Merge remote-tracking branch 'origin/main' into fix-whigher-precision-for-complex-division

Author: Maetveis

.ci/monolithic-linux.sh

@@ -33,6 +33,7 @@ function at-exit {
 
   mkdir -p artifacts
   ccache --print-stats > artifacts/ccache_stats.txt
+  cp "${BUILD_DIR}"/.ninja_log artifacts/.ninja_log
 
   # If building fails there will be no results files.
   shopt -s nullglob

.ci/monolithic-windows.sh

@@ -32,6 +32,7 @@ function at-exit {
 
   mkdir -p artifacts
   sccache --show-stats >> artifacts/sccache_stats.txt
+  cp "${BUILD_DIR}"/.ninja_log artifacts/.ninja_log
 
   # If building fails there will be no results files.
   shopt -s nullglob

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

@@ -32,7 +32,7 @@ RUN cmake -B ./build -G Ninja ./llvm \
   -DLLVM_ENABLE_RUNTIMES="compiler-rt" \
   -DCMAKE_INSTALL_PREFIX="$LLVM_SYSROOT" \
   -DLLVM_ENABLE_PROJECTS="bolt;clang;lld;clang-tools-extra" \
-  -DLLVM_DISTRIBUTION_COMPONENTS="lld;compiler-rt;clang-format;scan-build" \
+  -DLLVM_DISTRIBUTION_COMPONENTS="lld;compiler-rt;clang-format;scan-build;llvm-symbolizer" \
   -DCLANG_DEFAULT_LINKER="lld"
 
 RUN ninja -C ./build stage2-clang-bolt stage2-install-distribution && ninja -C ./build install-distribution

.github/workflows/libcxx-build-and-test.yaml

@@ -216,7 +216,8 @@ jobs:
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
       - uses: maxim-lobanov/setup-xcode@60606e260d2fc5762a71e64e74b2174e8ea3c8bd # v1.6.0
         with:
-          xcode-version: 'latest'
+          # https://github.com/actions/runner-images/blob/main/images/macos/macos-15-Readme.md
+          xcode-version: '16.3'
       - uses: seanmiddleditch/gha-setup-ninja@3b1f8f94a2f8254bd26914c4ab9474d4f0015f67 # v6
       - name: Build and test
         run: |

.github/workflows/premerge.yaml

@@ -65,6 +65,14 @@ jobs:
           export CXX=/opt/llvm/bin/clang++
 
           ./.ci/monolithic-linux.sh "${projects_to_build}" "${project_check_targets}" "${runtimes_to_build}" "${runtimes_check_targets}"
+            - name: "Upload artifact"
+      - name: Upload Artifacts
+        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08 # v4.6.0
+        with:
+          name: Premerge Artifacts
+          path: artifacts/
+          retention-days: 5
+          include-hidden-files: 'true'
 
   premerge-checks-windows:
     name: Windows Premerge Checks (Test Only - Please Ignore Results)
@@ -113,6 +121,13 @@ jobs:
           set MAX_PARALLEL_LINK_JOBS=64
           call C:\\BuildTools\\Common7\\Tools\\VsDevCmd.bat -arch=amd64 -host_arch=amd64
           bash .ci/monolithic-windows.sh "${{ steps.vars.outputs.windows-projects }}" "${{ steps.vars.outputs.windows-check-targets }}"
+      - name: Upload Artifacts
+        uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08 # v4.6.0
+        with:
+          name: Premerge Artifacts
+          path: artifacts/
+          retention-days: 5
+          include-hidden-files: 'true'
 
   premerge-check-macos:
     name: MacOS Premerge Checks

.mailmap

@@ -29,7 +29,8 @@
 <[email protected]> <[email protected]>
 <[email protected]> <[email protected]>
 <[email protected]> <[email protected]>
-<[email protected]> <[email protected]>
+Min Hsu <[email protected]> <[email protected]>
+Min Hsu <[email protected]> <[email protected]>
 <[email protected]> <[email protected]> <[email protected]>
 <[email protected]> <[email protected]>
 <[email protected]> <[email protected]>
@@ -48,6 +49,6 @@ Ramkumar Ramachandra <[email protected]> <[email protected]>
 Ramkumar Ramachandra <[email protected]> <[email protected]>
 Saleem Abdulrasool <[email protected]>
 Tommy Chiang <[email protected]> <[email protected]>
-Wang Pengcheng <[email protected]>
-Wang Pengcheng <[email protected]> <[email protected]>
-Wang Pengcheng <[email protected]> <[email protected]>
+Pengcheng Wang <[email protected]>
+Pengcheng Wang <[email protected]> <[email protected]>
+Pengcheng Wang <[email protected]> <[email protected]>

bolt/lib/Core/BinaryContext.cpp

@@ -663,6 +663,7 @@ bool BinaryContext::analyzeJumpTable(const uint64_t Address,
     const BinaryFunction *TargetBF = getBinaryFunctionContainingAddress(Value);
     if (!TargetBF || !areRelatedFragments(TargetBF, &BF)) {
       LLVM_DEBUG(printEntryDiagnostics(dbgs(), TargetBF));
+      (void)printEntryDiagnostics;
       break;
     }
 
@@ -843,6 +844,7 @@ BinaryContext::getOrCreateJumpTable(BinaryFunction &Function, uint64_t Address,
                                &Function, std::placeholders::_1);
     assert(llvm::all_of(JT->Parents, isSibling) &&
            "cannot re-use jump table of a different function");
+    (void)isSibling;
     if (opts::Verbosity > 2) {
       this->outs() << "BOLT-INFO: multiple fragments access the same jump table"
                    << ": " << *JT->Parents[0] << "; " << Function << '\n';

bolt/lib/Core/Relocation.cpp

@@ -96,6 +96,7 @@ static bool isSupportedAArch64(uint32_t Type) {
   case ELF::R_AARCH64_MOVW_UABS_G2:
   case ELF::R_AARCH64_MOVW_UABS_G2_NC:
   case ELF::R_AARCH64_MOVW_UABS_G3:
+  case ELF::R_AARCH64_PLT32:
     return true;
   }
 }
@@ -202,6 +203,7 @@ static size_t getSizeForTypeAArch64(uint32_t Type) {
   case ELF::R_AARCH64_MOVW_UABS_G2_NC:
   case ELF::R_AARCH64_MOVW_UABS_G3:
   case ELF::R_AARCH64_ABS32:
+  case ELF::R_AARCH64_PLT32:
     return 4;
   case ELF::R_AARCH64_ABS64:
   case ELF::R_AARCH64_PREL64:
@@ -354,6 +356,7 @@ static uint64_t extractValueAArch64(uint32_t Type, uint64_t Contents,
   case ELF::R_AARCH64_PREL16:
     return static_cast<int64_t>(PC) + SignExtend64<16>(Contents & 0xffff);
   case ELF::R_AARCH64_PREL32:
+  case ELF::R_AARCH64_PLT32:
     return static_cast<int64_t>(PC) + SignExtend64<32>(Contents & 0xffffffff);
   case ELF::R_AARCH64_PREL64:
     return static_cast<int64_t>(PC) + Contents;
@@ -676,6 +679,7 @@ static bool isPCRelativeAArch64(uint32_t Type) {
   case ELF::R_AARCH64_PREL16:
   case ELF::R_AARCH64_PREL32:
   case ELF::R_AARCH64_PREL64:
+  case ELF::R_AARCH64_PLT32:
     return true;
   }
 }

bolt/lib/Profile/DataAggregator.cpp

@@ -871,18 +871,27 @@ DataAggregator::getFallthroughsInTrace(BinaryFunction &BF,
 
   BinaryContext &BC = BF.getBinaryContext();
 
-  if (!BF.isSimple())
-    return std::nullopt;
-
-  assert(BF.hasCFG() && "can only record traces in CFG state");
-
   // Offsets of the trace within this function.
   const uint64_t From = FirstLBR.To - BF.getAddress();
   const uint64_t To = SecondLBR.From - BF.getAddress();
 
   if (From > To)
     return std::nullopt;
 
+  // Accept fall-throughs inside pseudo functions (PLT/thunks).
+  // This check has to be above BF.empty as pseudo functions would pass it:
+  // pseudo => ignored => CFG not built => empty.
+  // If we return nullopt, trace would be reported as mismatching disassembled
+  // function contents which it is not. To avoid this, return an empty
+  // fall-through list instead.
+  if (BF.isPseudo())
+    return Branches;
+
+  if (!BF.isSimple())
+    return std::nullopt;
+
+  assert(BF.hasCFG() && "can only record traces in CFG state");
+
   const BinaryBasicBlock *FromBB = BF.getBasicBlockContainingOffset(From);
   const BinaryBasicBlock *ToBB = BF.getBasicBlockContainingOffset(To);
 

bolt/lib/Rewrite/RewriteInstance.cpp

@@ -2603,7 +2603,9 @@ void RewriteInstance::readRelocations(const SectionRef &Section) {
 void RewriteInstance::handleRelocation(const SectionRef &RelocatedSection,
                                        const RelocationRef &Rel) {
   const bool IsAArch64 = BC->isAArch64();
+  const bool IsX86 = BC->isX86();
   const bool IsFromCode = RelocatedSection.isText();
+  const bool IsWritable = BinarySection(*BC, RelocatedSection).isWritable();
 
   SmallString<16> TypeName;
   Rel.getTypeName(TypeName);
@@ -2612,15 +2614,15 @@ void RewriteInstance::handleRelocation(const SectionRef &RelocatedSection,
     return;
 
   // Adjust the relocation type as the linker might have skewed it.
-  if (BC->isX86() && (RType & ELF::R_X86_64_converted_reloc_bit)) {
+  if (IsX86 && (RType & ELF::R_X86_64_converted_reloc_bit)) {
     if (opts::Verbosity >= 1)
       dbgs() << "BOLT-WARNING: ignoring R_X86_64_converted_reloc_bit\n";
     RType &= ~ELF::R_X86_64_converted_reloc_bit;
   }
 
   if (Relocation::isTLS(RType)) {
     // No special handling required for TLS relocations on X86.
-    if (BC->isX86())
+    if (IsX86)
       return;
 
     // The non-got related TLS relocations on AArch64 and RISC-V also could be
@@ -2661,6 +2663,30 @@ void RewriteInstance::handleRelocation(const SectionRef &RelocatedSection,
     return;
   }
 
+  if (!IsFromCode && !IsWritable && (IsX86 || IsAArch64) &&
+      Relocation::isPCRelative(RType)) {
+    BinaryData *BD = BC->getBinaryDataContainingAddress(Rel.getOffset());
+    if (BD && (BD->nameStartsWith("_ZTV") ||   // vtable
+               BD->nameStartsWith("_ZTCN"))) { // construction vtable
+      BinaryFunction *BF = BC->getBinaryFunctionContainingAddress(
+          SymbolAddress, /*CheckPastEnd*/ false, /*UseMaxSize*/ true);
+      if (!BF || BF->getAddress() != SymbolAddress) {
+        BC->errs()
+            << "BOLT-ERROR: the virtual function table entry at offset 0x"
+            << Twine::utohexstr(Rel.getOffset());
+        if (BF)
+          BC->errs() << " points to the middle of a function @ 0x"
+                     << Twine::utohexstr(BF->getAddress()) << "\n";
+        else
+          BC->errs() << " does not point to any function\n";
+        exit(1);
+      }
+      BC->addRelocation(Rel.getOffset(), BF->getSymbol(), RType, Addend,
+                        ExtractedValue);
+      return;
+    }
+  }
+
   const uint64_t Address = SymbolAddress + Addend;
 
   LLVM_DEBUG({
@@ -2724,7 +2750,7 @@ void RewriteInstance::handleRelocation(const SectionRef &RelocatedSection,
   const bool IsToCode = ReferencedSection && ReferencedSection->isText();
 
   // Special handling of PC-relative relocations.
-  if (BC->isX86() && Relocation::isPCRelative(RType)) {
+  if (IsX86 && Relocation::isPCRelative(RType)) {
     if (!IsFromCode && IsToCode) {
       // PC-relative relocations from data to code are tricky since the
       // original information is typically lost after linking, even with