Merge branch 'sycl' into peter/windows-e2e

Author: ProGTX

.github/ISSUE_TEMPLATE/2-bug-report-cuda.yml

@@ -1,5 +1,7 @@
 name: CUDA-specific bug report
-description: Please use this template in case of CUDA-specific issue
+description: Please use this template in case of CUDA-specific issue.
+Before submitting an issue check that it is not already covered in the
+[troubleshooting guide](https://developer.codeplay.com/products/oneapi/nvidia/latest/guides/troubleshooting)
 labels: ["bug", "cuda"]
 body:
   - type: textarea

.github/ISSUE_TEMPLATE/3-bug-report-hip.yml

@@ -1,5 +1,7 @@
 name: HIP-specific bug report
 description: Please use this template in case of HIP-specific issue
+Before submitting an issue check that it is not already covered in the
+[troubleshooting guide](https://developer.codeplay.com/products/oneapi/amd/latest/guides/troubleshooting)
 labels: ["bug", "hip"]
 body:
   - type: textarea

.github/new-prs-labeler.yml

@@ -846,6 +846,26 @@ backend:PowerPC:
   - clang/lib/Driver/ToolChains/Arch/PPC.*
   - clang/test/CodeGen/PowerPC/**
 
+backend:SystemZ:
+  - llvm/include/llvm/BinaryFormat/ELFRelocs/SystemZ*
+  - llvm/include/llvm/BinaryFormat/GOFF.h
+  - llvm/include/llvm/IR/IntrinsicsSystemZ.td
+  - llvm/lib/Target/SystemZ/**
+  - llvm/test/Analysis/**/SystemZ/**
+  - llvm/test/CodeGen/SystemZ/**
+  - llvm/test/DebugInfo/SystemZ/**
+  - llvm/test/ExecutionEngine/**/SystemZ/**
+  - llvm/test/MC/Disassembler/SystemZ/**
+  - llvm/test/MC/GOFF/**
+  - llvm/test/MC/SystemZ/**
+  - llvm/test/Transforms/**/SystemZ/**
+  - clang/include/clang/Basic/BuiltinsSystemZ.*
+  - clang/lib/Basic/Targets/SystemZ.*
+  - clang/lib/CodeGen/Targets/SystemZ.cpp
+  - clang/lib/Driver/ToolChains/ZOS*
+  - clang/lib/Driver/ToolChains/Arch/SystemZ.*
+  - clang/test/CodeGen/SystemZ/**
+
 third-party:unittests:
   - third-party/unittests/**
 

bolt/include/bolt/Core/BinaryContext.h

@@ -997,6 +997,10 @@ class BinaryContext {
     return getUniqueSectionByName(".gdb_index");
   }
 
+  ErrorOr<BinarySection &> getDebugNamesSection() const {
+    return getUniqueSectionByName(".debug_names");
+  }
+
   /// @}
 
   /// Register \p TargetFunction as a fragment of \p Function if checks pass:

bolt/include/bolt/Core/DIEBuilder.h

@@ -16,6 +16,7 @@
 #define BOLT_CORE_DIE_BUILDER_H
 
 #include "bolt/Core/BinaryContext.h"
+#include "bolt/Core/DebugNames.h"
 #include "llvm/CodeGen/DIE.h"
 #include "llvm/DebugInfo/DWARF/DWARFAbbreviationDeclaration.h"
 #include "llvm/DebugInfo/DWARF/DWARFDie.h"
@@ -127,6 +128,7 @@ class DIEBuilder {
   DWARFUnit *SkeletonCU{nullptr};
   uint64_t UnitSize{0};
   llvm::DenseSet<uint64_t> AllProcessed;
+  DWARF5AcceleratorTable &DebugNamesTable;
 
   /// Returns current state of the DIEBuilder
   State &getState() { return *BuilderState.get(); }
@@ -206,8 +208,8 @@ class DIEBuilder {
   /// Update references once the layout is finalized.
   void updateReferences();
 
-  /// Update the Offset and Size of DIE.
-  uint32_t computeDIEOffset(const DWARFUnit &CU, DIE &Die, uint32_t &CurOffset);
+  /// Update the Offset and Size of DIE, populate DebugNames table.
+  uint32_t finalizeDIEs(DWARFUnit &CU, DIE &Die, uint32_t &CurOffset);
 
   void registerUnit(DWARFUnit &DU, bool NeedSort);
 
@@ -269,6 +271,7 @@ class DIEBuilder {
 
 public:
   DIEBuilder(BinaryContext &BC, DWARFContext *DwarfContext,
+             DWARF5AcceleratorTable &DebugNamesTable,
              DWARFUnit *SkeletonCU = nullptr);
 
   /// Returns enum to what we are currently processing.

bolt/include/bolt/Core/DebugData.h

@@ -439,6 +439,8 @@ class DebugStrOffsetsWriter {
   /// Update Str offset in .debug_str in .debug_str_offsets.
   void updateAddressMap(uint32_t Index, uint32_t Address);
 
+  /// Get offset for given index in original .debug_str_offsets section.
+  uint64_t getOffset(uint32_t Index) const { return StrOffsets[Index]; }
   /// Writes out current sections entry into .debug_str_offsets.
   void finalizeSection(DWARFUnit &Unit, DIEBuilder &DIEBldr);
 
@@ -463,7 +465,7 @@ class DebugStrOffsetsWriter {
   std::unique_ptr<DebugStrOffsetsBufferVector> StrOffsetsBuffer;
   std::unique_ptr<raw_svector_ostream> StrOffsetsStream;
   std::map<uint32_t, uint32_t> IndexToAddressMap;
-  std::vector<uint32_t> StrOffsets;
+  SmallVector<uint32_t, 5> StrOffsets;
   std::unordered_map<uint64_t, uint64_t> ProcessedBaseOffsets;
   bool StrOffsetSectionWasModified = false;
 };
@@ -484,11 +486,12 @@ class DebugStrWriter {
   /// Returns False if no strings were added to .debug_str.
   bool isInitialized() const { return !StrBuffer->empty(); }
 
+  /// Initializes Buffer and Stream.
+  void initialize();
+
 private:
   /// Mutex used for parallel processing of debug info.
   std::mutex WriterMutex;
-  /// Initializes Buffer and Stream.
-  void initialize();
   /// Creates internal data structures.
   void create();
   std::unique_ptr<DebugStrBufferVector> StrBuffer;

bolt/include/bolt/Core/DebugNames.h

@@ -0,0 +1,172 @@
+//===- bolt/Core/DebugNames.h - Debug names support ---*- C++
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file contains declaration of classes required for generation of
+// .debug_names section.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef BOLT_CORE_DEBUG_NAMES_H
+#define BOLT_CORE_DEBUG_NAMES_H
+
+#include "DebugData.h"
+#include "llvm/CodeGen/AccelTable.h"
+
+namespace llvm {
+namespace bolt {
+class BOLTDWARF5AccelTableData : public DWARF5AccelTableData {
+public:
+  BOLTDWARF5AccelTableData(const uint64_t DieOffset,
+                           const std::optional<uint64_t> DefiningParentOffset,
+                           const unsigned DieTag, const unsigned UnitID,
+                           const bool IsTU,
+                           const std::optional<unsigned> SecondUnitID)
+      : DWARF5AccelTableData(DieOffset, DefiningParentOffset, DieTag, UnitID,
+                             IsTU),
+        SecondUnitID(SecondUnitID) {}
+
+  uint64_t getDieOffset() const { return DWARF5AccelTableData::getDieOffset(); }
+  unsigned getDieTag() const { return DWARF5AccelTableData::getDieTag(); }
+  unsigned getUnitID() const { return DWARF5AccelTableData::getUnitID(); }
+  bool isTU() const { return DWARF5AccelTableData::isTU(); }
+  std::optional<unsigned> getSecondUnitID() const { return SecondUnitID; }
+
+private:
+  std::optional<unsigned> SecondUnitID;
+};
+
+class DWARF5AcceleratorTable {
+public:
+  DWARF5AcceleratorTable(const bool CreateDebugNames, BinaryContext &BC,
+                         DebugStrWriter &MainBinaryStrWriter);
+  ~DWARF5AcceleratorTable() {
+    for (DebugNamesAbbrev *Abbrev : AbbreviationsVector)
+      Abbrev->~DebugNamesAbbrev();
+  }
+  /// Add DWARF5 Accelerator table entry.
+  /// Input is DWARFUnit being processed, DIE that belongs to it, and potential
+  /// SkeletonCU if the Unit comes from a DWO section.
+  void addAccelTableEntry(DWARFUnit &Unit, const DIE &Die,
+                          const std::optional<uint64_t> &DWOID);
+  /// Set current unit being processed.
+  void setCurrentUnit(DWARFUnit &Unit, const uint64_t UnitStartOffset);
+  /// Emit Accelerator table.
+  void emitAccelTable();
+  /// Returns true if the table was crated.
+  bool isCreated() const { return NeedToCreate; }
+  /// Returns buffer containing the accelerator table.
+  std::unique_ptr<DebugBufferVector> releaseBuffer() {
+    return std::move(FullTableBuffer);
+  }
+
+private:
+  BinaryContext &BC;
+  bool NeedToCreate = false;
+  BumpPtrAllocator Allocator;
+  DebugStrWriter &MainBinaryStrWriter;
+  StringRef StrSection;
+  uint64_t CurrentUnitOffset = 0;
+  const DWARFUnit *CurrentUnit = nullptr;
+  std::unordered_map<uint32_t, uint32_t> AbbrevTagToIndexMap;
+
+  /// Represents a group of entries with identical name (and hence, hash value).
+  struct HashData {
+    uint64_t StrOffset;
+    uint32_t HashValue;
+    uint32_t EntryOffset;
+    std::vector<BOLTDWARF5AccelTableData *> Values;
+  };
+  using HashList = std::vector<HashData *>;
+  using BucketList = std::vector<HashList>;
+  /// Contains all the offsets of CUs.
+  SmallVector<uint32_t, 1> CUList;
+  /// Contains all the offsets of local TUs.
+  SmallVector<uint32_t, 1> LocalTUList;
+  /// Contains all the type hashes for split dwarf TUs.
+  SmallVector<uint64_t, 1> ForeignTUList;
+  using StringEntries =
+      MapVector<std::string, HashData, llvm::StringMap<unsigned>>;
+  StringEntries Entries;
+  /// FoldingSet that uniques the abbreviations.
+  FoldingSet<DebugNamesAbbrev> AbbreviationsSet;
+  /// Vector containing DebugNames abbreviations for iteration in order.
+  SmallVector<DebugNamesAbbrev *, 5> AbbreviationsVector;
+  /// The bump allocator to use when creating DIEAbbrev objects in the uniqued
+  /// storage container.
+  BumpPtrAllocator Alloc;
+  uint32_t BucketCount = 0;
+  uint32_t UniqueHashCount = 0;
+  uint32_t AbbrevTableSize = 0;
+  uint32_t CUIndexEncodingSize = 4;
+  uint32_t TUIndexEncodingSize = 4;
+  uint32_t AugmentationStringSize = 0;
+  dwarf::Form CUIndexForm = dwarf::DW_FORM_data4;
+  dwarf::Form TUIndexForm = dwarf::DW_FORM_data4;
+
+  BucketList Buckets;
+
+  std::unique_ptr<DebugBufferVector> FullTableBuffer;
+  std::unique_ptr<raw_svector_ostream> FullTableStream;
+  std::unique_ptr<DebugBufferVector> StrBuffer;
+  std::unique_ptr<raw_svector_ostream> StrStream;
+  std::unique_ptr<DebugBufferVector> EntriesBuffer;
+  std::unique_ptr<raw_svector_ostream> Entriestream;
+  std::unique_ptr<DebugBufferVector> AugStringBuffer;
+  std::unique_ptr<raw_svector_ostream> AugStringtream;
+  llvm::DenseMap<llvm::hash_code, uint64_t> StrCacheToOffsetMap;
+  // Contains DWO ID to CUList Index.
+  llvm::DenseMap<uint64_t, uint32_t> CUOffsetsToPatch;
+  /// Adds Unit to either CUList, LocalTUList or ForeignTUList.
+  /// Input Unit being processed, and DWO ID if Unit is being processed comes
+  /// from a DWO section.
+  void addUnit(DWARFUnit &Unit, const std::optional<uint64_t> &DWOID);
+  /// Returns number of buckets in .debug_name table.
+  ArrayRef<HashList> getBuckets() const { return Buckets; }
+  /// Get encoding for a given attribute.
+  std::optional<DWARF5AccelTable::UnitIndexAndEncoding>
+  getIndexForEntry(const BOLTDWARF5AccelTableData &Value) const;
+  /// Get encoding for a given attribute for second index.
+  /// Returns nullopt if there is no second index.
+  std::optional<DWARF5AccelTable::UnitIndexAndEncoding>
+  getSecondIndexForEntry(const BOLTDWARF5AccelTableData &Value) const;
+  /// Uniquify Entries.
+  void finalize();
+  /// Computes bucket count.
+  void computeBucketCount();
+  /// Populate Abbreviations Map.
+  void populateAbbrevsMap();
+  /// Write Entries.
+  void writeEntries();
+  /// Write an Entry.
+  void writeEntry(const BOLTDWARF5AccelTableData &Entry);
+  /// Write augmentation_string for BOLT.
+  void writeAugmentationString();
+  /// Emit out Header for DWARF5 Accelerator table.
+  void emitHeader() const;
+  /// Emit out CU list.
+  void emitCUList() const;
+  /// Emit out TU List. Combination of LocalTUList and ForeignTUList.
+  void emitTUList() const;
+  /// Emit buckets.
+  void emitBuckets() const;
+  /// Emit hashes for hash table.
+  void emitHashes() const;
+  /// Emit string offsets for hash table.
+  void emitStringOffsets() const;
+  /// Emit Entry Offsets for hash table.
+  void emitOffsets() const;
+  /// Emit abbreviation table.
+  void emitAbbrevs();
+  /// Emit entries.
+  void emitData();
+  /// Emit augmentation string.
+  void emitAugmentationString() const;
+};
+} // namespace bolt
+} // namespace llvm
+#endif

bolt/include/bolt/Core/MCPlusBuilder.h

@@ -620,7 +620,17 @@ class MCPlusBuilder {
     return Info->get(Inst.getOpcode()).mayStore();
   }
 
-  virtual bool isAArch64Exclusive(const MCInst &Inst) const {
+  virtual bool isAArch64ExclusiveLoad(const MCInst &Inst) const {
+    llvm_unreachable("not implemented");
+    return false;
+  }
+
+  virtual bool isAArch64ExclusiveStore(const MCInst &Inst) const {
+    llvm_unreachable("not implemented");
+    return false;
+  }
+
+  virtual bool isAArch64ExclusiveClear(const MCInst &Inst) const {
     llvm_unreachable("not implemented");
     return false;
   }
@@ -1173,11 +1183,16 @@ class MCPlusBuilder {
   bool clearOffset(MCInst &Inst) const;
 
   /// Return the label of \p Inst, if available.
-  MCSymbol *getLabel(const MCInst &Inst) const;
+  MCSymbol *getInstLabel(const MCInst &Inst) const;
+
+  /// Set the label of \p Inst or return the existing label for the instruction.
+  /// This label will be emitted right before \p Inst is emitted to MCStreamer.
+  MCSymbol *getOrCreateInstLabel(MCInst &Inst, const Twine &Name,
+                                 MCContext *Ctx) const;
 
   /// Set the label of \p Inst. This label will be emitted right before \p Inst
   /// is emitted to MCStreamer.
-  bool setLabel(MCInst &Inst, MCSymbol *Label) const;
+  void setInstLabel(MCInst &Inst, MCSymbol *Label) const;
 
   /// Get instruction size specified via annotation.
   std::optional<uint32_t> getSize(const MCInst &Inst) const;

bolt/include/bolt/Rewrite/DWARFRewriter.h

@@ -11,6 +11,7 @@
 
 #include "bolt/Core/DIEBuilder.h"
 #include "bolt/Core/DebugData.h"
+#include "bolt/Core/DebugNames.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/CodeGen/DIE.h"
 #include "llvm/DWP/DWP.h"
@@ -140,8 +141,10 @@ class DWARFRewriter {
                             const std::list<DWARFUnit *> &CUs);
 
   /// Finalize debug sections in the main binary.
-  void finalizeDebugSections(DIEBuilder &DIEBlder, DIEStreamer &Streamer,
-                             raw_svector_ostream &ObjOS, CUOffsetMap &CUMap);
+  void finalizeDebugSections(DIEBuilder &DIEBlder,
+                             DWARF5AcceleratorTable &DebugNamesTable,
+                             DIEStreamer &Streamer, raw_svector_ostream &ObjOS,
+                             CUOffsetMap &CUMap);
 
   /// Patches the binary for DWARF address ranges (e.g. in functions and lexical
   /// blocks) to be updated.

bolt/lib/Core/BinaryContext.cpp

@@ -1967,7 +1967,7 @@ void BinaryContext::printInstruction(raw_ostream &OS, const MCInst &Instruction,
     OS << " # Offset: " << *Offset;
   if (std::optional<uint32_t> Size = MIB->getSize(Instruction))
     OS << " # Size: " << *Size;
-  if (MCSymbol *Label = MIB->getLabel(Instruction))
+  if (MCSymbol *Label = MIB->getInstLabel(Instruction))
     OS << " # Label: " << *Label;
 
   MIB->printAnnotations(Instruction, OS);

bolt/lib/Core/BinaryEmitter.cpp

@@ -489,7 +489,7 @@ void BinaryEmitter::emitFunctionBody(BinaryFunction &BF, FunctionFragment &FF,
 
       if (!EmitCodeOnly) {
         // A symbol to be emitted before the instruction to mark its location.
-        MCSymbol *InstrLabel = BC.MIB->getLabel(Instr);
+        MCSymbol *InstrLabel = BC.MIB->getInstLabel(Instr);
 
         if (opts::UpdateDebugSections && BF.getDWARFUnit()) {
           LastLocSeen = emitLineInfo(BF, Instr.getLoc(), LastLocSeen,

bolt/lib/Core/BinaryFunction.cpp

@@ -1424,7 +1424,7 @@ Error BinaryFunction::disassemble() {
     InstrMapType::iterator II = Instructions.find(Offset);
     assert(II != Instructions.end() && "reference to non-existing instruction");
 
-    BC.MIB->setLabel(II->second, Label);
+    BC.MIB->setInstLabel(II->second, Label);
   }
 
   // Reset symbolizer for the disassembler.

bolt/lib/Core/CMakeLists.txt

@@ -20,6 +20,7 @@ add_llvm_library(LLVMBOLTCore
   BinaryFunctionProfile.cpp
   BinarySection.cpp
   DebugData.cpp
+  DebugNames.cpp
   DIEBuilder.cpp
   DynoStats.cpp
   Exceptions.cpp