15 @@ set(BUILTINS_${TOOLCHAIN_TARGET_TRIPLE}_CMAKE_INSTALL_RPATH
 set(BUILTINS_${TOOLCHAIN_TARGET_TRIPLE}_CMAKE_BUILD_WITH_INSTALL_RPATH            ON  CACHE BOOL "")
 set(BUILTINS_${TOOLCHAIN_TARGET_TRIPLE}_LLVM_CMAKE_DIR                            "${LLVM_PROJECT_DIR}/llvm/cmake/modules" CACHE PATH "")
 
+if (DEFINED TOOLCHAIN_TARGET_COMPILER_FLAGS)
+  foreach(lang C;CXX;ASM)
+    set(BUILTINS_${TOOLCHAIN_TARGET_TRIPLE}_CMAKE_${lang}_FLAGS         "${TOOLCHAIN_TARGET_COMPILER_FLAGS}" CACHE STRING "")
+  endforeach()
+endif()
+foreach(type SHARED;MODULE;EXE)
+  set(BUILTINS_${TOOLCHAIN_TARGET_TRIPLE}_CMAKE_${type}_LINKER_FLAGS    "-fuse-ld=lld" CACHE STRING "")
+endforeach()
+
 set(LLVM_RUNTIME_TARGETS                    "${TOOLCHAIN_TARGET_TRIPLE}" CACHE STRING "")
 set(LLVM_ENABLE_PER_TARGET_RUNTIME_DIR      ON CACHE BOOL "")
 
@@ -137,6 +160,15 @@ set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_CMAKE_SYSROOT
 set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_CMAKE_INSTALL_RPATH                       "${RUNTIMES_INSTALL_RPATH}"  CACHE STRING "")
 set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_CMAKE_BUILD_WITH_INSTALL_RPATH            ON  CACHE BOOL "")
 
+if (DEFINED TOOLCHAIN_TARGET_COMPILER_FLAGS)
+  foreach(lang C;CXX;ASM)
+    set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_CMAKE_${lang}_FLAGS         "${TOOLCHAIN_TARGET_COMPILER_FLAGS}" CACHE STRING "")
+  endforeach()
+endif()
+foreach(type SHARED;MODULE;EXE)
+  set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_CMAKE_${type}_LINKER_FLAGS    "-fuse-ld=lld" CACHE STRING "")
+endforeach()
+
 set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_COMPILER_RT_BUILD_BUILTINS                ON CACHE BOOL "")
 set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_COMPILER_RT_BUILD_SANITIZERS              OFF CACHE BOOL "")
 set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_COMPILER_RT_BUILD_XRAY                    OFF CACHE BOOL "")
@@ -164,7 +196,7 @@ set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_LIBCXXABI_ENABLE_SHARED
 
 set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_LIBCXX_USE_COMPILER_RT                    ON CACHE BOOL "")
 set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_LIBCXX_ENABLE_SHARED                      OFF CACHE BOOL "")
-set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_LIBCXX_ABI_VERSION                        2 CACHE STRING "")
+set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_LIBCXX_ABI_VERSION                        ${LIBCXX_ABI_VERSION} CACHE STRING "")
 set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_LIBCXX_CXX_ABI                            "libcxxabi" CACHE STRING "")    #!!!
 set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_LIBCXX_ENABLE_NEW_DELETE_DEFINITIONS      ON CACHE BOOL "")
 

diff --git a/clang/cmake/caches/Fuchsia-stage2.cmake b/clang/cmake/caches/Fuchsia-stage2.cmake
@@ -19,7 +19,6 @@ set(LLVM_ENABLE_LLD ON CACHE BOOL "")
 set(LLVM_ENABLE_LTO ON CACHE BOOL "")
 set(LLVM_ENABLE_PER_TARGET_RUNTIME_DIR ON CACHE BOOL "")
 set(LLVM_ENABLE_PLUGINS OFF CACHE BOOL "")
-set(LLVM_ENABLE_TERMINFO OFF CACHE BOOL "")
 set(LLVM_ENABLE_UNWIND_TABLES OFF CACHE BOOL "")
 set(LLVM_ENABLE_Z3_SOLVER OFF CACHE BOOL "")
 set(LLVM_ENABLE_ZLIB ON CACHE BOOL "")

diff --git a/clang/cmake/caches/Fuchsia.cmake b/clang/cmake/caches/Fuchsia.cmake
@@ -12,7 +12,6 @@ set(LLVM_ENABLE_DIA_SDK OFF CACHE BOOL "")
 set(LLVM_ENABLE_LIBEDIT OFF CACHE BOOL "")
 set(LLVM_ENABLE_LIBXML2 OFF CACHE BOOL "")
 set(LLVM_ENABLE_PER_TARGET_RUNTIME_DIR ON CACHE BOOL "")
-set(LLVM_ENABLE_TERMINFO OFF CACHE BOOL "")
 set(LLVM_ENABLE_UNWIND_TABLES OFF CACHE BOOL "")
 set(LLVM_ENABLE_Z3_SOLVER OFF CACHE BOOL "")
 set(LLVM_ENABLE_ZLIB OFF CACHE BOOL "")
@@ -34,7 +33,6 @@ set(_FUCHSIA_BOOTSTRAP_PASSTHROUGH
   LibXml2_ROOT
   LLVM_ENABLE_CURL
   LLVM_ENABLE_HTTPLIB
-  LLVM_ENABLE_TERMINFO
   LLVM_ENABLE_LIBEDIT
   CURL_ROOT
   OpenSSL_ROOT
@@ -47,11 +45,6 @@ set(_FUCHSIA_BOOTSTRAP_PASSTHROUGH
   CURSES_LIBRARIES
   PANEL_LIBRARIES
 
-  # Deprecated
-  Terminfo_ROOT
-
-  Terminfo_LIBRARIES
-
   # Deprecated
   LibEdit_ROOT
-Original file line number
+Diff line change
@@ Expand Up / @@ -598,8 +598,6 @@ void DataReader::readSampleData(BinaryFunction &BF) { @@
       }
       BF.ExecutionCount = TotalEntryCount;
-      estimateEdgeCounts(BF);
     }
     void DataReader::convertBranchData(BinaryFunction &BF) const {
@@ Expand Down @@