Merge upstream llvm into amd-debug

.github/workflows/containers/github-action-ci-tooling/Dockerfile

@@ -1,6 +1,6 @@
-ARG LLVM_VERSION=22.1.0
+ARG LLVM_VERSION=23.1.0
 # FIXME: Use "${LLVM_VERSION%%.*}" instead of "LLVM_VERSION_MAJOR" once we update runners to Ubuntu-26.04 with Buildah >= 1.37
-ARG LLVM_VERSION_MAJOR=22
+ARG LLVM_VERSION_MAJOR=23
 
 FROM docker.io/library/ubuntu:24.04@sha256:c4a8d5503dfb2a3eb8ab5f807da5bc69a85730fb49b5cfca2330194ebcc41c7b AS llvm-downloader
 ARG LLVM_VERSION

.github/workflows/libcxx-pr-conformance-tests.yaml

@@ -45,7 +45,7 @@ jobs:
         config: [
           'frozen-cxx03-headers',
           'generic-cxx03',
-          'generic-cxx26',
+          'generic-cxx29',
           'generic-modules'
         ]
         cc: [  'clang-23' ]
@@ -91,7 +91,8 @@ jobs:
           'generic-cxx14',
           'generic-cxx17',
           'generic-cxx20',
-          'generic-cxx23'
+          'generic-cxx23',
+          'generic-cxx26',
         ]
         cc: [ 'clang-23' ]
         cxx: [ 'clang++-23' ]

.github/workflows/pr-code-format.yml

@@ -13,7 +13,7 @@ jobs:
   code_formatter:
     runs-on: ubuntu-24.04
     container:
-      image: 'ghcr.io/llvm/ci-ubuntu-24.04-format'
+      image: 'ghcr.io/llvm/ci-ubuntu-24.04-format:latest@sha256:7d17cad6fb4fcf9b23bef3a4c6debe9e516e9a38298d8244bdd486322f2b1cc8'
     timeout-minutes: 30
     concurrency:
       group: ${{ github.workflow }}-${{ github.event.pull_request.number }}

.github/workflows/pr-code-lint.yml

@@ -21,7 +21,7 @@ jobs:
       run:
         shell: bash
     container:
-      image: 'ghcr.io/llvm/ci-ubuntu-24.04-lint:latest@sha256:416874e477b33c1cb66a30e77e0b9942b7333a7bd304c2f1d51743e7adaedd54'
+      image: 'ghcr.io/llvm/ci-ubuntu-24.04-lint:latest@sha256:1b62569840b08f7c966a20effdba36845b9890dc5661a8ed6f271b0f003e398f'
     timeout-minutes: 60
     concurrency:
       group: ${{ github.workflow }}-${{ github.ref }}

.github/workflows/release-binaries.yml

@@ -22,6 +22,8 @@ on:
           - ubuntu-22.04
           - ubuntu-22.04-arm
           - macos-14
+          - windows-2022
+          - windows-11-arm
 
   workflow_call:
     inputs:
@@ -153,13 +155,6 @@ jobs:
           target_cmake_flags="$target_cmake_flags -DBOOTSTRAP_BOOTSTRAP_COMPILER_RT_ENABLE_IOS=OFF"
           if [ "$RUNNER_ARCH" = "ARM64" ]; then
             arches=arm64
-          else
-            arches=x86_64
-            # Disable Flang builds on macOS x86_64.  The FortranLower library takes
-            # 2-3 hours to build on macOS, much slower than on Linux.
-            # The long build time causes the release build to time out on x86_64,
-            # so we need to disable flang there.
-            target_cmake_flags="$target_cmake_flags -DLLVM_RELEASE_ENABLE_PROJECTS='clang;lld;lldb;clang-tools-extra;polly;mlir'"
           fi
           target_cmake_flags="$target_cmake_flags -DBOOTSTRAP_BOOTSTRAP_DARWIN_osx_ARCHS=$arches -DBOOTSTRAP_BOOTSTRAP_DARWIN_osx_BUILTIN_ARCHS=$arches"
         fi
@@ -292,7 +287,7 @@ jobs:
       run: |
         choco install wixtoolset --version 3.14.1.20250415
         # Add wix install directory to PATH.
-        Split-Path (Get-ChildItem -Path "C:\" -Filter candle.exe -Recurse -ErrorAction SilentlyContinue | Select-Object -First 1).FullName
+        Split-Path (Get-ChildItem -Path "C:\" -Filter candle.exe -Recurse -ErrorAction SilentlyContinue | Select-Object -First 1).FullName >> $env:GITHUB_PATH
 
     - name: Build Windows
       id: build-windows
@@ -316,7 +311,7 @@ jobs:
         echo "windows-installer-filename=$(Split-Path -Path $installer -Leaf)" >> $env:GITHUB_OUTPUT
 
     - name: Dump Wix logs
-      if: runner.os == 'Windows' && always()
+      if: runner.os == 'Windows' && failure()
       env:
         LLVM_VERSION: ${{ needs.prepare.outputs.release-version }}
         BUILD_DIR_SUFFIX: ${{ case(runner.arch == 'ARM64', 'arm64', 'amd64') }}

CODE_OF_CONDUCT.md

@@ -1,3 +1,5 @@
 # Code of Conduct
 
 The LLVM Community Code of Conduct can be found at https://llvm.org/docs/CodeOfConduct.html.
+For inquiries about moderation, bans, or related issues, please
+contact the LLVM Code of Conduct Committee at conduct@llvm.org

bolt/include/bolt/Utils/CommandLineOpts.h

@@ -119,6 +119,7 @@ extern llvm::cl::opt<SplitFunctionsStrategy> SplitStrategy;
 enum ProfileFormatKind { PF_Fdata, PF_YAML, PF_PreAgg, PF_PerfScript };
 
 extern llvm::cl::opt<ProfileFormatKind> ProfileFormat;
+extern llvm::cl::list<std::string> ReorderData;
 extern llvm::cl::opt<bool> ShowDensity;
 extern llvm::cl::opt<bool> SplitEH;
 extern llvm::cl::opt<bool> StrictMode;

bolt/lib/Core/BinaryContext.cpp

@@ -1540,21 +1540,23 @@ void BinaryContext::processInterproceduralReferences() {
 }
 
 void BinaryContext::postProcessSymbolTable() {
-  fixBinaryDataHoles();
-  bool Valid = true;
-  for (auto &Entry : BinaryDataMap) {
-    BinaryData *BD = Entry.second;
-    if ((BD->getName().starts_with("SYMBOLat") ||
-         BD->getName().starts_with("DATAat")) &&
-        !BD->getParent() && !BD->getSize() && !BD->isAbsolute() &&
-        BD->getSection()) {
-      this->errs() << "BOLT-WARNING: zero-sized top level symbol: " << *BD
-                   << "\n";
-      Valid = false;
-    }
-  }
-  assert(Valid);
-  (void)Valid;
+  if (!opts::ReorderData.empty()) {
+    fixBinaryDataHoles();
+    bool Valid = true;
+    for (auto &Entry : BinaryDataMap) {
+      BinaryData *BD = Entry.second;
+      if ((BD->getName().starts_with("SYMBOLat") ||
+           BD->getName().starts_with("DATAat")) &&
+          !BD->getParent() && !BD->getSize() && !BD->isAbsolute() &&
+          BD->getSection()) {
+        this->errs() << "BOLT-WARNING: zero-sized top level symbol: " << *BD
+                     << "\n";
+        Valid = false;
+      }
+    }
+    assert(Valid);
+    (void)Valid;
+  }
   generateSymbolHashes();
 }
 

bolt/lib/Core/BinaryFunction.cpp

@@ -1330,13 +1330,6 @@ Error 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;
@@ -1423,7 +1416,8 @@ Error BinaryFunction::disassemble() {
 
     if (MIB->isBranch(Instruction) || MIB->isCall(Instruction)) {
       uint64_t TargetAddress = 0;
-      if (MIB->evaluateBranch(Instruction, AbsoluteInstrAddr, Size,
+      if (!MIB->isIndirectBranch(Instruction) &&
+          MIB->evaluateBranch(Instruction, AbsoluteInstrAddr, Size,
                               TargetAddress)) {
         // Check if the target is within the same function. Otherwise it's
         // a call, possibly a tail call.
@@ -1501,42 +1495,6 @@ Error BinaryFunction::disassemble() {
         if (BC.isAArch64())
           handleAArch64IndirectCall(Instruction, Offset);
       }
-    } else if (BC.isRISCV()) {
-      // Check if there's a relocation associated with this instruction.
-      for (auto Itr = Relocations.lower_bound(Offset),
-                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;
-          }
-        }
-
-        uint64_t Addend = Relocation.Addend;
-
-        // For GOT relocations, create a reference against GOT entry ignoring
-        // the relocation symbol.
-        if (Relocation::isGOT(Relocation.Type)) {
-          assert(Relocation::isPCRelative(Relocation.Type) &&
-                 "GOT relocation must be PC-relative on RISC-V");
-          Symbol = BC.registerNameAtAddress("__BOLT_got_zero", 0, 0, 0);
-          Addend = Relocation.Value + Relocation.Offset + getAddress();
-        }
-        int64_t Value = Relocation.Value;
-        const bool Result = BC.MIB->replaceImmWithSymbolRef(
-            Instruction, Symbol, Addend, Ctx.get(), Value, Relocation.Type);
-        (void)Result;
-        assert(Result && "cannot replace immediate with relocation");
-      }
     }
 
 add_instruction:
@@ -1578,13 +1536,6 @@ Error BinaryFunction::disassemble() {
   // Scope-boundary markers are only consulted while assigning offsets above.
   DebugScopeBoundaryOffsets.clear();
 
-  for (auto [Offset, Label] : InstructionLabels) {
-    InstrMapType::iterator II = Instructions.find(Offset);
-    assert(II != Instructions.end() && "reference to non-existing instruction");
-
-    BC.MIB->setInstLabel(II->second, Label);
-  }
-
   // Reset symbolizer for the disassembler.
   BC.SymbolicDisAsm->setSymbolizer(nullptr);
 

bolt/lib/Core/Relocation.cpp

@@ -892,7 +892,7 @@ bool Relocation::isTLS(uint32_t Type) {
 }
 
 bool Relocation::isInstructionReference(uint32_t Type) {
-  if (Arch != Triple::riscv64)
+  if (Arch != Triple::riscv64 && Arch != Triple::riscv32)
     return false;
 
   switch (Type) {

bolt/lib/Passes/ReorderData.cpp

@@ -16,6 +16,7 @@
 // - estimate temporal locality by looking at CFG?
 
 #include "bolt/Passes/ReorderData.h"
+#include "bolt/Utils/CommandLineOpts.h"
 #include "llvm/ADT/MapVector.h"
 #include <algorithm>
 
@@ -35,13 +36,6 @@ static cl::opt<bool>
                        cl::desc("print section contents after reordering"),
                        cl::Hidden, cl::cat(BoltCategory));
 
-cl::list<std::string>
-ReorderData("reorder-data",
-  cl::CommaSeparated,
-  cl::desc("list of sections to reorder"),
-  cl::value_desc("section1,section2,section3,..."),
-  cl::cat(BoltOptCategory));
-
 enum ReorderAlgo : char {
   REORDER_COUNT         = 0,
   REORDER_FUNCS         = 1

bolt/lib/Target/RISCV/CMakeLists.txt

@@ -1,5 +1,6 @@
 set(LLVM_LINK_COMPONENTS
   MC
+  MCDisassembler
   Support
   RISCVDesc
   )
@@ -19,6 +20,7 @@ endif()
 
 add_llvm_library(LLVMBOLTTargetRISCV
   RISCVMCPlusBuilder.cpp
+  RISCVMCSymbolizer.cpp
 
   NO_EXPORT
   DISABLE_LLVM_LINK_LLVM_DYLIB

bolt/lib/Target/RISCV/RISCVMCPlusBuilder.cpp

@@ -12,6 +12,7 @@
 
 #include "MCTargetDesc/RISCVMCAsmInfo.h"
 #include "MCTargetDesc/RISCVMCTargetDesc.h"
+#include "RISCVMCSymbolizer.h"
 #include "bolt/Core/MCPlusBuilder.h"
 #include "llvm/BinaryFormat/ELF.h"
 #include "llvm/MC/MCContext.h"
@@ -39,6 +40,12 @@ class RISCVMCPlusBuilder : public MCPlusBuilder {
 public:
   using MCPlusBuilder::MCPlusBuilder;
 
+  std::unique_ptr<MCSymbolizer>
+  createTargetSymbolizer(BinaryFunction &Function,
+                         bool CreateNewSymbols) const override {
+    return std::make_unique<RISCVMCSymbolizer>(Function, CreateNewSymbols);
+  }
+
   bool equals(const MCSpecifierExpr &A, const MCSpecifierExpr &B,
               CompFuncTy Comp) const override {
     const auto &RISCVExprA = cast<MCSpecifierExpr>(A);