Merge remote-tracking branch 'origin/main' into vplan-remove-region-before-cg

Author: fhahn

bolt/include/bolt/Profile/DataReader.h

@@ -246,6 +246,9 @@ struct FuncSampleData {
   /// Get the number of samples recorded in [Start, End)
   uint64_t getSamples(uint64_t Start, uint64_t End) const;
 
+  /// Returns the total number of samples recorded in this function.
+  uint64_t getSamples() const;
+
   /// Aggregation helper
   DenseMap<uint64_t, size_t> Index;
 

bolt/lib/Profile/DataAggregator.cpp

@@ -565,15 +565,14 @@ void DataAggregator::processProfile(BinaryContext &BC) {
   processMemEvents();
 
   // Mark all functions with registered events as having a valid profile.
-  const auto Flags = opts::BasicAggregation ? BinaryFunction::PF_SAMPLE
-                                            : BinaryFunction::PF_LBR;
   for (auto &BFI : BC.getBinaryFunctions()) {
     BinaryFunction &BF = BFI.second;
-    FuncBranchData *FBD = getBranchData(BF);
-    if (FBD || getFuncSampleData(BF.getNames())) {
-      BF.markProfiled(Flags);
-      if (FBD)
-        BF.RawBranchCount = FBD->getNumExecutedBranches();
+    if (FuncBranchData *FBD = getBranchData(BF)) {
+      BF.markProfiled(BinaryFunction::PF_LBR);
+      BF.RawBranchCount = FBD->getNumExecutedBranches();
+    } else if (FuncSampleData *FSD = getFuncSampleData(BF.getNames())) {
+      BF.markProfiled(BinaryFunction::PF_SAMPLE);
+      BF.RawBranchCount = FSD->getSamples();
     }
   }
 
@@ -630,10 +629,18 @@ StringRef DataAggregator::getLocationName(const BinaryFunction &Func,
 
 bool DataAggregator::doSample(BinaryFunction &OrigFunc, uint64_t Address,
                               uint64_t Count) {
+  // To record executed bytes, use basic block size as is regardless of BAT.
+  uint64_t BlockSize = 0;
+  if (BinaryBasicBlock *BB = OrigFunc.getBasicBlockContainingOffset(
+          Address - OrigFunc.getAddress()))
+    BlockSize = BB->getOriginalSize();
+
   BinaryFunction *ParentFunc = getBATParentFunction(OrigFunc);
   BinaryFunction &Func = ParentFunc ? *ParentFunc : OrigFunc;
-  if (ParentFunc || (BAT && !BAT->isBATFunction(OrigFunc.getAddress())))
+  if (ParentFunc || (BAT && !BAT->isBATFunction(Func.getAddress())))
     NumColdSamples += Count;
+  // Attach executed bytes to parent function in case of cold fragment.
+  Func.SampleCountInBytes += Count * BlockSize;
 
   auto I = NamesToSamples.find(Func.getOneName());
   if (I == NamesToSamples.end()) {

bolt/lib/Profile/DataReader.cpp

@@ -128,6 +128,13 @@ uint64_t FuncSampleData::getSamples(uint64_t Start, uint64_t End) const {
   return Result;
 }
 
+uint64_t FuncSampleData::getSamples() const {
+  uint64_t Result = 0;
+  for (const SampleInfo &I : Data)
+    Result += I.Hits;
+  return Result;
+}
+
 void FuncSampleData::bumpCount(uint64_t Offset, uint64_t Count) {
   auto Iter = Index.find(Offset);
   if (Iter == Index.end()) {
@@ -1028,9 +1035,8 @@ ErrorOr<SampleInfo> DataReader::parseSampleInfo() {
 }
 
 ErrorOr<bool> DataReader::maybeParseNoLBRFlag() {
-  if (ParsingBuf.size() < 6 || ParsingBuf.substr(0, 6) != "no_lbr")
+  if (!ParsingBuf.consume_front("no_lbr"))
     return false;
-  ParsingBuf = ParsingBuf.drop_front(6);
   Col += 6;
 
   if (ParsingBuf.size() > 0 && ParsingBuf[0] == ' ')
@@ -1051,9 +1057,8 @@ ErrorOr<bool> DataReader::maybeParseNoLBRFlag() {
 }
 
 ErrorOr<bool> DataReader::maybeParseBATFlag() {
-  if (ParsingBuf.size() < 16 || ParsingBuf.substr(0, 16) != "boltedcollection")
+  if (!ParsingBuf.consume_front("boltedcollection"))
     return false;
-  ParsingBuf = ParsingBuf.drop_front(16);
   Col += 16;
 
   if (!checkAndConsumeNewLine()) {

bolt/lib/Rewrite/BuildIDRewriter.cpp

@@ -78,8 +78,7 @@ Error BuildIDRewriter::sectionInitializer() {
                                  "out of bounds while reading note section: %s",
                                  toString(Cursor.takeError()).c_str());
 
-      if (Type == ELF::NT_GNU_BUILD_ID && Name.substr(0, 3) == "GNU" &&
-          DescSz) {
+      if (Type == ELF::NT_GNU_BUILD_ID && Name.starts_with("GNU") && DescSz) {
         BuildIDSection = NoteSection;
         BuildID = Desc;
         BC.setFileBuildID(getPrintableBuildID(Desc));

bolt/test/perf2bolt/perf_test.test

@@ -8,6 +8,7 @@ RUN: perf2bolt %t -p=%t2 -o %t3 -nl -ignore-build-id 2>&1 | FileCheck %s
 
 CHECK-NOT: PERF2BOLT-ERROR
 CHECK-NOT: !! WARNING !! This high mismatch ratio indicates the input binary is probably not the same binary used during profiling collection.
+CHECK: BOLT-INFO: Functions with density >= {{.*}} account for 99.00% total sample counts.
 
 RUN: %clang %S/Inputs/perf_test.c -no-pie -fuse-ld=lld -o %t4
 RUN: perf record -Fmax -e cycles:u -o %t5 -- %t4

clang-tools-extra/clang-tidy/utils/IncludeSorter.cpp

@@ -19,9 +19,8 @@ namespace {
 
 StringRef removeFirstSuffix(StringRef Str, ArrayRef<const char *> Suffixes) {
   for (StringRef Suffix : Suffixes) {
-    if (Str.ends_with(Suffix)) {
-      return Str.substr(0, Str.size() - Suffix.size());
-    }
+    if (Str.consume_back(Suffix))
+      return Str;
   }
   return Str;
 }

clang-tools-extra/clangd/CompileCommands.cpp

@@ -404,8 +404,7 @@ enum DriverMode : unsigned char {
 DriverMode getDriverMode(const std::vector<std::string> &Args) {
   DriverMode Mode = DM_GCC;
   llvm::StringRef Argv0 = Args.front();
-  if (Argv0.ends_with_insensitive(".exe"))
-    Argv0 = Argv0.drop_back(strlen(".exe"));
+  Argv0.consume_back_insensitive(".exe");
   if (Argv0.ends_with_insensitive("cl"))
     Mode = DM_CL;
   for (const llvm::StringRef Arg : Args) {

clang-tools-extra/docs/clang-tidy/checks/bugprone/crtp-constructor-accessibility.rst

@@ -9,7 +9,7 @@ can be constructed outside itself and the derived class.
 The CRTP is an idiom, in which a class derives from a template class, where 
 itself is the template argument. It should be ensured that if a class is
 intended to be a base class in this idiom, it can only be instantiated if
-the derived class is it's template argument.
+the derived class is its template argument.
 
 Example:
 

clang-tools-extra/modularize/Modularize.cpp

@@ -406,11 +406,8 @@ struct Location {
   }
 
   friend bool operator<(const Location &X, const Location &Y) {
-    if (X.File != Y.File)
-      return X.File < Y.File;
-    if (X.Line != Y.Line)
-      return X.Line < Y.Line;
-    return X.Column < Y.Column;
+    return std::tie(X.File, X.Line, X.Column) <
+           std::tie(Y.File, Y.Line, Y.Column);
   }
   friend bool operator>(const Location &X, const Location &Y) { return Y < X; }
   friend bool operator<=(const Location &X, const Location &Y) {

clang-tools-extra/modularize/PreprocessorTracker.cpp

@@ -494,19 +494,8 @@ class PPItemKey {
     return Column == Other.Column;
   }
   bool operator<(const PPItemKey &Other) const {
-    if (Name < Other.Name)
-      return true;
-    else if (Name > Other.Name)
-      return false;
-    if (File < Other.File)
-      return true;
-    else if (File > Other.File)
-      return false;
-    if (Line < Other.Line)
-      return true;
-    else if (Line > Other.Line)
-      return false;
-    return Column < Other.Column;
+    return std::tie(Name, File, Line, Column) <
+           std::tie(Other.Name, Other.File, Other.Line, Other.Column);
   }
   StringHandle Name;
   HeaderHandle File;