Merge pull request #287 from sx-aurora-dev/feature/merge-upstream-202…

…31019

Feature/merge upstream 20231019

.ci/generate-buildkite-pipeline-premerge

@@ -21,8 +21,11 @@ set -eu
 set -o pipefail
 
 # Environment variables script works with:
+
+# Fetch origin/main to have an up to date merge base for main...HEAD diff.
+git fetch origin main:main
 # List of files affected by this commit
-: ${MODIFIED_FILES:=$(git diff --name-only HEAD~1)}
+: ${MODIFIED_FILES:=$(git diff --name-only main...HEAD)}
 # Filter rules for generic windows tests
 : ${WINDOWS_AGENTS:='{"queue": "windows"}'}
 # Filter rules for generic linux tests

.github/CODEOWNERS

@@ -30,3 +30,5 @@
 /clang/test/CXX/drs/ @Endilll
 /clang/www/cxx_dr_status.html @Endilll
 /clang/www/make_cxx_dr_status @Endilll
+
+/lldb/ @JDevlieghere

.github/new-prs-labeler.yml

@@ -794,3 +794,35 @@ llvm:binary-utilities:
   - llvm/tools/llvm-size/**
   - llvm/tools/llvm-strings/**
   - llvm/tools/llvm-symbolizer/**
+
+clang:openmp:
+  - clang/include/clang/Basic/OpenMP*
+  - clang/include/clang/AST/OpenMPClause.h
+  - clang/include/clang/AST/DeclOpenMP.h
+  - clang/include/clang/AST/ExprOpenMP.h
+  - clang/include/clang/AST/StmtOpenMP.h
+  - clang/lib/AST/DeclOpenMP.cpp
+  - clang/lib/AST/OpenMPClause.cpp
+  - clang/lib/AST/StmtOpenMP.cpp
+  - clang/lib/Headers/openmp_wrappers/**
+  - clang/lib/Parse/ParseOpenMP.cpp
+  - clang/lib/Basic/OpenMPKinds.cpp
+  - clang/lib/Driver/ToolChains/AMDGPUOpenMP.cpp
+  - clang/lib/Driver/ToolChains/AMDGPUOpenMP.h
+  - clang/lib/CodeGen/CgStmtOpenMP.cpp
+  - clang/lib/CodeGen/CGOpenMP*
+  - clang/lib/Sema/SemaOpenMP.cpp
+  - clang/test/OpenMP/**
+  - clang/test/AST/ast-dump-openmp-*
+  - llvm/lib/Frontend/OpenMP/**
+  - llvm/lib/Transforms/IPO/OpenMPOpt.cpp
+  - llvm/include/llvm/Frontend/OpenMP/**
+  - llvm/include/llvm/Transforms/IPO/OpenMPOpt.h
+  - llvm/unittests/Frontend/OpenMP*
+  - llvm/test/Transforms/OpenMP/**
+
+openmp:libomp:
+  - any: ['openmp/**', '!openmp/libomptarget/**']
+
+openmp:libomptarget:
+  - any: ['openmp/**', '!openmp/runtime/**']

.github/workflows/docs.yml

@@ -0,0 +1,63 @@
+# LLVM Documentation CI
+# 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
+
+name: "Test documentation build"
+
+permissions:
+  contents: read
+
+on:
+  push:
+    branches:
+      - 'main'
+    paths:
+      - 'llvm/docs/**'
+      - 'clang/docs/**'
+  pull_request:
+    paths:
+      - 'llvm/docs/**'
+      - 'clang/docs/**'
+
+jobs:
+  check-docs-build:
+    name: "Test documentation build"
+    runs-on: ubuntu-latest
+    steps:
+      - name: Fetch LLVM sources
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 1
+      - name: Setup Python env
+        uses: actions/setup-python@v4
+        with:
+          python-version: '3.11'
+          cache: 'pip'
+          cache-dependency-path: 'llvm/docs/requirements.txt'
+      - name: Install python dependencies
+        run: pip install -r llvm/docs/requirements.txt
+      - name: Install system dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y cmake ninja-build
+      - name: Get subprojects that have doc changes
+        id: docs-changed-subprojects
+        uses: tj-actions/changed-files@v39
+        with:
+          files_yaml: |
+            llvm:
+              - 'llvm/docs/**'
+            clang:
+              - 'clang/docs/**'
+      - name: Build LLVM docs
+        if: steps.docs-changed-subprojects.outputs.llvm_any_changed == 'true'
+        run: |
+          cmake -B llvm-build -GNinja -DCMAKE_BUILD_TYPE=Release -DLLVM_ENABLE_SPHINX=ON -DSPHINX_OUTPUT_HTML=ON -DSPHINX_OUTPUT_MAN=ON ./llvm
+          TZ=UTC ninja -C llvm-build docs-llvm-html docs-llvm-man
+      - name: Build Clang docs
+        if: steps.docs-changed-subprojects.outputs.clang_any_changed == 'true'
+        run: |
+          cmake -B clang-build -GNinja -DCMAKE_BUILD_TYPE=Release -DLLVM_ENABLE_PROJECTS="clang" -DLLVM_ENABLE_SPHINX=ON -DSPHINX_OUTPUT_HTML=ON -DSPHINX_OUTPUT_MAN=ON ./llvm
+          TZ=UTC ninja -C clang-build docs-clang-html docs-clang-man
+

.github/workflows/libcxx-check-generated-files.yml

@@ -0,0 +1,21 @@
+name: "Check libc++ generated files"
+on:
+  pull_request:
+    paths:
+      - 'libcxx/**'
+
+jobs:
+  check_generated_files:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Fetch LLVM sources
+        uses: actions/checkout@v4
+
+      - name: Install dependencies
+        uses: aminya/setup-cpp@v1
+        with:
+          clangformat: 17.0.1
+          ninja: true
+
+      - name: Check generated files
+        run: libcxx/utils/ci/run-buildbot check-generated-output

.github/workflows/release-tasks.yml

@@ -104,16 +104,18 @@ jobs:
         run: |
           cd llvm/utils/lit
           # Remove 'dev' suffix from lit version.
-          sed -i "s/ + 'dev'//g" lit/__init__.py
+          sed -i 's/ + "dev"//g' lit/__init__.py
           python3 setup.py sdist
 
       - name: Upload lit to test.pypi.org
         uses: pypa/gh-action-pypi-publish@release/v1
         with:
           password: ${{ secrets.LLVM_LIT_TEST_PYPI_API_TOKEN }}
           repository-url: https://test.pypi.org/legacy/
+          packages-dir: llvm/utils/lit/dist/
 
       - name: Upload lit to pypi.org
         uses: pypa/gh-action-pypi-publish@release/v1
         with:
           password: ${{ secrets.LLVM_LIT_PYPI_API_TOKEN }}
+          packages-dir: llvm/utils/lit/dist/

bolt/include/bolt/Core/BinaryFunction.h

@@ -192,9 +192,6 @@ class BinaryFunction {
   static constexpr uint64_t COUNT_NO_PROFILE =
       BinaryBasicBlock::COUNT_NO_PROFILE;
 
-  /// We have to use at least 2-byte alignment for functions because of C++ ABI.
-  static constexpr unsigned MinAlign = 2;
-
   static const char TimerGroupName[];
   static const char TimerGroupDesc[];
 
@@ -1720,8 +1717,24 @@ class BinaryFunction {
     return *this;
   }
 
-  Align getAlign() const { return Align(Alignment); }
+  uint16_t getMinAlignment() const {
+    // Align data in code BFs minimum to CI alignment
+    if (!size() && hasIslandsInfo())
+      return getConstantIslandAlignment();
+
+    // Minimal code alignment on AArch64 and RISCV is 4
+    if (BC.isAArch64() || BC.isRISCV())
+      return 4;
+
+    // We have to use at least 2-byte alignment for functions because
+    // of C++ ABI.
+    return 2;
+  }
+
+  Align getMinAlign() const { return Align(getMinAlignment()); }
+
   uint16_t getAlignment() const { return Alignment; }
+  Align getAlign() const { return Align(getAlignment()); }
 
   BinaryFunction &setMaxAlignmentBytes(uint16_t MaxAlignBytes) {
     MaxAlignmentBytes = MaxAlignBytes;

bolt/include/bolt/Core/MCPlus.h

@@ -66,6 +66,7 @@ class MCAnnotation {
     kTailCall,            /// Tail call.
     kConditionalTailCall, /// CTC.
     kOffset,              /// Offset in the function.
+    kLabel,               /// MCSymbol pointing to this instruction.
     kGeneric              /// First generic annotation.
   };
 

bolt/include/bolt/Core/MCPlusBuilder.h

@@ -160,6 +160,7 @@ class MCPlusBuilder {
   const MCInstrAnalysis *Analysis;
   const MCInstrInfo *Info;
   const MCRegisterInfo *RegInfo;
+  const MCSubtargetInfo *STI;
 
   /// Map annotation name into an annotation index.
   StringMap<uint64_t> AnnotationNameIndexMap;
@@ -331,8 +332,8 @@ class MCPlusBuilder {
 
 public:
   MCPlusBuilder(const MCInstrAnalysis *Analysis, const MCInstrInfo *Info,
-                const MCRegisterInfo *RegInfo)
-      : Analysis(Analysis), Info(Info), RegInfo(RegInfo) {
+                const MCRegisterInfo *RegInfo, const MCSubtargetInfo *STI)
+      : Analysis(Analysis), Info(Info), RegInfo(RegInfo), STI(STI) {
     // Initialize the default annotation allocator with id 0
     AnnotationAllocators.emplace(0, AnnotationAllocator());
     MaxAllocatorId++;
@@ -1179,6 +1180,13 @@ class MCPlusBuilder {
   /// Remove offset annotation.
   bool clearOffset(MCInst &Inst);
 
+  /// Return the label of \p Inst, if available.
+  std::optional<MCSymbol *> getLabel(const MCInst &Inst) 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, AllocatorIdTy AllocatorId = 0);
+
   /// Return MCSymbol that represents a target of this instruction at a given
   /// operand number \p OpNum. If there's no symbol associated with
   /// the operand - return nullptr.
@@ -2079,15 +2087,18 @@ class MCPlusBuilder {
 
 MCPlusBuilder *createX86MCPlusBuilder(const MCInstrAnalysis *,
                                       const MCInstrInfo *,
-                                      const MCRegisterInfo *);
+                                      const MCRegisterInfo *,
+                                      const MCSubtargetInfo *);
 
 MCPlusBuilder *createAArch64MCPlusBuilder(const MCInstrAnalysis *,
                                           const MCInstrInfo *,
-                                          const MCRegisterInfo *);
+                                          const MCRegisterInfo *,
+                                          const MCSubtargetInfo *);
 
 MCPlusBuilder *createRISCVMCPlusBuilder(const MCInstrAnalysis *,
                                         const MCInstrInfo *,
-                                        const MCRegisterInfo *);
+                                        const MCRegisterInfo *,
+                                        const MCSubtargetInfo *);
 
 } // namespace bolt
 } // namespace llvm

bolt/include/bolt/Core/Relocation.h

@@ -98,6 +98,10 @@ struct Relocation {
   /// Return true if relocation type is for thread local storage.
   static bool isTLS(uint64_t Type);
 
+  /// Return true of relocation type is for referencing a specific instruction
+  /// (as opposed to a function, basic block, etc).
+  static bool isInstructionReference(uint64_t Type);
+
   /// Return code for a NONE relocation
   static uint64_t getNone();
 

bolt/include/bolt/Rewrite/RewriteInstance.h

@@ -584,7 +584,8 @@ class RewriteInstance {
 MCPlusBuilder *createMCPlusBuilder(const Triple::ArchType Arch,
                                    const MCInstrAnalysis *Analysis,
                                    const MCInstrInfo *Info,
-                                   const MCRegisterInfo *RegInfo);
+                                   const MCRegisterInfo *RegInfo,
+                                   const MCSubtargetInfo *STI);
 
 } // namespace bolt
 } // namespace llvm

bolt/lib/Core/BinaryContext.cpp

@@ -1803,6 +1803,10 @@ MarkerSymType BinaryContext::getMarkerType(const SymbolRef &Symbol) const {
   if (*NameOrError == "$x" || NameOrError->startswith("$x."))
     return MarkerSymType::CODE;
 
+  // $x<ISA>
+  if (isRISCV() && NameOrError->startswith("$x"))
+    return MarkerSymType::CODE;
+
   if (*NameOrError == "$d" || NameOrError->startswith("$d."))
     return MarkerSymType::DATA;
 
@@ -1888,6 +1892,8 @@ void BinaryContext::printInstruction(raw_ostream &OS, const MCInst &Instruction,
   }
   if (std::optional<uint32_t> Offset = MIB->getOffset(Instruction))
     OS << " # Offset: " << *Offset;
+  if (auto Label = MIB->getLabel(Instruction))
+    OS << " # Label: " << **Label;
 
   MIB->printAnnotations(Instruction, OS);
 

bolt/lib/Core/BinaryEmitter.cpp

@@ -309,7 +309,7 @@ bool BinaryEmitter::emitFunction(BinaryFunction &Function,
     // tentative layout.
     Section->ensureMinAlignment(Align(opts::AlignFunctions));
 
-    Streamer.emitCodeAlignment(Align(BinaryFunction::MinAlign), &*BC.STI);
+    Streamer.emitCodeAlignment(Function.getMinAlign(), &*BC.STI);
     uint16_t MaxAlignBytes = FF.isSplitFragment()
                                  ? Function.getMaxColdAlignmentBytes()
                                  : Function.getMaxAlignmentBytes();
@@ -498,6 +498,9 @@ void BinaryEmitter::emitFunctionBody(BinaryFunction &BF, FunctionFragment &FF,
         BB->getLocSyms().emplace_back(Offset, LocSym);
       }
 
+      if (auto Label = BC.MIB->getLabel(Instr))
+        Streamer.emitLabel(*Label);
+
       Streamer.emitInstruction(Instr, *BC.STI);
       LastIsPrefix = BC.MIB->isPrefix(Instr);
     }

bolt/lib/Core/BinaryFunction.cpp

@@ -164,8 +164,6 @@ bool shouldPrint(const BinaryFunction &Function) {
 namespace llvm {
 namespace bolt {
 
-constexpr unsigned BinaryFunction::MinAlign;
-
 template <typename R> static bool emptyRange(const R &Range) {
   return Range.begin() == Range.end();
 }
@@ -1173,6 +1171,13 @@ bool BinaryFunction::disassemble() {
   // basic block.
   Labels[0] = Ctx->createNamedTempSymbol("BB0");
 
+  // Map offsets in the function to a label that should always point to the
+  // corresponding instruction. This is used for labels that shouldn't point to
+  // the start of a basic block but always to a specific instruction. This is
+  // used, for example, on RISC-V where %pcrel_lo relocations point to the
+  // corresponding %pcrel_hi.
+  LabelsMapType InstructionLabels;
+
   uint64_t Size = 0; // instruction size
   for (uint64_t Offset = 0; Offset < getSize(); Offset += Size) {
     MCInst Instruction;
@@ -1329,9 +1334,23 @@ bool BinaryFunction::disassemble() {
                 ItrE = Relocations.lower_bound(Offset + Size);
            Itr != ItrE; ++Itr) {
         const Relocation &Relocation = Itr->second;
+        MCSymbol *Symbol = Relocation.Symbol;
+
+        if (Relocation::isInstructionReference(Relocation.Type)) {
+          uint64_t RefOffset = Relocation.Value - getAddress();
+          LabelsMapType::iterator LI = InstructionLabels.find(RefOffset);
+
+          if (LI == InstructionLabels.end()) {
+            Symbol = BC.Ctx->createNamedTempSymbol();
+            InstructionLabels.emplace(RefOffset, Symbol);
+          } else {
+            Symbol = LI->second;
+          }
+        }
+
         int64_t Value = Relocation.Value;
         const bool Result = BC.MIB->replaceImmWithSymbolRef(
-            Instruction, Relocation.Symbol, Relocation.Addend, Ctx.get(), Value,
+            Instruction, Symbol, Relocation.Addend, Ctx.get(), Value,
             Relocation.Type);
         (void)Result;
         assert(Result && "cannot replace immediate with relocation");
@@ -1366,6 +1385,13 @@ bool BinaryFunction::disassemble() {
     addInstruction(Offset, std::move(Instruction));
   }
 
+  for (auto [Offset, Label] : InstructionLabels) {
+    InstrMapType::iterator II = Instructions.find(Offset);
+    assert(II != Instructions.end() && "reference to non-existing instruction");
+
+    BC.MIB->setLabel(II->second, Label);
+  }
+
   // Reset symbolizer for the disassembler.
   BC.SymbolicDisAsm->setSymbolizer(nullptr);
 
@@ -4489,7 +4515,7 @@ void BinaryFunction::addRelocation(uint64_t Address, MCSymbol *Symbol,
   uint64_t Offset = Address - getAddress();
   LLVM_DEBUG(dbgs() << "BOLT-DEBUG: addRelocation in "
                     << formatv("{0}@{1:x} against {2}\n", *this, Offset,
-                               Symbol->getName()));
+                               (Symbol ? Symbol->getName() : "<undef>")));
   bool IsCI = BC.isAArch64() && isInConstantIsland(Address);
   std::map<uint64_t, Relocation> &Rels =
       IsCI ? Islands->Relocations : Relocations;