Merge remote-tracking branch 'origin/sycl-web' into llvmspirv_pulldown

Author: sys-ce-bb

.github/workflows/issue-write.yml

@@ -5,6 +5,7 @@ on:
     workflows:
       - "Check code formatting"
       - "Check for private emails used in PRs"
+      - "PR Request Release Note"
     types:
       - completed
 
@@ -17,7 +18,11 @@ jobs:
     permissions:
       pull-requests: write
     if: >
-      github.event.workflow_run.event == 'pull_request'
+      github.event.workflow_run.event == 'pull_request' &&
+      (
+        github.event.workflow_run.conclusion == 'success' ||
+        github.event.workflow_run.conclusion == 'failure'
+      )
     steps:
       - name: 'Download artifact'
         uses: actions/download-artifact@6b208ae046db98c579e8a3aa621ab581ff575935 # v4.1.1
@@ -92,7 +97,11 @@ jobs:
 
             var pr_number = 0;
             gql_result.repository.ref.associatedPullRequests.nodes.forEach((pr) => {
-              if (pr.baseRepository.owner.login = context.repo.owner && pr.state == 'OPEN') {
+
+              // The largest PR number is the one we care about.  The only way
+              // to have more than one associated pull requests is if all the
+              // old pull requests are in the closed state.
+              if (pr.baseRepository.owner.login = context.repo.owner && pr.number > pr_number) {
                 pr_number = pr.number;
               }
             });

.github/workflows/pr-request-release-note.yml

@@ -2,7 +2,6 @@ name: PR Request Release Note
 
 permissions:
   contents: read
-  pull-requests: write
 
 on:
   pull_request:
@@ -41,3 +40,10 @@ jobs:
             --token "$GITHUB_TOKEN" \
             request-release-note \
             --pr-number ${{ github.event.pull_request.number}}
+
+      - uses: actions/upload-artifact@26f96dfa697d77e81fd5907df203aa23a56210a8 #v4.3.0
+        if: always()
+        with:
+          name: workflow-args
+          path: |
+            comments

bolt/docs/CommandLineArgumentReference.md

@@ -259,6 +259,10 @@
 
   Always use long jumps/nops for Linux kernel static keys
 
+- `--match-profile-with-function-hash`
+
+  Match profile with function hash
+
 - `--max-data-relocations=<uint>`
 
   Maximum number of data relocations to process

bolt/include/bolt/Core/BinaryFunction.h

@@ -930,6 +930,10 @@ class BinaryFunction {
     return const_cast<BinaryFunction *>(this)->getInstructionAtOffset(Offset);
   }
 
+  /// When the function is in disassembled state, return an instruction that
+  /// contains the \p Offset.
+  MCInst *getInstructionContainingOffset(uint64_t Offset);
+
   std::optional<MCInst> disassembleInstructionAtOffset(uint64_t Offset) const;
 
   /// Return offset for the first instruction. If there is data at the

bolt/include/bolt/Passes/BinaryFunctionCallGraph.h renamed to bolt/include/bolt/Core/BinaryFunctionCallGraph.h

@@ -1,4 +1,4 @@
-//===- bolt/Passes/CallGraph.h ----------------------------------*- C++ -*-===//
+//===- bolt/Core/CallGraph.h ----------------------------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -9,7 +9,7 @@
 #ifndef BOLT_PASSES_BINARY_FUNCTION_CALLGRAPH_H
 #define BOLT_PASSES_BINARY_FUNCTION_CALLGRAPH_H
 
-#include "bolt/Passes/CallGraph.h"
+#include "bolt/Core/CallGraph.h"
 #include <deque>
 #include <functional>
 #include <unordered_map>

bolt/include/bolt/Passes/CallGraph.h renamed to bolt/include/bolt/Core/CallGraph.h

@@ -1,4 +1,4 @@
-//===- bolt/Passes/CallGraph.h ----------------------------------*- C++ -*-===//
+//===- bolt/Core/CallGraph.h ----------------------------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.

bolt/include/bolt/Passes/CallGraphWalker.h renamed to bolt/include/bolt/Core/CallGraphWalker.h

@@ -1,4 +1,4 @@
-//===- bolt/Passes/CallGraphWalker.h ----------------------------*- C++ -*-===//
+//===- bolt/Core/CallGraphWalker.h ----------------------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.

bolt/include/bolt/Core/MCPlusBuilder.h

@@ -439,7 +439,20 @@ class MCPlusBuilder {
   }
 
   /// Check whether this conditional branch can be reversed
-  virtual bool isReversibleBranch(const MCInst &Inst) const { return true; }
+  virtual bool isReversibleBranch(const MCInst &Inst) const {
+    assert(!isUnsupportedInstruction(Inst) && isConditionalBranch(Inst) &&
+           "Instruction is not known conditional branch");
+
+    if (isDynamicBranch(Inst))
+      return false;
+    return true;
+  }
+
+  /// Return true if this instruction inhibits analysis of the containing
+  /// function.
+  virtual bool isUnsupportedInstruction(const MCInst &Inst) const {
+    return false;
+  }
 
   /// Return true of the instruction is of pseudo kind.
   virtual bool isPseudo(const MCInst &Inst) const {

bolt/include/bolt/Passes/HFSort.h

@@ -19,7 +19,7 @@
 #ifndef BOLT_PASSES_HFSORT_H
 #define BOLT_PASSES_HFSORT_H
 
-#include "bolt/Passes/CallGraph.h"
+#include "bolt/Core/CallGraph.h"
 
 #include <string>
 #include <vector>

bolt/include/bolt/Passes/RegReAssign.h

@@ -9,7 +9,7 @@
 #ifndef BOLT_PASSES_REGREASSIGN_H
 #define BOLT_PASSES_REGREASSIGN_H
 
-#include "bolt/Passes/BinaryFunctionCallGraph.h"
+#include "bolt/Core/BinaryFunctionCallGraph.h"
 #include "bolt/Passes/BinaryPasses.h"
 #include "bolt/Passes/RegAnalysis.h"
 

bolt/include/bolt/Passes/ReorderFunctions.h

@@ -9,7 +9,7 @@
 #ifndef BOLT_PASSES_REORDER_FUNCTIONS_H
 #define BOLT_PASSES_REORDER_FUNCTIONS_H
 
-#include "bolt/Passes/BinaryFunctionCallGraph.h"
+#include "bolt/Core/BinaryFunctionCallGraph.h"
 #include "bolt/Passes/BinaryPasses.h"
 
 namespace llvm {

bolt/include/bolt/Rewrite/DWARFRewriter.h

@@ -16,7 +16,6 @@
 #include "llvm/ADT/StringRef.h"
 #include "llvm/CodeGen/DIE.h"
 #include "llvm/DWP/DWP.h"
-#include "llvm/MC/MCAsmLayout.h"
 #include "llvm/MC/MCContext.h"
 #include "llvm/Support/ToolOutputFile.h"
 #include <cstdint>
@@ -183,7 +182,7 @@ class DWARFRewriter {
   void updateDebugInfo();
 
   /// Update stmt_list for CUs based on the new .debug_line \p Layout.
-  void updateLineTableOffsets(const MCAsmLayout &Layout);
+  void updateLineTableOffsets(const MCAssembler &Asm);
 
   uint64_t getDwoRangesBase(uint64_t DWOId) { return DwoRangesBase[DWOId]; }
 

bolt/lib/Core/BinaryContext.cpp

@@ -20,7 +20,6 @@
 #include "llvm/DebugInfo/DWARF/DWARFCompileUnit.h"
 #include "llvm/DebugInfo/DWARF/DWARFFormValue.h"
 #include "llvm/DebugInfo/DWARF/DWARFUnit.h"
-#include "llvm/MC/MCAsmLayout.h"
 #include "llvm/MC/MCAssembler.h"
 #include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCDisassembler/MCDisassembler.h"
@@ -2416,20 +2415,19 @@ BinaryContext::calculateEmittedSize(BinaryFunction &BF, bool FixBranches) {
 
   MCAssembler &Assembler =
       static_cast<MCObjectStreamer *>(Streamer.get())->getAssembler();
-  MCAsmLayout Layout(Assembler);
-  Assembler.layout(Layout);
+  Assembler.layout();
 
   // Obtain fragment sizes.
   std::vector<uint64_t> FragmentSizes;
   // Main fragment size.
-  const uint64_t HotSize =
-      Layout.getSymbolOffset(*EndLabel) - Layout.getSymbolOffset(*StartLabel);
+  const uint64_t HotSize = Assembler.getSymbolOffset(*EndLabel) -
+                           Assembler.getSymbolOffset(*StartLabel);
   FragmentSizes.push_back(HotSize);
   // Split fragment sizes.
   uint64_t ColdSize = 0;
   for (const auto &Labels : SplitLabels) {
-    uint64_t Size = Layout.getSymbolOffset(*Labels.second) -
-                    Layout.getSymbolOffset(*Labels.first);
+    uint64_t Size = Assembler.getSymbolOffset(*Labels.second) -
+                    Assembler.getSymbolOffset(*Labels.first);
     FragmentSizes.push_back(Size);
     ColdSize += Size;
   }
@@ -2439,7 +2437,8 @@ BinaryContext::calculateEmittedSize(BinaryFunction &BF, bool FixBranches) {
   for (FunctionFragment &FF : BF.getLayout().fragments()) {
     BinaryBasicBlock *PrevBB = nullptr;
     for (BinaryBasicBlock *BB : FF) {
-      const uint64_t BBStartOffset = Layout.getSymbolOffset(*(BB->getLabel()));
+      const uint64_t BBStartOffset =
+          Assembler.getSymbolOffset(*(BB->getLabel()));
       BB->setOutputStartAddress(BBStartOffset);
       if (PrevBB)
         PrevBB->setOutputEndAddress(BBStartOffset);

bolt/lib/Core/BinaryFunction.cpp

@@ -543,7 +543,7 @@ void BinaryFunction::print(raw_ostream &OS, std::string Annotation) {
         else
           OS << "<unknown>\n";
       }
-      if (BB->getCFIState() >= 0)
+      if (hasCFI())
         OS << "  CFI State : " << BB->getCFIState() << '\n';
       if (opts::EnableBAT) {
         OS << "  Input offset: 0x" << Twine::utohexstr(BB->getInputOffset())
@@ -611,7 +611,7 @@ void BinaryFunction::print(raw_ostream &OS, std::string Annotation) {
       }
 
       // In CFG_Finalized state we can miscalculate CFI state at exit.
-      if (CurrentState == State::CFG) {
+      if (CurrentState == State::CFG && hasCFI()) {
         const int32_t CFIStateAtExit = BB->getCFIStateAtExit();
         if (CFIStateAtExit >= 0)
           OS << "  CFI State: " << CFIStateAtExit << '\n';
@@ -1276,6 +1276,10 @@ Error BinaryFunction::disassemble() {
       }
     }
 
+    bool IsUnsupported = BC.MIB->isUnsupportedInstruction(Instruction);
+    if (IsUnsupported)
+      setIgnored();
+
     if (MIB->isBranch(Instruction) || MIB->isCall(Instruction)) {
       uint64_t TargetAddress = 0;
       if (MIB->evaluateBranch(Instruction, AbsoluteInstrAddr, Size,
@@ -1289,12 +1293,10 @@ Error BinaryFunction::disassemble() {
         const bool IsCondBranch = MIB->isConditionalBranch(Instruction);
         MCSymbol *TargetSymbol = nullptr;
 
-        if (!BC.MIB->isReversibleBranch(Instruction)) {
-          setIgnored();
-          if (BinaryFunction *TargetFunc =
+        if (IsUnsupported)
+          if (auto *TargetFunc =
                   BC.getBinaryFunctionContainingAddress(TargetAddress))
             TargetFunc->setIgnored();
-        }
 
         if (IsCall && containsAddress(TargetAddress)) {
           if (TargetAddress == getAddress()) {
@@ -4472,6 +4474,18 @@ MCInst *BinaryFunction::getInstructionAtOffset(uint64_t Offset) {
   }
 }
 
+MCInst *BinaryFunction::getInstructionContainingOffset(uint64_t Offset) {
+  assert(CurrentState == State::Disassembled && "Wrong function state");
+
+  if (Offset > Size)
+    return nullptr;
+
+  auto II = Instructions.upper_bound(Offset);
+  assert(II != Instructions.begin() && "First instruction not at offset 0");
+  --II;
+  return &II->second;
+}
+
 void BinaryFunction::printLoopInfo(raw_ostream &OS) const {
   if (!opts::shouldPrint(*this))
     return;

bolt/lib/Passes/BinaryFunctionCallGraph.cpp renamed to bolt/lib/Core/BinaryFunctionCallGraph.cpp

@@ -1,4 +1,4 @@
-//===- bolt/Passes/BinaryFunctionCallGraph.cpp ----------------------------===//
+//===- bolt/Core/BinaryFunctionCallGraph.cpp ----------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -10,7 +10,7 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "bolt/Passes/BinaryFunctionCallGraph.h"
+#include "bolt/Core/BinaryFunctionCallGraph.h"
 #include "bolt/Core/BinaryContext.h"
 #include "bolt/Core/BinaryFunction.h"
 #include "llvm/Support/CommandLine.h"

bolt/lib/Core/CMakeLists.txt

@@ -17,8 +17,11 @@ add_llvm_library(LLVMBOLTCore
   BinaryData.cpp
   BinaryEmitter.cpp
   BinaryFunction.cpp
+  BinaryFunctionCallGraph.cpp
   BinaryFunctionProfile.cpp
   BinarySection.cpp
+  CallGraph.cpp
+  CallGraphWalker.cpp
   DebugData.cpp
   DebugNames.cpp
   DIEBuilder.cpp

bolt/lib/Passes/CallGraph.cpp renamed to bolt/lib/Core/CallGraph.cpp

@@ -1,4 +1,4 @@
-//===- bolt/Passes/CallGraph.cpp ------------------------------------------===//
+//===- bolt/Core/CallGraph.cpp ------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -10,16 +10,16 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "bolt/Passes/CallGraph.h"
+#include "bolt/Core/CallGraph.h"
 
 #define DEBUG_TYPE "callgraph"
 
 #if defined(__x86_64__) && !defined(_MSC_VER)
-#  if (!defined USE_SSECRC)
-#    define USE_SSECRC
-#  endif
+#if (!defined USE_SSECRC)
+#define USE_SSECRC
+#endif
 #else
-#  undef USE_SSECRC
+#undef USE_SSECRC
 #endif
 
 static LLVM_ATTRIBUTE_UNUSED inline size_t hash_int64_fallback(int64_t k) {
@@ -50,7 +50,7 @@ static inline size_t hash_int64_pair(int64_t k1, int64_t k2) {
   // crc32 is commutative, so we need to perturb k1 so that (k1, k2) hashes
   // differently from (k2, k1).
   k1 += k1;
-  __asm("crc32q %1, %0\n" : "+r" (k1) : "rm"(k2));
+  __asm("crc32q %1, %0\n" : "+r"(k1) : "rm"(k2));
   return k1;
 #else
   return (hash_int64(k1) << 1) ^ hash_int64(k2);
@@ -118,5 +118,5 @@ void CallGraph::adjustArcWeights() {
   }
 }
 
-}
-}
+} // namespace bolt
+} // namespace llvm

bolt/lib/Passes/CallGraphWalker.cpp renamed to bolt/lib/Core/CallGraphWalker.cpp

@@ -1,4 +1,4 @@
-//===- bolt/Passes/CallGraphWalker.cpp ------------------------------------===//
+//===- bolt/Core/CallGraphWalker.cpp ------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -10,8 +10,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "bolt/Passes/CallGraphWalker.h"
-#include "bolt/Passes/BinaryFunctionCallGraph.h"
+#include "bolt/Core/CallGraphWalker.h"
+#include "bolt/Core/BinaryFunctionCallGraph.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/Timer.h"
 #include <queue>

bolt/lib/Passes/CMakeLists.txt

@@ -4,11 +4,8 @@ add_llvm_library(LLVMBOLTPasses
   AllocCombiner.cpp
   AsmDump.cpp
   BinaryPasses.cpp
-  BinaryFunctionCallGraph.cpp
   CMOVConversion.cpp
   CacheMetrics.cpp
-  CallGraph.cpp
-  CallGraphWalker.cpp
   DataflowAnalysis.cpp
   DataflowInfoManager.cpp
   FrameAnalysis.cpp

bolt/lib/Passes/FrameAnalysis.cpp

@@ -11,8 +11,8 @@
 //===----------------------------------------------------------------------===//
 
 #include "bolt/Passes/FrameAnalysis.h"
+#include "bolt/Core/CallGraphWalker.h"
 #include "bolt/Core/ParallelUtilities.h"
-#include "bolt/Passes/CallGraphWalker.h"
 #include "llvm/MC/MCRegisterInfo.h"
 #include "llvm/Support/Timer.h"
 #include <fstream>

bolt/lib/Passes/FrameOptimizer.cpp

@@ -11,8 +11,8 @@
 //===----------------------------------------------------------------------===//
 
 #include "bolt/Passes/FrameOptimizer.h"
+#include "bolt/Core/BinaryFunctionCallGraph.h"
 #include "bolt/Core/ParallelUtilities.h"
-#include "bolt/Passes/BinaryFunctionCallGraph.h"
 #include "bolt/Passes/DataflowInfoManager.h"
 #include "bolt/Passes/ShrinkWrapping.h"
 #include "bolt/Passes/StackAvailableExpressions.h"