LLVM and SPIRV-LLVM-Translator pulldown (WW50)

LLVM: llvm/llvm-project@9bd32d7
SPIRV-LLVM-Translator: KhronosGroup/SPIRV-LLVM-Translator@0166a0f

Author: bb-sycl

.git-blame-ignore-revs

@@ -67,3 +67,6 @@ f6d557ee34b6bbdb1dc32f29e34b4a4a8ad35e81
 
 # [libc++] Rename _LIBCPP_INLINE_VISIBILITY to _LIBCPP_HIDE_FROM_ABI
 4c198542226223f6a5c5511a1f89b37d15ee10b9
+
+# [libc++] Replace uses of _VSTD:: by std:: (#74331)
+77a00c0d546cd4aa8311b5b9031ae9ea8cdb050c

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

@@ -160,6 +160,7 @@ jobs:
           'generic-no-tzdb',
           'generic-no-unicode',
           'generic-no-wide-characters',
+          'generic-no-rtti',
           'generic-static',
           'generic-with_llvm_unwinder',
           # TODO Find a better place for the benchmark and bootstrapping builds to live. They're either very expensive

.github/workflows/llvm-project-tests.yml

@@ -10,6 +10,11 @@ on:
         required: false
       projects:
         required: false
+      extra_cmake_args:
+        required: false
+      os_list:
+        required: false
+        default: '["ubuntu-latest", "windows-2019", "macOS-11"]'
   workflow_call:
     inputs:
       build_target:
@@ -20,6 +25,19 @@ on:
         required: true
         type: string
 
+      extra_cmake_args:
+        required: false
+        type: string
+
+      os_list:
+        required: false
+        type: string
+        # Use windows-2019 due to:
+        # https://developercommunity.visualstudio.com/t/Prev-Issue---with-__assume-isnan-/1597317
+        # We're using a specific version of macOS due to:
+        # https://github.com/actions/virtual-environments/issues/5900
+        default: '["ubuntu-latest", "windows-2019", "macOS-11"]'
+
 concurrency:
   # Skip intermediate builds: always.
   # Cancel intermediate builds: only if it is a pull request build.
@@ -35,14 +53,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os:
-          - ubuntu-latest
-          # Use windows-2019 due to:
-          # https://developercommunity.visualstudio.com/t/Prev-Issue---with-__assume-isnan-/1597317
-          - windows-2019
-          # We're using a specific version of macOS due to:
-          # https://github.com/actions/virtual-environments/issues/5900
-          - macOS-11
+        os: ${{ fromJSON(inputs.os_list) }}
     steps:
       - name: Setup Windows
         if: startsWith(matrix.os, 'windows')
@@ -85,7 +96,7 @@ jobs:
           # This should be a no-op for non-mac OSes
           PKG_CONFIG_PATH: /usr/local/Homebrew/Library/Homebrew/os/mac/pkgconfig//12
         with:
-          cmake_args: '-GNinja -DLLVM_ENABLE_PROJECTS="${{ inputs.projects }}" -DCMAKE_BUILD_TYPE=Release -DLLDB_INCLUDE_TESTS=OFF -DCMAKE_C_COMPILER_LAUNCHER=sccache -DCMAKE_CXX_COMPILER_LAUNCHER=sccache'
+          cmake_args: '-GNinja -DLLVM_ENABLE_PROJECTS="${{ inputs.projects }}" -DCMAKE_BUILD_TYPE=Release -DLLVM_ENABLE_ASSERTIONS=ON -DLLDB_INCLUDE_TESTS=OFF -DCMAKE_C_COMPILER_LAUNCHER=sccache -DCMAKE_CXX_COMPILER_LAUNCHER=sccache ${{ inputs.extra_cmake_args }}'
           build_target: '${{ inputs.build_target }}'
 
       - name: Build and Test libclc

.github/workflows/new-prs.yml

@@ -20,12 +20,19 @@ jobs:
     permissions:
       pull-requests: write
     # Only comment on PRs that have been opened for the first time, by someone
-    # new to LLVM or to GitHub as a whole.
+    # new to LLVM or to GitHub as a whole. Ideally we'd look for FIRST_TIMER
+    # or FIRST_TIME_CONTRIBUTOR, but this does not appear to work. Instead check
+    # that we do not have any of the other author associations.
+    # See https://docs.github.com/en/webhooks/webhook-events-and-payloads?actionType=opened#pull_request
+    # for all the possible values.
     if: >-
       (github.repository == 'llvm/llvm-project') &&
       (github.event.action == 'opened') &&
-      (github.event.pull_request.author_association == 'FIRST_TIME_CONTRIBUTOR' ||
-       github.event.pull_request.author_association == 'FIRST_TIMER')
+      (github.event.pull_request.author_association != 'COLLABORATOR') &&
+      (github.event.pull_request.author_association != 'CONTRIBUTOR') &&
+      (github.event.pull_request.author_association != 'MANNEQUIN') &&
+      (github.event.pull_request.author_association != 'MEMBER') &&
+      (github.event.pull_request.author_association != 'OWNER')
     steps:
       - name: Setup Automation Script
         run: |

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

@@ -67,10 +67,14 @@ jobs:
           START_REV: ${{ github.event.pull_request.base.sha }}
           END_REV: ${{ github.event.pull_request.head.sha }}
           CHANGED_FILES: ${{ steps.changed-files.outputs.all_changed_files }}
+        # TODO(boomanaiden154): Once clang v18 is released, we should be able
+        # to take advantage of the new --diff_from_common_commit option
+        # explicitly in code-format-helper.py and not have to diff starting at
+        # the merge base.
         run: |
           python ./code-format-tools/llvm/utils/git/code-format-helper.py \
             --token ${{ secrets.GITHUB_TOKEN }} \
             --issue-number $GITHUB_PR_NUMBER \
-            --start-rev $START_REV \
+            --start-rev $(git merge-base $START_REV $END_REV) \
             --end-rev $END_REV \
             --changed-files "$CHANGED_FILES"

.github/workflows/spirv-tests.yml

@@ -0,0 +1,29 @@
+name: SPIR-V Tests
+
+permissions:
+  contents: read
+
+on:
+  workflow_dispatch:
+  pull_request:
+    paths:
+      - 'llvm/lib/Target/SPIRV/**'
+      - 'llvm/test/CodeGen/SPIRV/**'
+      - '.github/workflows/spirv-tests.yml'
+
+concurrency:
+  # Skip intermediate builds: always.
+  # Cancel intermediate builds: only if it is a pull request build.
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ startsWith(github.ref, 'refs/pull/') }}
+
+jobs:
+  check_spirv:
+    if: github.repository_owner == 'llvm'
+    name: Test SPIR-V
+    uses: ./.github/workflows/llvm-project-tests.yml
+    with:
+      build_target: check-llvm-codegen-spirv
+      projects:
+      extra_cmake_args: '-DLLVM_TARGETS_TO_BUILD="" -DLLVM_EXPERIMENTAL_TARGETS_TO_BUILD="SPIRV"'
+      os_list: '["ubuntu-latest"]'

.mailmap

@@ -33,6 +33,8 @@
 <[email protected]> <[email protected]>
 <[email protected]> <[email protected]>
 <[email protected]> <[email protected]>
+Jianjian GUAN <[email protected]>
+Jianjian GUAN <[email protected]> <[email protected]>
 Jon Roelofs <[email protected]> <[email protected]>
 Jon Roelofs <[email protected]> <[email protected]>
 Jonathan Thackray <[email protected]> <[email protected]>

bolt/include/bolt/Core/BinaryFunction.h

@@ -75,6 +75,14 @@ enum IndirectCallPromotionType : char {
   ICP_ALL          /// Perform ICP on calls and jump tables.
 };
 
+/// Hash functions supported for BF/BB hashing.
+enum class HashFunction : char {
+  StdHash, /// std::hash, implementation is platform-dependent. Provided for
+           /// backwards compatibility.
+  XXH3,    /// llvm::xxh3_64bits, the default.
+  Default = XXH3,
+};
+
 /// Information on a single indirect call to a particular callee.
 struct IndirectCallProfile {
   MCSymbol *Symbol;
@@ -2234,18 +2242,21 @@ class BinaryFunction {
   ///
   /// If \p UseDFS is set, process basic blocks in DFS order. Otherwise, use
   /// the existing layout order.
+  /// \p HashFunction specifies which function is used for BF hashing.
   ///
   /// By default, instruction operands are ignored while calculating the hash.
   /// The caller can change this via passing \p OperandHashFunc function.
   /// The return result of this function will be mixed with internal hash.
   size_t computeHash(
-      bool UseDFS = false,
+      bool UseDFS = false, HashFunction HashFunction = HashFunction::Default,
       OperandHashFuncTy OperandHashFunc = [](const MCOperand &) {
         return std::string();
       }) const;
 
   /// Compute hash values for each block of the function.
-  void computeBlockHashes() const;
+  /// \p HashFunction specifies which function is used for BB hashing.
+  void
+  computeBlockHashes(HashFunction HashFunction = HashFunction::Default) const;
 
   void setDWARFUnit(DWARFUnit *Unit) { DwarfUnit = Unit; }
 

bolt/include/bolt/Core/DebugData.h

@@ -459,8 +459,6 @@ class DebugStrOffsetsWriter {
   std::unique_ptr<raw_svector_ostream> StrOffsetsStream;
   std::map<uint32_t, uint32_t> IndexToAddressMap;
   std::unordered_map<uint64_t, uint64_t> ProcessedBaseOffsets;
-  // Section size not including header.
-  uint32_t CurrentSectionSize{0};
   bool StrOffsetSectionWasModified = false;
 };
 

bolt/include/bolt/Profile/ProfileYAMLMapping.h

@@ -178,6 +178,14 @@ template <> struct ScalarBitSetTraits<PROFILE_PF> {
   }
 };
 
+template <> struct ScalarEnumerationTraits<llvm::bolt::HashFunction> {
+  using HashFunction = llvm::bolt::HashFunction;
+  static void enumeration(IO &io, HashFunction &value) {
+    io.enumCase(value, "std-hash", HashFunction::StdHash);
+    io.enumCase(value, "xxh3", HashFunction::XXH3);
+  }
+};
+
 namespace bolt {
 struct BinaryProfileHeader {
   uint32_t Version{1};
@@ -188,6 +196,7 @@ struct BinaryProfileHeader {
   std::string Origin;     // How the profile was obtained.
   std::string EventNames; // Events used for sample profile.
   bool IsDFSOrder{true};  // Whether using DFS block order in function profile
+  llvm::bolt::HashFunction HashFunction; // Hash used for BB/BF hashing
 };
 } // end namespace bolt
 
@@ -200,6 +209,8 @@ template <> struct MappingTraits<bolt::BinaryProfileHeader> {
     YamlIO.mapOptional("profile-origin", Header.Origin);
     YamlIO.mapOptional("profile-events", Header.EventNames);
     YamlIO.mapOptional("dfs-order", Header.IsDFSOrder);
+    YamlIO.mapOptional("hash-func", Header.HashFunction,
+                       llvm::bolt::HashFunction::StdHash);
   }
 };
 

bolt/lib/Core/BinaryFunction.cpp

@@ -3633,7 +3633,7 @@ BinaryFunction::BasicBlockListType BinaryFunction::dfs() const {
   return DFS;
 }
 
-size_t BinaryFunction::computeHash(bool UseDFS,
+size_t BinaryFunction::computeHash(bool UseDFS, HashFunction HashFunction,
                                    OperandHashFuncTy OperandHashFunc) const {
   if (size() == 0)
     return 0;
@@ -3652,7 +3652,13 @@ size_t BinaryFunction::computeHash(bool UseDFS,
   for (const BinaryBasicBlock *BB : Order)
     HashString.append(hashBlock(BC, *BB, OperandHashFunc));
 
-  return Hash = llvm::xxh3_64bits(HashString);
+  switch (HashFunction) {
+  case HashFunction::StdHash:
+    return Hash = std::hash<std::string>{}(HashString);
+  case HashFunction::XXH3:
+    return Hash = llvm::xxh3_64bits(HashString);
+  }
+  llvm_unreachable("Unhandled HashFunction");
 }
 
 void BinaryFunction::insertBasicBlocks(

bolt/lib/Core/BinaryFunctionProfile.cpp

@@ -225,6 +225,7 @@ void BinaryFunction::mergeProfileDataInto(BinaryFunction &BF) const {
     for (const BinaryBasicBlock *BBSucc : BB->successors()) {
       (void)BBSucc;
       assert(getIndex(BBSucc) == BF.getIndex(*BBMergeSI));
+      (void)BBMergeSI;
 
       // At this point no branch count should be set to COUNT_NO_PROFILE.
       assert(BII->Count != BinaryBasicBlock::COUNT_NO_PROFILE &&

bolt/lib/Core/DebugData.cpp

@@ -889,8 +889,10 @@ void DebugStrOffsetsWriter::finalizeSection(DWARFUnit &Unit,
   // Handling re-use of str-offsets section.
   if (RetVal == ProcessedBaseOffsets.end() || StrOffsetSectionWasModified) {
     // Writing out the header for each section.
-    support::endian::write(*StrOffsetsStream, CurrentSectionSize + 4,
-                           llvm::endianness::little);
+    support::endian::write(
+        *StrOffsetsStream,
+        static_cast<uint32_t>(IndexToAddressMap.size() * 4 + 4),
+        llvm::endianness::little);
     support::endian::write(*StrOffsetsStream, static_cast<uint16_t>(5),
                            llvm::endianness::little);
     support::endian::write(*StrOffsetsStream, static_cast<uint16_t>(0),

bolt/lib/Core/Exceptions.cpp

@@ -108,7 +108,8 @@ void BinaryFunction::parseLSDA(ArrayRef<uint8_t> LSDASectionData,
   DWARFDataExtractor Data(
       StringRef(reinterpret_cast<const char *>(LSDASectionData.data()),
                 LSDASectionData.size()),
-      BC.DwCtx->getDWARFObj().isLittleEndian(), 8);
+      BC.DwCtx->getDWARFObj().isLittleEndian(),
+      BC.DwCtx->getDWARFObj().getAddressSize());
   uint64_t Offset = getLSDAAddress() - LSDASectionAddress;
   assert(Data.isValidOffset(Offset) && "wrong LSDA address");
 

bolt/lib/Passes/IdenticalCodeFolding.cpp

@@ -360,9 +360,9 @@ void IdenticalCodeFolding::runOnFunctions(BinaryContext &BC) {
 
       // Pre-compute hash before pushing into hashtable.
       // Hash instruction operands to minimize hash collisions.
-      BF.computeHash(opts::ICFUseDFS, [&BC](const MCOperand &Op) {
-        return hashInstOperand(BC, Op);
-      });
+      BF.computeHash(
+          opts::ICFUseDFS, HashFunction::Default,
+          [&BC](const MCOperand &Op) { return hashInstOperand(BC, Op); });
     };
 
     ParallelUtilities::PredicateTy SkipFunc = [&](const BinaryFunction &BF) {

bolt/lib/Passes/IndirectCallPromotion.cpp

@@ -158,6 +158,7 @@ static cl::opt<bool> ICPPeelForInline(
 
 } // namespace opts
 
+#ifndef NDEBUG
 static bool verifyProfile(std::map<uint64_t, BinaryFunction> &BFs) {
   bool IsValid = true;
   for (auto &BFI : BFs) {
@@ -182,6 +183,7 @@ static bool verifyProfile(std::map<uint64_t, BinaryFunction> &BFs) {
   }
   return IsValid;
 }
+#endif
 
 namespace llvm {
 namespace bolt {
@@ -1467,7 +1469,6 @@ void IndirectCallPromotion::runOnFunctions(BinaryContext &BC) {
                                std::max<uint64_t>(TotalIndexBasedCandidates, 1))
          << "%\n";
 
-  (void)verifyProfile;
 #ifndef NDEBUG
   verifyProfile(BFs);
 #endif

bolt/lib/Passes/LongJmp.cpp

@@ -202,10 +202,23 @@ LongJmpPass::replaceTargetWithStub(BinaryBasicBlock &BB, MCInst &Inst,
     }
   } else if (LocalStubsIter != Stubs.end() &&
              LocalStubsIter->second.count(TgtBB)) {
-    // If we are replacing a local stub (because it is now out of range),
-    // use its target instead of creating a stub to jump to another stub
+    // The TgtBB and TgtSym now are the local out-of-range stub and its label.
+    // So, we are attempting to restore BB to its previous state without using
+    // this stub.
     TgtSym = BC.MIB->getTargetSymbol(*TgtBB->begin());
-    TgtBB = BB.getSuccessor(TgtSym, BI);
+    assert(TgtSym &&
+           "First instruction is expected to contain a target symbol.");
+    BinaryBasicBlock *TgtBBSucc = TgtBB->getSuccessor(TgtSym, BI);
+
+    // TgtBB might have no successor. e.g. a stub for a function call.
+    if (TgtBBSucc) {
+      BB.replaceSuccessor(TgtBB, TgtBBSucc, BI.Count, BI.MispredictedCount);
+      assert(TgtBB->getExecutionCount() >= BI.Count &&
+             "At least equal or greater than the branch count.");
+      TgtBB->setExecutionCount(TgtBB->getExecutionCount() - BI.Count);
+    }
+
+    TgtBB = TgtBBSucc;
   }
 
   BinaryBasicBlock *StubBB = lookupLocalStub(BB, Inst, TgtSym, DotAddress);

bolt/lib/Passes/ReorderAlgorithm.cpp

@@ -20,7 +20,7 @@
 #include <random>
 #include <stack>
 
-#undef  DEBUG_TYPE
+#undef DEBUG_TYPE
 #define DEBUG_TYPE "bolt"
 
 using namespace llvm;
@@ -425,7 +425,7 @@ void TSPReorderAlgorithm::reorderBasicBlocks(BinaryFunction &BF,
   }
 
   std::vector<std::vector<int64_t>> DP;
-  DP.resize(1 << N);
+  DP.resize(static_cast<size_t>(1) << N);
   for (std::vector<int64_t> &Elmt : DP)
     Elmt.resize(N, -1);
 

bolt/lib/Passes/VeneerElimination.cpp

@@ -89,6 +89,7 @@ void VeneerElimination::runOnFunctions(BinaryContext &BC) {
   LLVM_DEBUG(
       dbgs() << "BOLT-INFO: number of linker-inserted veneers call sites: "
              << VeneerCallers << "\n");
+  (void)VeneerCallers;
 }
 
 } // namespace bolt