Merged main:5bd1b93cb220 into amd-gfx:18b1fa897044

Local branch amd-gfx 18b1fa8 Merged main:455bf3d1ccd6 into amd-gfx:c7691189ce5b
Remote branch main 5bd1b93 Move CallInst::CreateFree to IRBuilderBase

Author: SC llvm team

.github/CODEOWNERS

@@ -0,0 +1,4 @@
+/libcxx/ @llvm/reviewers-libcxx
+/libcxxabi/ @llvm/reviewers-libcxxabi
+/libunwind/ @llvm/reviewers-libunwind
+/runtimes/ @llvm/reviewers-libcxx

.github/new-issues-labeler.yml

@@ -1,6 +1,3 @@
-'new issue':
-  - '/.*/'
-
 'clang':
   - '/\bclang(?!\-)\b/i'
 

.github/new-prs-labeler.yml

@@ -110,16 +110,26 @@ mlir:
   - mlir/**
 
 mlir:core:
-  - mlir/**/Support/**
-  - mlir/**/Parser/**
-  - mlir/**/IR/**
-  - mlir/**/Bytecode/**
-  - mlir/**/AsmParser/**
-  - mlir/**/Pass/**
-  - mlir/**/tools/**
-  - mlir/**/Reducer/**
-  - mlir/**/Transforms/**
-  - mlir/**/Debug/**
+  - mlir/include/mlir/Support/**
+  - mlir/lib/Support/**
+  - mlir/include/mlir/Parser/**
+  - mlir/lib/Parser/**
+  - mlir/include/mlir/IR/**
+  - mlir/lib/IR/**
+  - mlir/include/mlir/Bytecode/**
+  - mlir/lib/Bytecode/**
+  - mlir/include/mlir/AsmParser/**
+  - mlir/lib/AsmParser/**
+  - mlir/include/mlir/Pass/**
+  - mlir/lib/Pass/**
+  - mlir/include/mlir/Tools/**
+  - mlir/lib/Tools/**
+  - mlir/include/mlir/Reducer/**
+  - mlir/lib/Reducer/**
+  - mlir/include/mlir/Transforms/**
+  - mlir/lib/Transforms/**
+  - mlir/include/mlir/Debug/**
+  - mlir/lib/Debug/**
   - mlir/tools/**
 
 mlir:ods:
@@ -300,7 +310,10 @@ vectorizers:
   - llvm/include/llvm/Transforms/Vectorize/**
 
 clang:
-  - clang/**
+  - any:
+    - clang/**
+    - '!clang/**/Format/**'
+    - '!clang/tools/clang-format/**'
 
 testing-tools:
   - llvm/include/llvm/FileCheck/**
@@ -510,6 +523,12 @@ llvm:analysis:
   - llvm/test/Analysis/**
   - llvm/unittests/Analysis/**
 
+llvm:adt:
+  - llvm/**/ADT/*
+
+llvm:support:
+  - llvm/**/Support/**
+
 llvm:transforms:
   - llvm/lib/Transforms/**
   - llvm/include/llvm/Transforms/**
@@ -554,6 +573,13 @@ backend:DirectX:
   - '**/*DXIL*/**'
   - '**/*dxil*/**'
 
+backend:SPIR-V:
+  - clang/lib/Driver/ToolChains/SPIRV.*
+  - llvm/lib/Target/SPIRV/**
+  - llvm/test/CodeGen/SPIRV/**
+  - llvm/test/Frontend/HLSL/**
+  - llvm/docs/SPIRVUsage.rst
+
 mlgo:
   - llvm/lib/Analysis/ML*
   - llvm/include/llvm/Analysis/ML*

.github/workflows/new-issues.yml

@@ -13,7 +13,7 @@ jobs:
     runs-on: ubuntu-latest
     if: github.repository == 'llvm/llvm-project'
     steps:
-      - uses: github/issue-labeler@v3.2
+      - uses: llvm/actions/issue-labeler@main
         with:
           repo-token: ${{ secrets.GITHUB_TOKEN }}
           configuration-path: .github/new-issues-labeler.yml

.github/workflows/pr-receive.yml

@@ -10,7 +10,12 @@ permissions:
 jobs:
   pr-target:
     runs-on: ubuntu-latest
-    if: github.repository == 'llvm/llvm-project'
+    # Ignore PRs with more than 10 commits.  Pull requests with a lot of
+    # commits tend to be accidents usually when someone made a mistake while trying
+    # to rebase.  We want to ignore these pull requests to avoid excessive
+    # notifications.
+    if: github.repository == 'llvm/llvm-project' &&
+        github.event.pull_request.commits < 10
     steps:
       - name: Store PR Information
         run: |

bolt/include/bolt/Core/MCPlusBuilder.h

@@ -437,7 +437,7 @@ class MCPlusBuilder {
   virtual bool isUnsupportedBranch(const MCInst &Inst) const { return false; }
 
   /// Return true of the instruction is of pseudo kind.
-  bool isPseudo(const MCInst &Inst) const {
+  virtual bool isPseudo(const MCInst &Inst) const {
     return Info->get(Inst.getOpcode()).isPseudo();
   }
 

bolt/include/bolt/Passes/FixRISCVCallsPass.h

@@ -6,9 +6,13 @@
 //
 //===----------------------------------------------------------------------===//
 //
-// This file declares the FixRISCVCallsPass class, which sets the JALR immediate
-// to 0 for AUIPC/JALR pairs with a R_RISCV_CALL(_PLT) relocation. This is
-// necessary since MC expects it to be zero in order to or-in fixups.
+// This file declares the FixRISCVCallsPass class, which replaces all types of
+// calls with PseudoCALL pseudo instructions. This ensures that relaxed calls
+// get expanded to auipc/jalr pairs so that BOLT can freely reassign function
+// addresses without having to worry about the limited range of relaxed calls.
+// Using PseudoCALL also ensures that the RISC-V backend inserts the necessary
+// relaxation-related relocations to allow JITLink to relax instruction back to
+// shorter versions where possible.
 //===----------------------------------------------------------------------===//
 
 #ifndef BOLT_PASSES_FIXRISCVCALLSPASS_H

bolt/lib/Core/BinaryContext.cpp

@@ -131,7 +131,7 @@ BinaryContext::createBinaryContext(const ObjectFile *File, bool IsPIC,
   case llvm::Triple::riscv64:
     ArchName = "riscv64";
     // RV64GC
-    FeaturesStr = "+m,+a,+f,+d,+zicsr,+zifencei,+c";
+    FeaturesStr = "+m,+a,+f,+d,+zicsr,+zifencei,+c,+relax";
     break;
   default:
     return createStringError(std::errc::not_supported,

bolt/lib/Core/BinaryFunction.cpp

@@ -3156,6 +3156,10 @@ void BinaryFunction::dumpGraphToFile(std::string Filename) const {
 }
 
 bool BinaryFunction::validateCFG() const {
+  // Skip the validation of CFG after it is finalized
+  if (CurrentState == State::CFG_Finalized)
+    return true;
+
   bool Valid = true;
   for (BinaryBasicBlock *BB : BasicBlocks)
     Valid &= BB->validateSuccessorInvariants();

bolt/lib/Core/Relocation.cpp

@@ -352,7 +352,7 @@ static uint64_t encodeValueAArch64(uint64_t Type, uint64_t Value, uint64_t PC) {
     assert(isInt<28>(Value) && "only PC +/- 128MB is allowed for direct call");
     // Immediate goes in bits 25:0 of BL.
     // OP 1001_01 goes in bits 31:26 of BL.
-    Value = (Value >> 2) | 0x94000000ULL;
+    Value = ((Value >> 2) & 0x3ffffff) | 0x94000000ULL;
     break;
   }
   return Value;

bolt/lib/Passes/FixRISCVCallsPass.cpp

@@ -10,23 +10,42 @@ namespace bolt {
 
 void FixRISCVCallsPass::runOnFunction(BinaryFunction &BF) {
   auto &BC = BF.getBinaryContext();
+  auto &MIB = BC.MIB;
+  auto *Ctx = BC.Ctx.get();
 
   for (auto &BB : BF) {
-    for (auto II = BB.begin(), IE = BB.end(); II != IE; ++II) {
+    for (auto II = BB.begin(); II != BB.end();) {
+      if (MIB->isCall(*II) && !MIB->isIndirectCall(*II)) {
+        auto *Target = MIB->getTargetSymbol(*II);
+        assert(Target && "Cannot find call target");
+
+        auto L = BC.scopeLock();
+
+        if (MIB->isTailCall(*II))
+          MIB->createTailCall(*II, Target, Ctx);
+        else
+          MIB->createCall(*II, Target, Ctx);
+
+        ++II;
+        continue;
+      }
+
       auto NextII = std::next(II);
 
-      if (NextII == IE)
+      if (NextII == BB.end())
         break;
 
-      if (!BC.MIB->isRISCVCall(*II, *NextII))
-        continue;
+      if (MIB->isRISCVCall(*II, *NextII)) {
+        auto *Target = MIB->getTargetSymbol(*II);
+        assert(Target && "Cannot find call target");
 
-      auto L = BC.scopeLock();
+        auto L = BC.scopeLock();
+        MIB->createCall(*II, Target, Ctx);
+        II = BB.eraseInstruction(NextII);
+        continue;
+      }
 
-      // The MC layer handles R_RISCV_CALL_PLT but assumes that the immediate
-      // in the JALR is zero (fixups are or'ed into instructions). Note that
-      // NextII is guaranteed to point to a JALR by isRISCVCall.
-      NextII->getOperand(2).setImm(0);
+      ++II;
     }
   }
 }

bolt/lib/Passes/TailDuplication.cpp

@@ -303,7 +303,7 @@ TailDuplication::aggressiveDuplicate(BinaryBasicBlock &BB,
   if (isInCacheLine(BB, Tail))
     return BlocksToDuplicate;
 
-  BinaryBasicBlock *CurrBB = &BB;
+  BinaryBasicBlock *CurrBB = &Tail;
   while (CurrBB) {
     LLVM_DEBUG(dbgs() << "Aggressive tail duplication: adding "
                       << CurrBB->getName() << " to duplication list\n";);

bolt/lib/Profile/YAMLProfileReader.cpp

@@ -294,9 +294,19 @@ Error YAMLProfileReader::preprocessProfile(BinaryContext &BC) {
   buildNameMaps(BC);
 
   // Preliminary assign function execution count.
-  for (auto [YamlBF, BF] : llvm::zip_equal(YamlBP.Functions, ProfileBFs))
-    if (BF)
+  for (auto [YamlBF, BF] : llvm::zip_equal(YamlBP.Functions, ProfileBFs)) {
+    if (!BF)
+      continue;
+    if (!BF->hasProfile()) {
       BF->setExecutionCount(YamlBF.ExecCount);
+    } else {
+      if (opts::Verbosity >= 1) {
+        errs() << "BOLT-WARNING: dropping duplicate profile for " << YamlBF.Name
+               << '\n';
+      }
+      BF = nullptr;
+    }
+  }
 
   return Error::success();
 }

bolt/lib/Rewrite/JITLinkLinker.cpp

@@ -8,6 +8,7 @@
 #include "bolt/Rewrite/JITLinkLinker.h"
 #include "bolt/Core/BinaryData.h"
 #include "bolt/Rewrite/RewriteInstance.h"
+#include "llvm/ExecutionEngine/JITLink/ELF_riscv.h"
 #include "llvm/ExecutionEngine/JITLink/JITLink.h"
 #include "llvm/ExecutionEngine/Orc/Shared/ExecutorAddress.h"
 #include "llvm/ExecutionEngine/Orc/Shared/ExecutorSymbolDef.h"
@@ -93,6 +94,18 @@ struct JITLinkLinker::Context : jitlink::JITLinkContext {
   Error modifyPassConfig(jitlink::LinkGraph &G,
                          jitlink::PassConfiguration &Config) override {
     Config.PrePrunePasses.push_back(markSectionsLive);
+    Config.PostAllocationPasses.push_back([this](auto &G) {
+      MapSections([&G](const BinarySection &Section, uint64_t Address) {
+        reassignSectionAddress(G, Section, Address);
+      });
+      return Error::success();
+    });
+
+    if (G.getTargetTriple().isRISCV()) {
+      Config.PostAllocationPasses.push_back(
+          jitlink::createRelaxationPass_ELF_riscv());
+    }
+
     return Error::success();
   }
 
@@ -137,10 +150,6 @@ struct JITLinkLinker::Context : jitlink::JITLinkContext {
   }
 
   Error notifyResolved(jitlink::LinkGraph &G) override {
-    MapSections([&G](const BinarySection &Section, uint64_t Address) {
-      reassignSectionAddress(G, Section, Address);
-    });
-
     for (auto *Symbol : G.defined_symbols()) {
       SymbolInfo Info{Symbol->getAddress().getValue(), Symbol->getSize()};
       Linker.Symtab.insert({Symbol->getName().str(), Info});

bolt/lib/Rewrite/MachORewriteInstance.cpp

@@ -564,8 +564,10 @@ void MachORewriteInstance::rewriteFile() {
   writeInstrumentationSection("I__literal16", OS);
 
   Out->keep();
-  EC = sys::fs::setPermissions(opts::OutputFilename,
-                               sys::fs::perms::all_all);
+  EC = sys::fs::setPermissions(
+      opts::OutputFilename,
+      static_cast<sys::fs::perms>(sys::fs::perms::all_all &
+                                  ~sys::fs::getUmask()));
   check_error(EC, "cannot set permissions of output file");
 }