Manually merged main:e2f9c1853349a5dc86ccd0e85e09af245c418aa8 into amd-gfx:857e6d3ada20

Local branch amd-gfx 857e6d3 Merged main:bb7143f6669345825c214b26fbe336857f4bf523 into amd-gfx:a7f5b4f85e9d
Remote branch main e2f9c18 [builtins] Rename sysauxv to getauxval to reflect the function called. NFCI (llvm#102796) -

Change-Id: I404fa805aec170126d48ce5c39aa2e054feadf6c

Author: perlfu

.github/new-prs-labeler.yml

@@ -990,6 +990,13 @@ clang:openmp:
   - llvm/unittests/Frontend/OpenMP*
   - llvm/test/Transforms/OpenMP/**
 
+clang:as-a-library:
+  - clang/tools/libclang/**
+  - clang/bindings/**
+  - clang/include/clang-c/**
+  - clang/test/LibClang/**
+  - clang/unittest/libclang/**
+
 openmp:libomp:
   - any: ['openmp/**', '!openmp/libomptarget/**']
 

.github/workflows/containers/github-action-ci/stage1.Dockerfile

@@ -2,7 +2,7 @@ FROM docker.io/library/ubuntu:22.04 as base
 ENV LLVM_SYSROOT=/opt/llvm
 
 FROM base as stage1-toolchain
-ENV LLVM_VERSION=17.0.6
+ENV LLVM_VERSION=18.1.8
 
 RUN apt-get update && \
     apt-get install -y \
@@ -21,11 +21,9 @@ WORKDIR /llvm-project-llvmorg-$LLVM_VERSION
 
 COPY bootstrap.patch /
 
-# TODO(boomanaiden154): Remove the patch pulled from a LLVM PR once we bump
-# the toolchain to version 18 and the patch is in-tree.
 # TODO(boomanaiden154): Remove the bootstrap patch once we unsplit the build
 # and no longer need to explicitly build the stage2 dependencies.
-RUN curl https://github.com/llvm/llvm-project/commit/dd0356d741aefa25ece973d6cc4b55dcb73b84b4.patch | patch -p1 && cat /bootstrap.patch | patch -p1
+RUN cat /bootstrap.patch | patch -p1
 
 RUN mkdir build
 

bolt/docs/CommandLineArgumentReference.md

@@ -113,6 +113,10 @@
 
   Prints out offsets for abbrev and debug_info of Skeleton CUs that get patched.
 
+- `--debug-thread-count=<uint>`
+
+  Specifies the number of threads to be used when processing DWO debug information.
+
 - `--dot-tooltip-code`
 
   Add basic block instructions as tool tips on nodes

bolt/docs/OptimizingLinux.md

@@ -37,7 +37,7 @@ Convert `perf` profile into a format suitable for BOLT passing the `vmlinux` bin
 
 
 ```bash
-$ sudo chwon $USER perf.data
+$ sudo chown $USER perf.data
 $ perf2bolt -p perf.data -o perf.fdata vmlinux
 ```
 

bolt/include/bolt/Core/ParallelUtilities.h

@@ -50,7 +50,8 @@ enum SchedulingPolicy {
 };
 
 /// Return the managed thread pool and initialize it if not initialized.
-ThreadPoolInterface &getThreadPool();
+ThreadPoolInterface &
+getThreadPool(const unsigned ThreadsCount = opts::ThreadCount);
 
 /// Perform the work on each BinaryFunction except those that are accepted
 /// by SkipPredicate, scheduling heuristic is based on SchedPolicy.

bolt/include/bolt/Rewrite/DWARFRewriter.h

@@ -184,7 +184,8 @@ class DWARFRewriter {
   /// Output .dwo files.
   void writeDWOFiles(DWARFUnit &, const OverriddenSectionsMap &,
                      const std::string &, DebugLocWriter &,
-                     DebugStrOffsetsWriter &, DebugStrWriter &);
+                     DebugStrOffsetsWriter &, DebugStrWriter &,
+                     DebugRangesSectionWriter &);
   using KnownSectionsEntry = std::pair<MCSection *, DWARFSectionKind>;
 };
 

bolt/lib/Core/ParallelUtilities.cpp

@@ -49,7 +49,7 @@ namespace ParallelUtilities {
 
 namespace {
 /// A single thread pool that is used to run parallel tasks
-std::unique_ptr<DefaultThreadPool> ThreadPoolPtr;
+std::unique_ptr<ThreadPoolInterface> ThreadPoolPtr;
 
 unsigned computeCostFor(const BinaryFunction &BF,
                         const PredicateTy &SkipPredicate,
@@ -102,12 +102,15 @@ inline unsigned estimateTotalCost(const BinaryContext &BC,
 
 } // namespace
 
-ThreadPoolInterface &getThreadPool() {
+ThreadPoolInterface &getThreadPool(const unsigned ThreadsCount) {
   if (ThreadPoolPtr.get())
     return *ThreadPoolPtr;
 
-  ThreadPoolPtr = std::make_unique<DefaultThreadPool>(
-      llvm::hardware_concurrency(opts::ThreadCount));
+  if (ThreadsCount > 1)
+    ThreadPoolPtr = std::make_unique<DefaultThreadPool>(
+        llvm::hardware_concurrency(ThreadsCount));
+  else
+    ThreadPoolPtr = std::make_unique<SingleThreadExecutor>();
   return *ThreadPoolPtr;
 }
 

bolt/lib/Rewrite/DWARFRewriter.cpp

@@ -329,6 +329,12 @@ static cl::opt<bool> KeepARanges(
         "keep or generate .debug_aranges section if .gdb_index is written"),
     cl::Hidden, cl::cat(BoltCategory));
 
+static cl::opt<unsigned>
+    DebugThreadCount("debug-thread-count",
+                     cl::desc("specifies thread count for the multithreading "
+                              "for updating DWO debug info"),
+                     cl::init(1), cl::cat(BoltCategory));
+
 static cl::opt<std::string> DwarfOutputPath(
     "dwarf-output-path",
     cl::desc("Path to where .dwo files will be written out to."), cl::init(""),
@@ -475,8 +481,8 @@ static void emitDWOBuilder(const std::string &DWOName,
                            DWARFUnit &SplitCU, DWARFUnit &CU,
                            DebugLocWriter &LocWriter,
                            DebugStrOffsetsWriter &StrOffstsWriter,
-                           DebugStrWriter &StrWriter,
-                           GDBIndex &GDBIndexSection) {
+                           DebugStrWriter &StrWriter, GDBIndex &GDBIndexSection,
+                           DebugRangesSectionWriter &TempRangesSectionWriter) {
   // Populate debug_info and debug_abbrev for current dwo into StringRef.
   DWODIEBuilder.generateAbbrevs();
   DWODIEBuilder.finish();
@@ -532,7 +538,7 @@ static void emitDWOBuilder(const std::string &DWOName,
     OverriddenSections[Kind] = Contents;
   }
   Rewriter.writeDWOFiles(CU, OverriddenSections, DWOName, LocWriter,
-                         StrOffstsWriter, StrWriter);
+                         StrOffstsWriter, StrWriter, TempRangesSectionWriter);
 }
 
 using DWARFUnitVec = std::vector<DWARFUnit *>;
@@ -646,7 +652,6 @@ void DWARFRewriter::updateDebugInfo() {
                                          *StrWriter);
   GDBIndex GDBIndexSection(BC);
   auto processSplitCU = [&](DWARFUnit &Unit, DWARFUnit &SplitCU,
-                            DIEBuilder &DIEBlder,
                             DebugRangesSectionWriter &TempRangesSectionWriter,
                             DebugAddrWriter &AddressWriter,
                             const std::string &DWOName,
@@ -669,7 +674,7 @@ void DWARFRewriter::updateDebugInfo() {
 
     emitDWOBuilder(DWOName, DWODIEBuilder, *this, SplitCU, Unit,
                    DebugLocDWoWriter, DWOStrOffstsWriter, DWOStrWriter,
-                   GDBIndexSection);
+                   GDBIndexSection, TempRangesSectionWriter);
   };
   auto processMainBinaryCU = [&](DWARFUnit &Unit, DIEBuilder &DIEBlder) {
     std::optional<DWARFUnit *> SplitCU;
@@ -716,9 +721,13 @@ void DWARFRewriter::updateDebugInfo() {
       finalizeTypeSections(DIEBlder, *Streamer, GDBIndexSection);
 
   CUPartitionVector PartVec = partitionCUs(*BC.DwCtx);
+  const unsigned int ThreadCount =
+      std::min(opts::DebugThreadCount, opts::ThreadCount);
   for (std::vector<DWARFUnit *> &Vec : PartVec) {
     DIEBlder.buildCompileUnits(Vec);
     llvm::SmallVector<std::unique_ptr<DIEBuilder>, 72> DWODIEBuildersByCU;
+    ThreadPoolInterface &ThreadPool =
+        ParallelUtilities::getThreadPool(ThreadCount);
     for (DWARFUnit *CU : DIEBlder.getProcessedCUs()) {
       createRangeLocListAddressWriters(*CU);
       std::optional<DWARFUnit *> SplitCU;
@@ -729,9 +738,9 @@ void DWARFRewriter::updateDebugInfo() {
         continue;
       DebugAddrWriter &AddressWriter =
           *AddressWritersByCU[CU->getOffset()].get();
-      DebugRangesSectionWriter *TempRangesSectionWriter =
-          CU->getVersion() >= 5 ? RangeListsWritersByCU[*DWOId].get()
-                                : LegacyRangesWritersByCU[*DWOId].get();
+      DebugRangesSectionWriter &TempRangesSectionWriter =
+          CU->getVersion() >= 5 ? *RangeListsWritersByCU[*DWOId].get()
+                                : *LegacyRangesWritersByCU[*DWOId].get();
       std::optional<std::string> DwarfOutputPath =
           opts::DwarfOutputPath.empty()
               ? std::nullopt
@@ -744,9 +753,17 @@ void DWARFRewriter::updateDebugInfo() {
           *DWODIEBuildersByCU.emplace_back(std::move(DWODIEBuilderPtr)).get();
       if (CU->getVersion() >= 5)
         StrOffstsWriter->finalizeSection(*CU, DIEBlder);
-      processSplitCU(*CU, **SplitCU, DIEBlder, *TempRangesSectionWriter,
-                     AddressWriter, DWOName, DwarfOutputPath, DWODIEBuilder);
+      // Important to capture CU and SplitCU by value here, otherwise when the
+      // thread is executed at some point after the current iteration of the
+      // loop, dereferencing CU/SplitCU in the call to processSplitCU means it
+      // will dereference a different variable than the one intended, causing a
+      // seg fault.
+      ThreadPool.async([&, DwarfOutputPath, DWOName, CU, SplitCU] {
+        processSplitCU(*CU, **SplitCU, TempRangesSectionWriter, AddressWriter,
+                       DWOName, DwarfOutputPath, DWODIEBuilder);
+      });
     }
+    ThreadPool.wait();
     for (std::unique_ptr<DIEBuilder> &DWODIEBuilderPtr : DWODIEBuildersByCU)
       DWODIEBuilderPtr->updateDebugNamesTable();
     for (DWARFUnit *CU : DIEBlder.getProcessedCUs())
@@ -1807,7 +1824,8 @@ std::optional<StringRef> updateDebugData(
 void DWARFRewriter::writeDWOFiles(
     DWARFUnit &CU, const OverriddenSectionsMap &OverridenSections,
     const std::string &DWOName, DebugLocWriter &LocWriter,
-    DebugStrOffsetsWriter &StrOffstsWriter, DebugStrWriter &StrWriter) {
+    DebugStrOffsetsWriter &StrOffstsWriter, DebugStrWriter &StrWriter,
+    DebugRangesSectionWriter &TempRangesSectionWriter) {
   // Setup DWP code once.
   DWARFContext *DWOCtx = BC.getDWOContext();
   const uint64_t DWOId = *CU.getDWOId();
@@ -1854,9 +1872,8 @@ void DWARFRewriter::writeDWOFiles(
 
   DebugRangeListsSectionWriter *RangeListssWriter = nullptr;
   if (CU.getVersion() == 5) {
-    assert(RangeListsWritersByCU.count(DWOId) != 0 &&
-           "No RangeListsWriter for DWO ID.");
-    RangeListssWriter = RangeListsWritersByCU[DWOId].get();
+    RangeListssWriter =
+        llvm::dyn_cast<DebugRangeListsSectionWriter>(&TempRangesSectionWriter);
 
     // Handling .debug_rnglists.dwo separately. The original .o/.dwo might not
     // have .debug_rnglists so won't be part of the loop below.

bolt/test/X86/dwarf4-cross-cu-backward-different-abbrev.test

@@ -3,7 +3,7 @@
 # RUN: llvm-mc -dwarf-version=4 -filetype=obj -triple x86_64-unknown-linux %p/Inputs/dwarf4-cross-reference-different-abbrev-dst.s -o %t.o
 # RUN: llvm-mc -dwarf-version=4 -filetype=obj -triple x86_64-unknown-linux %p/Inputs/dwarf4-cross-reference-different-abbrev-src.s -o %t1.o
 # RUN: %clang %cflags -gdwarf-4 %t.o %t1.o -o %t.exe
-# RUN: llvm-bolt %t.exe -o %t.bolt --update-debug-sections
+# RUN: llvm-bolt %t.exe -o %t.bolt --update-debug-sections --debug-thread-count=4 --cu-processing-batch-size=4
 # RUN: llvm-dwarfdump --show-form --verbose --debug-info %t.exe | FileCheck --check-prefix=PRECHECK %s
 # RUN: llvm-dwarfdump --show-form --verbose --debug-info %t.bolt | FileCheck --check-prefix=POSTCHECK %s
 

bolt/test/X86/dwarf4-cross-cu-forward-different-abbrev.test

@@ -3,7 +3,7 @@
 # RUN: llvm-mc -dwarf-version=4 -filetype=obj -triple x86_64-unknown-linux %p/Inputs/dwarf4-cross-reference-different-abbrev-dst.s -o %t.o
 # RUN: llvm-mc -dwarf-version=4 -filetype=obj -triple x86_64-unknown-linux %p/Inputs/dwarf4-cross-reference-different-abbrev-src.s -o %t1.o
 # RUN: %clang %cflags -gdwarf-4 %t1.o %t.o -o %t.exe
-# RUN: llvm-bolt %t.exe -o %t.bolt --update-debug-sections
+# RUN: llvm-bolt %t.exe -o %t.bolt --update-debug-sections --debug-thread-count=4 --cu-processing-batch-size=4
 # RUN: llvm-dwarfdump --show-form --verbose --debug-info %t.exe | FileCheck --check-prefix=PRECHECK %s
 # RUN: llvm-dwarfdump --show-form --verbose --debug-info %t.bolt | FileCheck --check-prefix=POSTCHECK %s