Merge branch 'main' into pr-riscv-shuffle-vcompress

Author: preames

bolt/include/bolt/Profile/DataAggregator.h

@@ -170,9 +170,6 @@ class DataAggregator : public DataReader {
   std::string BuildIDBinaryName;
 
   /// Memory map info for a single file as recorded in perf.data
-  /// When a binary has multiple text segments, the Size is computed as the
-  /// difference of the last address of these segments from the BaseAddress.
-  /// The base addresses of all text segments must be the same.
   struct MMapInfo {
     uint64_t BaseAddress{0}; /// Base address of the mapped binary.
     uint64_t MMapAddress{0}; /// Address of the executable segment.
@@ -496,11 +493,6 @@ class DataAggregator : public DataReader {
   /// and return a file name matching a given \p FileBuildID.
   std::optional<StringRef> getFileNameForBuildID(StringRef FileBuildID);
 
-  /// Get a constant reference to the parsed binary mmap entries.
-  const std::unordered_map<uint64_t, MMapInfo> &getBinaryMMapInfo() {
-    return BinaryMMapInfo;
-  }
-
   friend class YAMLProfileWriter;
 };
 } // namespace bolt

bolt/lib/Profile/DataAggregator.cpp

@@ -95,12 +95,6 @@ cl::opt<bool> ReadPreAggregated(
     "pa", cl::desc("skip perf and read data from a pre-aggregated file format"),
     cl::cat(AggregatorCategory));
 
-cl::opt<std::string>
-    ReadPerfEvents("perf-script-events",
-                   cl::desc("skip perf event collection by supplying a "
-                            "perf-script output in a textual format"),
-                   cl::ReallyHidden, cl::init(""), cl::cat(AggregatorCategory));
-
 static cl::opt<bool>
 TimeAggregator("time-aggr",
   cl::desc("time BOLT aggregator"),
@@ -173,9 +167,8 @@ void DataAggregator::findPerfExecutable() {
 void DataAggregator::start() {
   outs() << "PERF2BOLT: Starting data aggregation job for " << Filename << "\n";
 
-  // Don't launch perf for pre-aggregated files or when perf input is specified
-  // by the user.
-  if (opts::ReadPreAggregated || !opts::ReadPerfEvents.empty())
+  // Don't launch perf for pre-aggregated files
+  if (opts::ReadPreAggregated)
     return;
 
   findPerfExecutable();
@@ -471,13 +464,6 @@ void DataAggregator::filterBinaryMMapInfo() {
 
 int DataAggregator::prepareToParse(StringRef Name, PerfProcessInfo &Process,
                                    PerfProcessErrorCallbackTy Callback) {
-  if (!opts::ReadPerfEvents.empty()) {
-    outs() << "PERF2BOLT: using pre-processed perf events for '" << Name
-           << "' (perf-script-events)\n";
-    ParsingBuf = opts::ReadPerfEvents;
-    return 0;
-  }
-
   std::string Error;
   outs() << "PERF2BOLT: waiting for perf " << Name
          << " collection to finish...\n";
@@ -2070,6 +2056,15 @@ std::error_code DataAggregator::parseMMapEvents() {
     if (FileMMapInfo.first == "(deleted)")
       continue;
 
+    // Consider only the first mapping of the file for any given PID
+    auto Range = GlobalMMapInfo.equal_range(FileMMapInfo.first);
+    bool PIDExists = llvm::any_of(make_range(Range), [&](const auto &MI) {
+      return MI.second.PID == FileMMapInfo.second.PID;
+    });
+
+    if (PIDExists)
+      continue;
+
     GlobalMMapInfo.insert(FileMMapInfo);
   }
 
@@ -2121,22 +2116,12 @@ std::error_code DataAggregator::parseMMapEvents() {
                << " using file offset 0x" << Twine::utohexstr(MMapInfo.Offset)
                << ". Ignoring profile data for this mapping\n";
         continue;
+      } else {
+        MMapInfo.BaseAddress = *BaseAddress;
       }
-      MMapInfo.BaseAddress = *BaseAddress;
     }
 
-    // Try to add MMapInfo to the map and update its size. Large binaries may
-    // span to multiple text segments, so the mapping is inserted only on the
-    // first occurrence.
-    if (!BinaryMMapInfo.insert(std::make_pair(MMapInfo.PID, MMapInfo)).second)
-      assert(MMapInfo.BaseAddress == BinaryMMapInfo[MMapInfo.PID].BaseAddress &&
-             "Base address on multiple segment mappings should match");
-
-    // Update mapping size.
-    const uint64_t EndAddress = MMapInfo.MMapAddress + MMapInfo.Size;
-    const uint64_t Size = EndAddress - BinaryMMapInfo[MMapInfo.PID].BaseAddress;
-    if (Size > BinaryMMapInfo[MMapInfo.PID].Size)
-      BinaryMMapInfo[MMapInfo.PID].Size = Size;
+    BinaryMMapInfo.insert(std::make_pair(MMapInfo.PID, MMapInfo));
   }
 
   if (BinaryMMapInfo.empty()) {

bolt/unittests/Core/CMakeLists.txt

@@ -8,7 +8,6 @@ set(LLVM_LINK_COMPONENTS
 add_bolt_unittest(CoreTests
   BinaryContext.cpp
   MCPlusBuilder.cpp
-  MemoryMaps.cpp
   DynoStats.cpp
 
   DISABLE_LLVM_LINK_LLVM_DYLIB
@@ -18,8 +17,6 @@ target_link_libraries(CoreTests
   PRIVATE
   LLVMBOLTCore
   LLVMBOLTRewrite
-  LLVMBOLTProfile
-  LLVMTestingSupport
   )
 
 foreach (tgt ${BOLT_TARGETS_TO_BUILD})

bolt/unittests/Core/MemoryMaps.cpp

@@ -467,18 +467,9 @@ class DefineOutline : public Tweak {
       }
     }
 
-    // For function templates, the same limitations as for class templates
-    // apply.
-    if (const TemplateParameterList *Params =
-            MD->getDescribedTemplateParams()) {
-      // FIXME: Is this really needed? It inhibits application on
-      //        e.g. std::enable_if.
-      for (NamedDecl *P : *Params) {
-        if (!P->getIdentifier())
-          return false;
-      }
+    // Function templates must be defined in the same file.
+    if (MD->getDescribedTemplate())
       SameFile = true;
-    }
 
     // The refactoring is meaningless for unnamed classes and namespaces,
     // unless we're outlining in the same file

clang-tools-extra/clangd/refactor/tweaks/DefineOutline.cpp

@@ -118,12 +118,6 @@ TEST_F(DefineOutlineTest, TriggersOnFunctionDecl) {
     template <> void fo^o<int>() {}
   )cpp");
 
-  // Not available on member function templates with unnamed template
-  // parameters.
-  EXPECT_UNAVAILABLE(R"cpp(
-    struct Foo { template <typename> void ba^r() {} };
-  )cpp");
-
   // Not available on methods of unnamed classes.
   EXPECT_UNAVAILABLE(R"cpp(
     struct Foo {
@@ -410,14 +404,14 @@ inline typename O1<T, U...>::template O2<V, A>::E O1<T, U...>::template O2<V, A>
       {
           R"cpp(
             struct Foo {
-              template <typename T, bool B = true>
+              template <typename T, typename, bool B = true>
               T ^bar() { return {}; }
             };)cpp",
           R"cpp(
             struct Foo {
-              template <typename T, bool B = true>
+              template <typename T, typename, bool B = true>
               T bar() ;
-            };template <typename T, bool B>
+            };template <typename T, typename, bool B>
 inline T Foo::bar() { return {}; }
 )cpp",
           ""},
@@ -426,13 +420,13 @@ inline T Foo::bar() { return {}; }
       {
           R"cpp(
             template <typename T> struct Foo {
-              template <typename U> T ^bar(const T& t, const U& u) { return {}; }
+              template <typename U, bool> T ^bar(const T& t, const U& u) { return {}; }
             };)cpp",
           R"cpp(
             template <typename T> struct Foo {
-              template <typename U> T bar(const T& t, const U& u) ;
+              template <typename U, bool> T bar(const T& t, const U& u) ;
             };template <typename T>
-template <typename U>
+template <typename U, bool>
 inline T Foo<T>::bar(const T& t, const U& u) { return {}; }
 )cpp",
           ""},

clang-tools-extra/clangd/unittests/tweaks/DefineOutlineTests.cpp

@@ -73,6 +73,8 @@ Hover
 Code completion
 ^^^^^^^^^^^^^^^
 
+- Added completion for C++20 keywords.
+
 Code actions
 ^^^^^^^^^^^^
 

clang-tools-extra/docs/ReleaseNotes.rst

@@ -647,6 +647,9 @@ elementwise to the input.
 
 Unless specified otherwise operation(±0) = ±0 and operation(±infinity) = ±infinity
 
+The integer elementwise intrinsics, including ``__builtin_elementwise_popcount``,
+can be called in a ``constexpr`` context.
+
 ============================================== ====================================================================== =========================================
          Name                                   Operation                                                             Supported element types
 ============================================== ====================================================================== =========================================

clang/docs/LanguageExtensions.rst

@@ -372,6 +372,7 @@ Non-comprehensive list of changes in this release
 - ``__builtin_reduce_mul`` function can now be used in constant expressions.
 - ``__builtin_reduce_and`` function can now be used in constant expressions.
 - ``__builtin_reduce_or`` and ``__builtin_reduce_xor`` functions can now be used in constant expressions.
+- ``__builtin_elementwise_popcount`` function can now be used in constant expressions.
 
 New Compiler Flags
 ------------------
@@ -587,6 +588,8 @@ Improvements to Clang's diagnostics
 - For an rvalue reference bound to a temporary struct with an integer member, Clang will detect constant integer overflow
   in the initializer for the integer member (#GH46755).
 
+- Fixed a false negative ``-Wunused-private-field`` diagnostic when a defaulted comparison operator is defined out of class (#GH116961).
+
 Improvements to Clang's time-trace
 ----------------------------------
 
@@ -728,6 +731,8 @@ Bug Fixes to AST Handling
   sometimes incorrectly return null even if a comment was present. (#GH108145)
 - Clang now correctly parses the argument of the ``relates``, ``related``, ``relatesalso``,
   and ``relatedalso`` comment commands.
+- Clang now uses the location of the begin of the member expression for ``CallExpr``
+  involving deduced ``this``. (#GH116928)
 
 Miscellaneous Bug Fixes
 ^^^^^^^^^^^^^^^^^^^^^^^