Merge branch 'main' into xegpu_wg_sg_main

Author: nbpatel

bolt/include/bolt/Core/BinaryContext.h

@@ -1499,7 +1499,7 @@ class BinaryContext {
     MCEInstance.LocalCtx.reset(
         new MCContext(*TheTriple, AsmInfo.get(), MRI.get(), STI.get()));
     MCEInstance.LocalMOFI.reset(
-        TheTarget->createMCObjectFileInfo(*MCEInstance.LocalCtx.get(),
+        TheTarget->createMCObjectFileInfo(*MCEInstance.LocalCtx,
                                           /*PIC=*/!HasFixedLoadAddress));
     MCEInstance.LocalCtx->setObjectFileInfo(MCEInstance.LocalMOFI.get());
     MCEInstance.MCE.reset(

bolt/include/bolt/Core/BinaryFunction.h

@@ -867,15 +867,15 @@ class BinaryFunction {
   /// Returns if BinaryDominatorTree has been constructed for this function.
   bool hasDomTree() const { return BDT != nullptr; }
 
-  BinaryDominatorTree &getDomTree() { return *BDT.get(); }
+  BinaryDominatorTree &getDomTree() { return *BDT; }
 
   /// Constructs DomTree for this function.
   void constructDomTree();
 
   /// Returns if loop detection has been run for this function.
   bool hasLoopInfo() const { return BLI != nullptr; }
 
-  const BinaryLoopInfo &getLoopInfo() { return *BLI.get(); }
+  const BinaryLoopInfo &getLoopInfo() { return *BLI; }
 
   bool isLoopFree() {
     if (!hasLoopInfo())

bolt/include/bolt/Core/DIEBuilder.h

@@ -137,7 +137,7 @@ class DIEBuilder {
   std::unordered_map<std::string, uint32_t> NameToIndexMap;
 
   /// Returns current state of the DIEBuilder
-  State &getState() { return *BuilderState.get(); }
+  State &getState() { return *BuilderState; }
 
   /// Resolve the reference in DIE, if target is not loaded into IR,
   /// pre-allocate it. \p RefCU will be updated to the Unit specific by \p

bolt/include/bolt/Passes/FrameAnalysis.h

@@ -10,6 +10,7 @@
 #define BOLT_PASSES_FRAMEANALYSIS_H
 
 #include "bolt/Passes/StackPointerTracking.h"
+#include <tuple>
 
 namespace llvm {
 namespace bolt {
@@ -53,9 +54,7 @@ struct ArgInStackAccess {
   uint8_t Size;
 
   bool operator<(const ArgInStackAccess &RHS) const {
-    if (StackOffset != RHS.StackOffset)
-      return StackOffset < RHS.StackOffset;
-    return Size < RHS.Size;
+    return std::tie(StackOffset, Size) < std::tie(RHS.StackOffset, RHS.Size);
   }
 };
 

bolt/include/bolt/Passes/PAuthGadgetScanner.h

@@ -43,9 +43,7 @@ struct MCInstInBBReference {
     return BB == RHS.BB && BBIndex == RHS.BBIndex;
   }
   bool operator<(const MCInstInBBReference &RHS) const {
-    if (BB != RHS.BB)
-      return BB < RHS.BB;
-    return BBIndex < RHS.BBIndex;
+    return std::tie(BB, BBIndex) < std::tie(RHS.BB, RHS.BBIndex);
   }
   operator MCInst &() const {
     assert(BB != nullptr);

bolt/include/bolt/Profile/DataReader.h

@@ -90,13 +90,7 @@ struct BranchInfo {
   }
 
   bool operator<(const BranchInfo &RHS) const {
-    if (From < RHS.From)
-      return true;
-
-    if (From == RHS.From)
-      return (To < RHS.To);
-
-    return false;
+    return std::tie(From, To) < std::tie(RHS.From, RHS.To);
   }
 
   /// Merges branch and misprediction counts of \p BI with those of this object.
@@ -252,6 +246,9 @@ struct FuncSampleData {
   /// Get the number of samples recorded in [Start, End)
   uint64_t getSamples(uint64_t Start, uint64_t End) const;
 
+  /// Returns the total number of samples recorded in this function.
+  uint64_t getSamples() const;
+
   /// Aggregation helper
   DenseMap<uint64_t, size_t> Index;
 

bolt/lib/Core/DIEBuilder.cpp

@@ -292,10 +292,10 @@ void DIEBuilder::buildTypeUnits(DebugStrOffsetsWriter *StrOffsetWriter,
 
   getState().Type = ProcessingType::DWARF4TUs;
   for (std::unique_ptr<DWARFUnit> &DU : CU4TURanges)
-    registerUnit(*DU.get(), false);
+    registerUnit(*DU, false);
 
   for (std::unique_ptr<DWARFUnit> &DU : CU4TURanges)
-    constructFromUnit(*DU.get());
+    constructFromUnit(*DU);
 
   DWARFContext::unit_iterator_range CURanges =
       isDWO() ? DwarfContext->dwo_info_section_units()
@@ -308,7 +308,7 @@ void DIEBuilder::buildTypeUnits(DebugStrOffsetsWriter *StrOffsetWriter,
   for (std::unique_ptr<DWARFUnit> &DU : CURanges) {
     if (!DU->isTypeUnit())
       continue;
-    registerUnit(*DU.get(), false);
+    registerUnit(*DU, false);
   }
 
   for (DWARFUnit *DU : getState().DWARF5TUVector) {
@@ -335,7 +335,7 @@ void DIEBuilder::buildCompileUnits(const bool Init) {
   for (std::unique_ptr<DWARFUnit> &DU : CURanges) {
     if (DU->isTypeUnit())
       continue;
-    registerUnit(*DU.get(), false);
+    registerUnit(*DU, false);
   }
 
   // Using DULIst since it can be modified by cross CU refrence resolution.

bolt/lib/Core/DebugData.cpp

@@ -142,7 +142,7 @@ DebugRangesSectionWriter::DebugRangesSectionWriter() {
 
 void DebugRangesSectionWriter::initSection() {
   // Adds an empty range to the buffer.
-  writeAddressRanges(*RangesStream.get(), DebugAddressRangesVector{});
+  writeAddressRanges(*RangesStream, DebugAddressRangesVector{});
 }
 
 uint64_t DebugRangesSectionWriter::addRanges(
@@ -169,7 +169,7 @@ uint64_t DebugRangesSectionWriter::addRanges(DebugAddressRangesVector &Ranges) {
   // unique and correct offsets in patches.
   std::lock_guard<std::mutex> Lock(WriterMutex);
   const uint32_t EntryOffset = RangesBuffer->size();
-  writeAddressRanges(*RangesStream.get(), Ranges);
+  writeAddressRanges(*RangesStream, Ranges);
 
   return EntryOffset;
 }
@@ -321,8 +321,8 @@ void DebugRangeListsSectionWriter::finalizeSection() {
                            llvm::endianness::little);
 
   std::unique_ptr<DebugBufferVector> Header = getDWARF5Header(
-      {static_cast<uint32_t>(SizeOfArraySection + CUBodyBuffer.get()->size()),
-       5, 8, 0, static_cast<uint32_t>(RangeEntries.size())});
+      {static_cast<uint32_t>(SizeOfArraySection + CUBodyBuffer->size()), 5, 8,
+       0, static_cast<uint32_t>(RangeEntries.size())});
   *RangesStream << *Header;
   *RangesStream << *CUArrayBuffer;
   *RangesStream << *CUBodyBuffer;
@@ -747,8 +747,8 @@ void DebugLoclistWriter::finalizeDWARF5(DIEBuilder &DIEBldr, DIE &Die) {
         llvm::endianness::little);
 
   std::unique_ptr<DebugBufferVector> Header = getDWARF5Header(
-      {static_cast<uint32_t>(SizeOfArraySection + LocBodyBuffer.get()->size()),
-       5, 8, 0, NumberOfEntries});
+      {static_cast<uint32_t>(SizeOfArraySection + LocBodyBuffer->size()), 5, 8,
+       0, NumberOfEntries});
   *LocStream << *Header;
   *LocStream << *LocArrayBuffer;
   *LocStream << *LocBodyBuffer;

bolt/lib/Core/DebugNames.cpp

@@ -648,8 +648,8 @@ void DWARF5AcceleratorTable::writeEntries() {
         if (const auto Iter = EntryRelativeOffsets.find(*ParentOffset);
             Iter != EntryRelativeOffsets.end()) {
           const uint64_t PatchOffset = Entry->getPatchOffset();
-          uint32_t *Ptr = reinterpret_cast<uint32_t *>(
-              &EntriesBuffer.get()->data()[PatchOffset]);
+          uint32_t *Ptr =
+              reinterpret_cast<uint32_t *>(&EntriesBuffer->data()[PatchOffset]);
           *Ptr = Iter->second;
         } else {
           BC.errs() << "BOLT-WARNING: [internal-dwarf-warning]: Could not find "

bolt/lib/Core/ParallelUtilities.cpp

@@ -103,7 +103,7 @@ inline unsigned estimateTotalCost(const BinaryContext &BC,
 } // namespace
 
 ThreadPoolInterface &getThreadPool(const unsigned ThreadsCount) {
-  if (ThreadPoolPtr.get())
+  if (ThreadPoolPtr)
     return *ThreadPoolPtr;
 
   if (ThreadsCount > 1)