Merge pull request #149 from sx-aurora-dev/feature/merge-upstream-20220202

Feature/merge upstream 20220202

Author: kaz7

.clang-tidy

@@ -6,6 +6,10 @@ CheckOptions:
     value:           CamelCase
   - key:             readability-identifier-naming.FunctionCase
     value:           camelBack
+  # Exclude from scanning as this is an exported symbol used for fuzzing
+  # throughout the code base.
+  - key:             readability-identifier-naming.FunctionIgnoredRegexp
+    value:           "LLVMFuzzerTestOneInput"
   - key:             readability-identifier-naming.MemberCase
     value:           CamelCase
   - key:             readability-identifier-naming.ParameterCase

.github/workflows/closed-issues.yml

@@ -0,0 +1,13 @@
+name: Labeling closed issues
+on:
+  issues:
+    types: ['closed']
+
+jobs:
+  automate-issues-labels:
+    runs-on: ubuntu-latest
+    if: github.repository == 'llvm/llvm-project'
+    steps:
+      - uses: andymckay/labeler@1.0.4
+        with:
+          remove-labels: "awaiting-review"

bolt/CMakeLists.txt

@@ -91,6 +91,10 @@ add_subdirectory(lib)
 add_subdirectory(tools)
 
 if (BOLT_INCLUDE_TESTS)
+  if (EXISTS ${LLVM_MAIN_SRC_DIR}/utils/unittest/googletest/include/gtest/gtest.h)
+    add_subdirectory(unittests)
+    list(APPEND BOLT_TEST_DEPS BoltUnitTests)
+  endif()
   add_subdirectory(test)
 endif()
 

bolt/README.md

@@ -30,6 +30,14 @@ compiler option. Since GCC8 enables this option by default, you have to
 explicitly disable it by adding `-fno-reorder-blocks-and-partition` flag if
 you are compiling with GCC8 or above.
 
+NOTE2: DWARF v5 is the new debugging format generated by the latest LLVM and GCC
+compilers. It offers several benefits over the previous DWARF v4. Currently, the
+support for v5 is a work in progress for BOLT. While you will be able to
+optimize binaries produced by the latest compilers, until the support is
+complete, you will not be able to update the debug info with
+`-update-debug-sections`. To temporarily work around the issue, we recommend
+compiling binaries with `-gdwarf-4` option that forces DWARF v4 output.
+
 PIE and .so support has been added recently. Please report bugs if you
 encounter any issues.
 

bolt/include/bolt/Core/DebugData.h

@@ -107,6 +107,13 @@ struct DebugLineTableRowRef {
 /// Common buffer vector used for debug info handling.
 using DebugBufferVector = SmallVector<char, 16>;
 
+/// Map of old CU offset to new offset and length.
+struct CUInfo {
+  uint32_t Offset;
+  uint32_t Length;
+};
+using CUOffsetMap = std::map<uint32_t, CUInfo>;
+
 /// Serializes the .debug_ranges DWARF section.
 class DebugRangesSectionWriter {
 public:
@@ -155,9 +162,8 @@ class DebugARangesSectionWriter {
   /// Writes .debug_aranges with the added ranges to the MCObjectWriter.
   /// Takes in \p RangesStream to write into, and \p CUMap which maps CU
   /// original offsets to new ones.
-  void
-  writeARangesSection(raw_svector_ostream &RangesStream,
-                      const std::unordered_map<uint32_t, uint32_t> CUMap) const;
+  void writeARangesSection(raw_svector_ostream &RangesStream,
+                           const CUOffsetMap &CUMap) const;
 
   /// Resets the writer to a clear state.
   void reset() { CUAddressRanges.clear(); }
@@ -647,8 +653,8 @@ class DebugInfoBinaryPatcher : public SimpleBinaryPatcher {
   void setDWPOffset(uint64_t DWPOffset) { DWPUnitOffset = DWPOffset; }
 
   /// When this function is invoked all of the DebugInfo Patches must be done.
-  /// Returns a map of old CU offsets to new ones.
-  std::unordered_map<uint32_t, uint32_t> computeNewOffsets();
+  /// Returns a map of old CU offsets to new offsets and new sizes.
+  CUOffsetMap computeNewOffsets(DWARFContext &DWCtx, bool IsDWOContext);
 
 private:
   struct PatchDeleter {
@@ -685,7 +691,7 @@ class DebugInfoBinaryPatcher : public SimpleBinaryPatcher {
   using UniquePatchPtrType = std::unique_ptr<Patch, PatchDeleter>;
 
   uint64_t DWPUnitOffset{0};
-  uint32_t ChangeInSize{0};
+  int32_t ChangeInSize{0};
   std::vector<UniquePatchPtrType> DebugPatches;
   /// Mutex used for parallel processing of debug info.
   std::mutex WriterMutex;
@@ -709,8 +715,11 @@ class DebugAbbrevWriter {
     std::unique_ptr<DebugBufferVector> Buffer;
     std::unique_ptr<raw_svector_ostream> Stream;
   };
-  /// Map original unit abbrev offset to abbreviations data.
-  std::map<uint64_t, AbbrevData> UnitsAbbrevData;
+  /// Map original unit to abbreviations data.
+  std::unordered_map<const DWARFUnit *, AbbrevData *> UnitsAbbrevData;
+
+  /// Map from Hash Signature to AbbrevData.
+  llvm::StringMap<std::unique_ptr<AbbrevData>> AbbrevDataCache;
 
   /// Attributes substitution (patch) information.
   struct PatchInfo {
@@ -777,10 +786,8 @@ class DebugAbbrevWriter {
   /// Return an offset in the finalized abbrev section corresponding to CU/TU.
   uint64_t getAbbreviationsOffsetForUnit(const DWARFUnit &Unit) {
     assert(!DWOId && "offsets are tracked for non-DWO units only");
-    assert(UnitsAbbrevData.find(Unit.getAbbreviationsOffset()) !=
-               UnitsAbbrevData.end() &&
-           "no abbrev data found for unit");
-    return UnitsAbbrevData[Unit.getAbbreviationsOffset()].Offset;
+    assert(UnitsAbbrevData.count(&Unit) && "no abbrev data found for unit");
+    return UnitsAbbrevData[&Unit]->Offset;
   }
 };
 

bolt/include/bolt/Core/MCPlusBuilder.h

@@ -1898,6 +1898,14 @@ class MCPlusBuilder {
   }
 };
 
+MCPlusBuilder *createX86MCPlusBuilder(const MCInstrAnalysis *,
+                                      const MCInstrInfo *,
+                                      const MCRegisterInfo *);
+
+MCPlusBuilder *createAArch64MCPlusBuilder(const MCInstrAnalysis *,
+                                          const MCInstrInfo *,
+                                          const MCRegisterInfo *);
+
 } // namespace bolt
 } // namespace llvm
 

bolt/include/bolt/Rewrite/DWARFRewriter.h

@@ -93,14 +93,14 @@ class DWARFRewriter {
   makeFinalLocListsSection(SimpleBinaryPatcher &DebugInfoPatcher);
 
   /// Finalize debug sections in the main binary.
-  void finalizeDebugSections(DebugInfoBinaryPatcher &DebugInfoPatcher);
+  CUOffsetMap finalizeDebugSections(DebugInfoBinaryPatcher &DebugInfoPatcher);
 
   /// Patches the binary for DWARF address ranges (e.g. in functions and lexical
   /// blocks) to be updated.
   void updateDebugAddressRanges();
 
   /// Rewrite .gdb_index section if present.
-  void updateGdbIndexSection();
+  void updateGdbIndexSection(CUOffsetMap &CUMap);
 
   /// Output .dwo files.
   void writeDWOFiles(std::unordered_map<uint64_t, std::string> &DWOIdToName);

bolt/include/bolt/Rewrite/RewriteInstance.h

@@ -545,6 +545,11 @@ class RewriteInstance {
   friend class RewriteInstanceDiff;
 };
 
+MCPlusBuilder *createMCPlusBuilder(const Triple::ArchType Arch,
+                                   const MCInstrAnalysis *Analysis,
+                                   const MCInstrInfo *Info,
+                                   const MCRegisterInfo *RegInfo);
+
 } // namespace bolt
 } // namespace llvm
 

bolt/include/bolt/Utils/CommandLineOpts.h

@@ -44,7 +44,6 @@ extern llvm::cl::opt<unsigned long long> HeatmapMinAddress;
 extern llvm::cl::opt<bool> HotData;
 extern llvm::cl::opt<bool> HotFunctionsAtEnd;
 extern llvm::cl::opt<bool> HotText;
-extern llvm::cl::opt<std::string> InputFilename;
 extern llvm::cl::opt<bool> Instrument;
 extern llvm::cl::opt<std::string> OutputFilename;
 extern llvm::cl::opt<std::string> PerfData;