AArch64TargetMachine.cpp

Go to the documentation of this file.

//===-- AArch64TargetMachine.cpp - Define TargetMachine for AArch64 -------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
//
//===----------------------------------------------------------------------===//
 
#include "AArch64TargetMachine.h"
#include "AArch64.h"
#include "AArch64AsmPrinter.h"
#include "AArch64MachineFunctionInfo.h"
#include "AArch64MachineScheduler.h"
#include "AArch64MacroFusion.h"
#include "AArch64Subtarget.h"
#include "AArch64TargetObjectFile.h"
#include "AArch64TargetTransformInfo.h"
#include "MCTargetDesc/AArch64MCTargetDesc.h"
#include "TargetInfo/AArch64TargetInfo.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/CodeGen/CSEConfigBase.h"
#include "llvm/CodeGen/GlobalISel/CSEInfo.h"
#include "llvm/CodeGen/GlobalISel/IRTranslator.h"
#include "llvm/CodeGen/GlobalISel/InstructionSelect.h"
#include "llvm/CodeGen/GlobalISel/Legalizer.h"
#include "llvm/CodeGen/GlobalISel/LoadStoreOpt.h"
#include "llvm/CodeGen/GlobalISel/Localizer.h"
#include "llvm/CodeGen/GlobalISel/RegBankSelect.h"
#include "llvm/CodeGen/MIRParser/MIParser.h"
#include "llvm/CodeGen/MachineScheduler.h"
#include "llvm/CodeGen/Passes.h"
#include "llvm/CodeGen/TargetInstrInfo.h"
#include "llvm/CodeGen/TargetPassConfig.h"
#include "llvm/IR/Attributes.h"
#include "llvm/IR/Function.h"
#include "llvm/InitializePasses.h"
#include "llvm/MC/MCAsmInfo.h"
#include "llvm/MC/MCTargetOptions.h"
#include "llvm/MC/TargetRegistry.h"
#include "llvm/Option/LibraryOptions.h"
#include "llvm/Pass.h"
#include "llvm/Passes/PassBuilder.h"
#include "llvm/Support/CodeGen.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Target/TargetLoweringObjectFile.h"
#include "llvm/Target/TargetOptions.h"
#include "llvm/TargetParser/Triple.h"
#include "llvm/Transforms/CFGuard.h"
#include "llvm/Transforms/Scalar.h"
#include "llvm/Transforms/Utils/LowerIFunc.h"
#include "llvm/Transforms/Vectorize/LoopIdiomVectorize.h"
#include <memory>
 
using namespace llvm;
 
extern "C" LLVM_ABI LLVM_EXTERNAL_VISIBILITY void
LLVMInitializeAArch64Target() {
  // Register the target.
  RegisterTargetMachine<AArch64leTargetMachine> X(getTheAArch64leTarget());
  RegisterTargetMachine<AArch64beTargetMachine> Y(getTheAArch64beTarget());
  RegisterTargetMachine<AArch64leTargetMachine> Z(getTheARM64Target());
  RegisterTargetMachine<AArch64leTargetMachine> W(getTheARM64_32Target());
  RegisterTargetMachine<AArch64leTargetMachine> V(getTheAArch64_32Target());
  static opt::RegisterLibraryOptions<AArch64Options> O;
  auto &PR = *PassRegistry::getPassRegistry();
  initializeGlobalISel(PR);
  initializeAArch64A53Fix835769LegacyPass(PR);
  initializeAArch64A57FPLoadBalancingLegacyPass(PR);
  initializeAArch64CodeLayoutOptPass(PR);
  initializeAArch64AdvSIMDScalarLegacyPass(PR);
  initializeAArch64AsmPrinterPass(PR);
  initializeAArch64BranchTargetsLegacyPass(PR);
  initializeAArch64CollectLOHLegacyPass(PR);
  initializeAArch64CompressJumpTablesLegacyPass(PR);
  initializeAArch64ConditionalComparesLegacyPass(PR);
  initializeAArch64ConditionOptimizerLegacyPass(PR);
  initializeAArch64DeadRegisterDefinitionsLegacyPass(PR);
  initializeAArch64ExpandPseudoLegacyPass(PR);
  initializeAArch64LoadStoreOptLegacyPass(PR);
  initializeAArch64MIPeepholeOptLegacyPass(PR);
  initializeAArch64PTrueCoalescingLegacyPass(PR);
  initializeAArch64SIMDInstrOptLegacyPass(PR);
  initializeAArch64O0PreLegalizerCombinerLegacyPass(PR);
  initializeAArch64PredicateAsCounterLoopRewritesPass(PR);
  initializeAArch64PreLegalizerCombinerLegacyPass(PR);
  initializeAArch64PointerAuthLegacyPass(PR);
  initializeAArch64PostCoalescerLegacyPass(PR);
  initializeAArch64PostLegalizerCombinerLegacyPass(PR);
  initializeAArch64PostSelectOptimizeLegacyPass(PR);
  initializeAArch64PostLegalizerLoweringLegacyPass(PR);
  initializeAArch64PromoteConstantPass(PR);
  initializeAArch64RedundantCopyEliminationLegacyPass(PR);
  initializeAArch64RedundantCondBranchLegacyPass(PR);
  initializeAArch64StorePairSuppressPass(PR);
  initializeFalkorHWPFFixPass(PR);
  initializeFalkorMarkStridedAccessesLegacyPass(PR);
  initializeLDTLSCleanupPass(PR);
  initializeMachineKCFILegacyPass(PR);
  initializeMachineSMEABIPass(PR);
  initializeAArch64SRLTDefineSuperRegsLegacyPass(PR);
  initializeSMEPeepholeOptPass(PR);
  initializeAArch64SpeculationHardeningPass(PR);
  initializeAArch64SLSHardeningLegacyPass(PR);
  initializeAArch64StackTaggingPass(PR);
  initializeAArch64StackTaggingPreRALegacyPass(PR);
  initializeAArch64LowerHomogeneousPrologEpilogLegacyPass(PR);
  initializeAArch64DAGToDAGISelLegacyPass(PR);
  initializeAArch64CondBrTuningPass(PR);
  initializeAArch64Arm64ECCallLoweringPass(PR);
  initializeSVEShuffleOptsPass(PR);
}
 
bool AArch64TargetMachine::isGlobalISelOptNone() const {
  const bool GlobalISelFlag = getCGPassBuilderOption().EnableGlobalISelOption ==
                              cl::boolOrDefault::BOU_TRUE;
 
  return getOptLevel() == CodeGenOptLevel::None ||
         (static_cast<unsigned>(getOptLevel()) >
              static_cast<unsigned>(CLOpts.enable_global_isel_at_O) &&
          !GlobalISelFlag);
}
 
void AArch64TargetMachine::reset() {
  SubtargetMap.clear();
  LastSubtarget = nullptr;
}
 
//===----------------------------------------------------------------------===//
// AArch64 Lowering public interface.
//===----------------------------------------------------------------------===//
static std::unique_ptr<TargetLoweringObjectFile> createTLOF(const Triple &TT) {
  if (TT.isOSBinFormatMachO())
    return std::make_unique<AArch64_MachoTargetObjectFile>();
  if (TT.isOSBinFormatCOFF())
    return std::make_unique<AArch64_COFFTargetObjectFile>();
 
  return std::make_unique<AArch64_ELFTargetObjectFile>();
}
 
static StringRef computeDefaultCPU(const Triple &TT, StringRef CPU) {
  if (CPU.empty() && TT.isArm64e())
    return "apple-a12";
  return CPU;
}
 
static Reloc::Model getEffectiveRelocModel(const Triple &TT,
                                           std::optional<Reloc::Model> RM) {
  // AArch64 Darwin and Windows are always PIC.
  if (TT.isOSDarwin() || TT.isOSWindows())
    return Reloc::PIC_;
  // On ELF platforms the default static relocation model has a smart enough
  // linker to cope with referencing external symbols defined in a shared
  // library. Hence DynamicNoPIC doesn't need to be promoted to PIC.
  if (!RM || *RM == Reloc::DynamicNoPIC)
    return Reloc::Static;
  return *RM;
}
 
static CodeModel::Model
getEffectiveAArch64CodeModel(const Triple &TT,
                             std::optional<CodeModel::Model> CM, bool JIT) {
  if (CM) {
    if (*CM != CodeModel::Small && *CM != CodeModel::Tiny &&
        *CM != CodeModel::Large) {
      report_fatal_error(
          "Only small, tiny and large code models are allowed on AArch64");
    } else if (*CM == CodeModel::Tiny && !TT.isOSBinFormatELF()) {
      report_fatal_error("tiny code model is only supported on ELF");
    }
    return *CM;
  }
  // The default MCJIT memory managers make no guarantees about where they can
  // find an executable page; JITed code needs to be able to refer to globals
  // no matter how far away they are.
  // We should set the CodeModel::Small for Windows ARM64 in JIT mode,
  // since with large code model LLVM generating 4 MOV instructions, and
  // Windows doesn't support relocating these long branch (4 MOVs).
  if (JIT && !TT.isOSWindows())
    return CodeModel::Large;
  return CodeModel::Small;
}
 
/// Create an AArch64 architecture model.
///
AArch64TargetMachine::AArch64TargetMachine(const Target &T, const Triple &TT,
                                           StringRef CPU, StringRef FS,
                                           const TargetOptions &Options,
                                           std::optional<Reloc::Model> RM,
                                           std::optional<CodeModel::Model> CM,
                                           CodeGenOptLevel OL, bool JIT,
                                           bool LittleEndian)
    : CodeGenTargetMachineImpl(T, TT, computeDefaultCPU(TT, CPU), FS, Options,
                               getEffectiveRelocModel(TT, RM),
                               getEffectiveAArch64CodeModel(TT, CM, JIT), OL),
      CLOpts(AArch64Options::Global), TLOF(createTLOF(getTargetTriple())),
      isLittle(LittleEndian) {
  initAsmInfo();
 
  if (TT.isOSBinFormatMachO()) {
    this->Options.TrapUnreachable = true;
    this->Options.NoTrapAfterNoreturn = true;
  }
 
  if (getMCAsmInfo().getExceptionHandlingType() == ExceptionHandling::WinEH &&
      getMCAsmInfo().usesWindowsCFI()) {
    // Unwinding can get confused if the last instruction in an
    // exception-handling region (function, funclet, try block, etc.)
    // is a call.
    //
    // FIXME: We could elide the trap if the next instruction would be in
    // the same region anyway.
    this->Options.TrapUnreachable = true;
  }
 
  if (this->Options.TLSSize == 0) // default
    this->Options.TLSSize = 24;
  if ((getCodeModel() == CodeModel::Small ||
       getCodeModel() == CodeModel::Kernel) &&
      this->Options.TLSSize > 32)
    // for the small (and kernel) code model, the maximum TLS size is 4GiB
    this->Options.TLSSize = 32;
  else if (getCodeModel() == CodeModel::Tiny && this->Options.TLSSize > 24)
    // for the tiny code model, the maximum TLS size is 1MiB (< 16MiB)
    this->Options.TLSSize = 24;
 
  const bool TargetSupportsGISel =
      TT.getArch() != Triple::aarch64_32 &&
      TT.getEnvironment() != Triple::GNUILP32 &&
      !(getCodeModel() == CodeModel::Large && TT.isOSBinFormatMachO());
 
  const bool GlobalISelFlag = getCGPassBuilderOption().EnableGlobalISelOption ==
                              cl::boolOrDefault::BOU_TRUE;
 
  // Enable GlobalISel at or below EnableGlobalISelAt0, unless this is
  // MachO/CodeModel::Large, which GlobalISel does not support.
  if (TargetSupportsGISel && CLOpts.enable_global_isel_at_O != -1 &&
      (static_cast<int>(getOptLevel()) <= CLOpts.enable_global_isel_at_O ||
       (!GlobalISelFlag && !Options.EnableGlobalISel))) {
    setGlobalISel(true);
    setGlobalISelAbort(GlobalISelAbortMode::Disable);
  }
 
  LLT::setUseExtended(true);
 
  // AArch64 supports the MachineOutliner.
  setMachineOutliner(true);
 
  // AArch64 supports default outlining behaviour.
  setSupportsDefaultOutlining(true);
 
  // AArch64 supports the debug entry values.
  setSupportsDebugEntryValues(true);
 
  // AArch64 supports fixing up the DWARF unwind information.
  // enableCFIFixup() checks per function for DWARF unwinding.
  setCFIFixup(true);
}
 
unsigned AArch64TargetMachine::getEnableGlobalISelAtO() const {
  return CLOpts.enable_global_isel_at_O;
}
 
AArch64TargetMachine::~AArch64TargetMachine() = default;
 
const AArch64Subtarget *
AArch64TargetMachine::getSubtargetImpl(const Function &F) const {
  // Constructing the subtarget key is not cheap, avoid rebuilding it for
  // repeated queries with the same function attributes.
  AttributeSet FnAttrs = F.getAttributes().getFnAttrs();
  if (LastSubtarget && LastSubtargetAttrs == FnAttrs)
    return LastSubtarget;
 
  Attribute CPUAttr = F.getFnAttribute("target-cpu");
  Attribute TuneAttr = F.getFnAttribute("tune-cpu");
  Attribute FSAttr = F.getFnAttribute("target-features");
 
  StringRef CPU = CPUAttr.isValid() ? CPUAttr.getValueAsString() : TargetCPU;
  StringRef TuneCPU = TuneAttr.isValid() ? TuneAttr.getValueAsString() : CPU;
  StringRef FS = FSAttr.isValid() ? FSAttr.getValueAsString() : TargetFS;
  bool HasMinSize = F.hasMinSize();
 
  bool IsStreaming = CLOpts.force_streaming ||
                     F.hasFnAttribute("aarch64_pstate_sm_enabled") ||
                     F.hasFnAttribute("aarch64_pstate_sm_body");
  bool IsStreamingCompatible = CLOpts.force_streaming_compatible ||
                               F.hasFnAttribute("aarch64_pstate_sm_compatible");
 
  unsigned MinSVEVectorSize = 0;
  unsigned MaxSVEVectorSize = 0;
  if (F.hasFnAttribute(Attribute::VScaleRange)) {
    ConstantRange CR = getVScaleRange(&F, 64);
    MinSVEVectorSize = CR.getUnsignedMin().getZExtValue() * 128;
    MaxSVEVectorSize = CR.getUnsignedMax().getZExtValue() * 128;
  } else {
    MinSVEVectorSize = CLOpts.sve_vector_bits_min;
    MaxSVEVectorSize = CLOpts.sve_vector_bits_max;
  }
 
  assert(MinSVEVectorSize % 128 == 0 &&
         "SVE requires vector length in multiples of 128!");
  assert(MaxSVEVectorSize % 128 == 0 &&
         "SVE requires vector length in multiples of 128!");
  assert((MaxSVEVectorSize >= MinSVEVectorSize || MaxSVEVectorSize == 0) &&
         "Minimum SVE vector size should not be larger than its maximum!");
 
  // Sanitize user input in case of no asserts
  if (MaxSVEVectorSize != 0) {
    MinSVEVectorSize = std::min(MinSVEVectorSize, MaxSVEVectorSize);
    MaxSVEVectorSize = std::max(MinSVEVectorSize, MaxSVEVectorSize);
  }
 
  SmallString<512> Key;
  // This lookup is hot during repeated TTI queries, so build the key directly
  // instead of formatting through raw_svector_ostream.
  Key += "SVEMin";
  Key += utostr(MinSVEVectorSize);
  Key += "SVEMax";
  Key += utostr(MaxSVEVectorSize);
  Key += "IsStreaming=";
  Key += utostr(IsStreaming);
  Key += "IsStreamingCompatible=";
  Key += utostr(IsStreamingCompatible);
  Key += CPU;
  Key += TuneCPU;
  Key += FS;
  Key += "HasMinSize=";
  Key += utostr(HasMinSize);
 
  auto &I = SubtargetMap[Key];
  if (!I) {
    I = std::make_unique<AArch64Subtarget>(
        TargetTriple, CPU, TuneCPU, FS, *this, isLittle, MinSVEVectorSize,
        MaxSVEVectorSize, IsStreaming, IsStreamingCompatible, HasMinSize,
        CLOpts.srlt_mitigate_sr2r);
  }
 
  if (IsStreaming && !I->hasSME())
    reportFatalUsageError("streaming SVE functions require SME");
 
  LastSubtargetAttrs = FnAttrs;
  LastSubtarget = I.get();
  return LastSubtarget;
}
 
// Encourage placing FORM_TRANSPOSED_REG immediately before the instruction that
// uses/consumes it. This ensures its def has a short live range, which means
// we're more likely to allocate registers its operands first (which works best
// for the hints in AArch64RegisterInfo::getRegAllocationHints).
static bool scheduleFormTransposedTupleAdjacentToUsers(
    const TargetInstrInfo &TII, const TargetSubtargetInfo &TSI,
    const MachineInstr *FirstMI, const MachineInstr &SecondMI,
    const SDep *Dep) {
  if (isNonDataDep(Dep))
    return false;
  return !FirstMI ||
         FirstMI->getOpcode() == AArch64::FORM_TRANSPOSED_REG_TUPLE_X2_PSEUDO ||
         FirstMI->getOpcode() == AArch64::FORM_TRANSPOSED_REG_TUPLE_X4_PSEUDO;
}
 
ScheduleDAGInstrs *
AArch64TargetMachine::createMachineScheduler(MachineSchedContext *C) const {
  const AArch64Subtarget &ST = C->MF->getSubtarget<AArch64Subtarget>();
  ScheduleDAGMILive *DAG = createSchedLive(C);
  DAG->addMutation(createLoadClusterDAGMutation(DAG->TII));
  DAG->addMutation(createStoreClusterDAGMutation(DAG->TII));
  if (ST.hasFusion())
    DAG->addMutation(createAArch64MacroFusionDAGMutation());
  if (ST.hasSME() && ST.isStreaming())
    DAG->addMutation(createMacroFusionDAGMutation(
        scheduleFormTransposedTupleAdjacentToUsers));
  return DAG;
}
 
ScheduleDAGInstrs *
AArch64TargetMachine::createPostMachineScheduler(MachineSchedContext *C) const {
  const AArch64Subtarget &ST = C->MF->getSubtarget<AArch64Subtarget>();
  ScheduleDAGMI *DAG = createSchedPostRA<AArch64PostRASchedStrategy>(C);
  if (ST.hasFusion()) {
    // Run the Macro Fusion after RA again since literals are expanded from
    // pseudos then (v. addPreSched2()).
    DAG->addMutation(createAArch64MacroFusionDAGMutation());
    return DAG;
  }
 
  return DAG;
}
 
size_t AArch64TargetMachine::clearLinkerOptimizationHints(
    const SmallPtrSetImpl<MachineInstr *> &MIs) const {
  if (MIs.empty())
    return 0;
  auto *MI = *MIs.begin();
  auto *FuncInfo = MI->getMF()->getInfo<AArch64FunctionInfo>();
  return FuncInfo->clearLinkerOptimizationHints(MIs);
}
 
void AArch64leTargetMachine::anchor() { }
 
AArch64leTargetMachine::AArch64leTargetMachine(
    const Target &T, const Triple &TT, StringRef CPU, StringRef FS,
    const TargetOptions &Options, std::optional<Reloc::Model> RM,
    std::optional<CodeModel::Model> CM, CodeGenOptLevel OL, bool JIT)
    : AArch64TargetMachine(T, TT, CPU, FS, Options, RM, CM, OL, JIT, true) {}
 
void AArch64beTargetMachine::anchor() { }
 
AArch64beTargetMachine::AArch64beTargetMachine(
    const Target &T, const Triple &TT, StringRef CPU, StringRef FS,
    const TargetOptions &Options, std::optional<Reloc::Model> RM,
    std::optional<CodeModel::Model> CM, CodeGenOptLevel OL, bool JIT)
    : AArch64TargetMachine(T, TT, CPU, FS, Options, RM, CM, OL, JIT, false) {}
 
namespace {
 
/// AArch64 Code Generator Pass Configuration Options.
class AArch64PassConfig : public TargetPassConfig {
  const AArch64Options &CLOpts;
 
public:
  AArch64PassConfig(AArch64TargetMachine &TM, PassManagerBase &PM)
      : TargetPassConfig(TM, PM), CLOpts(TM.getCLOpts()) {
    if (TM.getOptLevel() != CodeGenOptLevel::None)
      substitutePass(&PostRASchedulerID, &PostMachineSchedulerID);
    setEnableSinkAndFold(CLOpts.enable_sink_fold);
  }
 
  AArch64TargetMachine &getAArch64TargetMachine() const {
    return getTM<AArch64TargetMachine>();
  }
 
  void addIRPasses()  override;
  bool addPreISel() override;
  void addCodeGenPrepare() override;
  bool addInstSelector() override;
  bool addIRTranslator() override;
  void addPreLegalizeMachineIR() override;
  bool addLegalizeMachineIR() override;
  void addPreRegBankSelect() override;
  bool addRegBankSelect() override;
  bool addGlobalInstructionSelect() override;
  void addMachineSSAOptimization() override;
  bool addILPOpts() override;
  void addPreRegAlloc() override;
  void addPostRewrite() override;
  void addPostRegAlloc() override;
  void addPreSched2() override;
  void addPreEmitPass() override;
  void addPostBBSections() override;
  void addPreEmitPass2() override;
  bool addRegAssignAndRewriteOptimized() override;
 
  std::unique_ptr<CSEConfigBase> getCSEConfig() const override;
};
 
} // end anonymous namespace
 
void AArch64TargetMachine::registerPassBuilderCallbacks(PassBuilder &PB) {
#define GET_PASS_REGISTRY "AArch64PassRegistry.def"
#include "llvm/Passes/TargetPassRegistry.inc"
 
  PB.registerLateLoopOptimizationsEPCallback(
      [=](LoopPassManager &LPM, OptimizationLevel Level) {
        if (Level != OptimizationLevel::O0)
          LPM.addPass(LoopIdiomVectorizePass());
      });
  if (getTargetTriple().isOSWindows())
    PB.registerPipelineEarlySimplificationEPCallback(
        [](ModulePassManager &PM, OptimizationLevel, ThinOrFullLTOPhase) {
          PM.addPass(LowerIFuncPass());
        });
}
 
TargetTransformInfo
AArch64TargetMachine::getTargetTransformInfo(const Function &F) const {
  return TargetTransformInfo(std::make_unique<AArch64TTIImpl>(this, F));
}
 
TargetPassConfig *AArch64TargetMachine::createPassConfig(PassManagerBase &PM) {
  return new AArch64PassConfig(*this, PM);
}
 
std::unique_ptr<CSEConfigBase> AArch64PassConfig::getCSEConfig() const {
  return getStandardCSEConfigForOpt(TM->getOptLevel());
}
 
void AArch64PassConfig::addIRPasses() {
  // Always expand atomic operations, we don't deal with atomicrmw or cmpxchg
  // ourselves.
  addPass(createAtomicExpandLegacyPass());
 
  if (getOptLevel() >= CodeGenOptLevel::Default &&
      CLOpts.enable_predicate_as_counter_loop_rewrites)
    addPass(createAArch64PredicateAsCounterLoopRewritesPass());
 
  // Cmpxchg instructions are often used with a subsequent comparison to
  // determine whether it succeeded. We can exploit existing control-flow in
  // ldrex/strex loops to simplify this, but it needs tidying up.
  if (TM->getOptLevel() != CodeGenOptLevel::None &&
      CLOpts.enable_atomic_cfg_tidy)
    addPass(createCFGSimplificationPass(SimplifyCFGOptions()
                                            .forwardSwitchCondToPhi(true)
                                            .convertSwitchRangeToICmp(true)
                                            .convertSwitchToLookupTable(true)
                                            .needCanonicalLoops(false)
                                            .hoistCommonInsts(true)
                                            .sinkCommonInsts(true)));
 
  // Run LoopDataPrefetch
  //
  // Run this before LSR to remove the multiplies involved in computing the
  // pointer values N iterations ahead.
  if (TM->getOptLevel() != CodeGenOptLevel::None) {
    if (CLOpts.enable_loop_data_prefetch)
      addPass(createLoopDataPrefetchPass());
    if (CLOpts.enable_falkor_hwpf_fix)
      addPass(createFalkorMarkStridedAccessesPass());
  }
 
  if (CLOpts.enable_gep_opt) {
    // Call SeparateConstOffsetFromGEP pass to extract constants within indices
    // and lower a GEP with multiple indices to either arithmetic operations or
    // multiple GEPs with single index.
    addPass(createSeparateConstOffsetFromGEPPass(true));
    // Call EarlyCSE pass to find and remove subexpressions in the lowered
    // result.
    addPass(createEarlyCSEPass());
    // Do loop invariant code motion in case part of the lowered result is
    // invariant.
    addPass(createLICMPass());
  }
 
  TargetPassConfig::addIRPasses();
 
  if (getOptLevel() == CodeGenOptLevel::Aggressive && CLOpts.select_opt)
    addPass(createSelectOptimizePass());
 
  addPass(createAArch64StackTaggingPass(
      /*IsOptNone=*/TM->getOptLevel() == CodeGenOptLevel::None));
 
  // Try to use tbl in place of other shuffling operations if doing so would
  // reduce the total number of instructions. Shuffle masks for big endian may
  // be different, so require a little endian target.
  if (getOptLevel() >= CodeGenOptLevel::Default &&
      CLOpts.enable_sve_shuffle_opts && TM->getTargetTriple().isLittleEndian())
    addPass(createSVEShuffleOptsPass());
 
  // Match complex arithmetic patterns
  if (TM->getOptLevel() >= CodeGenOptLevel::Default)
    addPass(createComplexDeinterleavingPass(TM));
 
  // Match interleaved memory accesses to ldN/stN intrinsics.
  if (TM->getOptLevel() != CodeGenOptLevel::None) {
    addPass(createInterleavedLoadCombinePass());
    addPass(createInterleavedAccessPass());
  }
 
  // Add Control Flow Guard checks.
  if (TM->getTargetTriple().isOSWindows()) {
    if (TM->getTargetTriple().isWindowsArm64EC())
      addPass(createAArch64Arm64ECCallLoweringPass());
    else
      addPass(createCFGuardPass());
  }
 
  if (TM->Options.JMCInstrument)
    addPass(createJMCInstrumenterPass());
}
 
// Pass Pipeline Configuration
bool AArch64PassConfig::addPreISel() {
  // Run promote constant before global merge, so that the promoted constants
  // get a chance to be merged
  if (TM->getOptLevel() != CodeGenOptLevel::None && CLOpts.enable_promote_const)
    addPass(createAArch64PromoteConstantPass());
  // FIXME: On AArch64, this depends on the type.
  // Basically, the addressable offsets are up to 4095 * Ty.getSizeInBytes().
  // and the offset has to be a multiple of the related size in bytes.
  if (valueOr(CLOpts.enable_global_merge,
              TM->getOptLevel() != CodeGenOptLevel::None)) {
    bool OnlyOptimizeForSize =
        (TM->getOptLevel() < CodeGenOptLevel::Aggressive) &&
        (CLOpts.enable_global_merge == BoolOrDefault::Default);
 
    // Merging of extern globals is enabled by default on non-Mach-O as we
    // expect it to be generally either beneficial or harmless. On Mach-O it
    // is disabled as we emit the .subsections_via_symbols directive which
    // means that merging extern globals is not safe.
    bool MergeExternalByDefault = !TM->getTargetTriple().isOSBinFormatMachO();
    addPass(createGlobalMergePass(TM, 4095, OnlyOptimizeForSize,
                                  MergeExternalByDefault));
  }
 
  return false;
}
 
void AArch64PassConfig::addCodeGenPrepare() {
  if (getOptLevel() != CodeGenOptLevel::None)
    addPass(createTypePromotionLegacyPass());
  TargetPassConfig::addCodeGenPrepare();
}
 
bool AArch64PassConfig::addInstSelector() {
  addPass(createAArch64ISelDag(getAArch64TargetMachine(), getOptLevel()));
  return false;
}
 
bool AArch64PassConfig::addIRTranslator() {
  addPass(new IRTranslatorLegacy(getOptLevel()));
  return false;
}
 
void AArch64PassConfig::addPreLegalizeMachineIR() {
  if (getAArch64TargetMachine().isGlobalISelOptNone()) {
    addPass(createAArch64O0PreLegalizerCombiner());
    addPass(new LocalizerLegacy());
  } else {
    addPass(createAArch64PreLegalizerCombiner());
    addPass(new LocalizerLegacy());
    if (CLOpts.enable_gisel_ldst_prelegal)
      addPass(new LoadStoreOptLegacy());
  }
}
 
bool AArch64PassConfig::addLegalizeMachineIR() {
  addPass(new LegalizerLegacy());
  return false;
}
 
void AArch64PassConfig::addPreRegBankSelect() {
  const bool IsGlobalISelOptNone =
      getAArch64TargetMachine().isGlobalISelOptNone();
  if (!IsGlobalISelOptNone) {
    addPass(createAArch64PostLegalizerCombinerLegacy(IsGlobalISelOptNone));
    if (CLOpts.enable_gisel_ldst_postlegal)
      addPass(new LoadStoreOptLegacy());
  }
  addPass(createAArch64PostLegalizerLowering());
}
 
bool AArch64PassConfig::addRegBankSelect() {
  addPass(new RegBankSelectLegacy());
  return false;
}
 
bool AArch64PassConfig::addGlobalInstructionSelect() {
  addPass(new InstructionSelectLegacy(getOptLevel()));
  if (!getAArch64TargetMachine().isGlobalISelOptNone())
    addPass(createAArch64PostSelectOptimize());
 
  return false;
}
 
void AArch64PassConfig::addMachineSSAOptimization() {
  // For ELF, cleanup any local-dynamic TLS accesses
  // (i.e. combine as many references to _TLS_MODULE_BASE_ as possible.
  if (TM->getTargetTriple().isOSBinFormatELF() &&
      getOptLevel() != CodeGenOptLevel::None)
    addPass(createAArch64CleanupLocalDynamicTLSPass());
 
  if (TM->getOptLevel() != CodeGenOptLevel::None)
    addPass(createMachineSMEABIPass(TM->getOptLevel()));
 
  if (TM->getOptLevel() != CodeGenOptLevel::None &&
      CLOpts.enable_aarch64_sme_peephole_opt)
    addPass(createSMEPeepholeOptPass());
 
  // Run default MachineSSAOptimization first.
  TargetPassConfig::addMachineSSAOptimization();
 
  if (TM->getOptLevel() != CodeGenOptLevel::None) {
    addPass(createAArch64MIPeepholeOptLegacyPass());
    addPass(createAArch64PTrueCoalescingLegacyPass());
  }
}
 
bool AArch64PassConfig::addILPOpts() {
  if (CLOpts.enable_condopt)
    addPass(createAArch64ConditionOptimizerLegacyPass());
  if (CLOpts.enable_ccmp)
    addPass(createAArch64ConditionalCompares());
  if (CLOpts.enable_mcr)
    addPass(&MachineCombinerID);
  if (CLOpts.enable_cond_br_tune)
    addPass(createAArch64CondBrTuning());
  if (CLOpts.enable_early_ifcvt)
    addPass(&EarlyIfConverterLegacyID);
  if (CLOpts.enable_stp_suppress)
    addPass(createAArch64StorePairSuppressPass());
  addPass(createAArch64SIMDInstrOptPass());
  if (TM->getOptLevel() != CodeGenOptLevel::None)
    addPass(createAArch64StackTaggingPreRALegacyPass());
  return true;
}
 
void AArch64PassConfig::addPreRegAlloc() {
  if (TM->getOptLevel() == CodeGenOptLevel::None)
    addPass(createMachineSMEABIPass(CodeGenOptLevel::None));
 
  // Change dead register definitions to refer to the zero register.
  if (TM->getOptLevel() != CodeGenOptLevel::None && CLOpts.enable_dead_defs)
    addPass(createAArch64DeadRegisterDefinitions());
 
  // Use AdvSIMD scalar instructions whenever profitable.
  if (TM->getOptLevel() != CodeGenOptLevel::None && CLOpts.enable_simd_scalar) {
    addPass(createAArch64AdvSIMDScalar());
    // The AdvSIMD pass may produce copies that can be rewritten to
    // be register coalescer friendly.
    addPass(&PeepholeOptimizerLegacyID);
  }
  if (TM->getOptLevel() != CodeGenOptLevel::None && CLOpts.enable_pipeliner)
    addPass(&MachinePipelinerID);
}
 
void AArch64PassConfig::addPostRewrite() {
  if (CLOpts.srlt_mitigate_sr2r)
    addPass(createAArch64SRLTDefineSuperRegsLegacyPass());
}
 
void AArch64PassConfig::addPostRegAlloc() {
  // Remove redundant copy instructions.
  if (TM->getOptLevel() != CodeGenOptLevel::None && CLOpts.enable_copyelim)
    addPass(createAArch64RedundantCopyEliminationPass());
 
  if (TM->getOptLevel() != CodeGenOptLevel::None && usingDefaultRegAlloc())
    // Improve performance for some FP/SIMD code for A57.
    addPass(createAArch64A57FPLoadBalancingLegacyPass());
}
 
void AArch64PassConfig::addPreSched2() {
  // Lower homogeneous frame instructions
  if (CLOpts.homogeneous_prolog_epilog)
    addPass(createAArch64LowerHomogeneousPrologEpilogPass());
  // Expand some pseudo instructions to allow proper scheduling.
  addPass(createAArch64ExpandPseudoLegacyPass());
  // Use load/store pair instructions when possible.
  if (TM->getOptLevel() != CodeGenOptLevel::None) {
    if (CLOpts.enable_ldst_opt)
      addPass(createAArch64LoadStoreOptLegacyPass());
  }
  // Emit KCFI checks for indirect calls.
  addPass(createKCFIPass());
 
  // The AArch64SpeculationHardeningPass destroys dominator tree and natural
  // loop info, which is needed for the FalkorHWPFFixPass and also later on.
  // Therefore, run the AArch64SpeculationHardeningPass before the
  // FalkorHWPFFixPass to avoid recomputing dominator tree and natural loop
  // info.
  addPass(createAArch64SpeculationHardeningPass());
 
  if (TM->getOptLevel() != CodeGenOptLevel::None) {
    if (CLOpts.enable_falkor_hwpf_fix)
      addPass(createFalkorHWPFFixPass());
  }
}
 
void AArch64PassConfig::addPreEmitPass() {
  // Machine Block Placement might have created new opportunities when run
  // at O3, where the Tail Duplication Threshold is set to 4 instructions.
  // Run the load/store optimizer once more.
  if (TM->getOptLevel() >= CodeGenOptLevel::Aggressive &&
      CLOpts.enable_ldst_opt)
    addPass(createAArch64LoadStoreOptLegacyPass());
 
  if (TM->getOptLevel() >= CodeGenOptLevel::Aggressive &&
      CLOpts.enable_copy_propagation)
    addPass(createMachineCopyPropagationPass(true));
  if (TM->getOptLevel() != CodeGenOptLevel::None)
    addPass(createAArch64RedundantCondBranchPass());
 
  addPass(createAArch64A53Fix835769LegacyPass());
 
  if (TM->getTargetTriple().isOSWindows()) {
    // Identify valid longjmp targets for Windows Control Flow Guard.
    addPass(createCFGuardLongjmpPass());
    // Identify valid eh continuation targets for Windows EHCont Guard.
    addPass(createEHContGuardTargetsLegacy());
  }
 
  if (TM->getOptLevel() != CodeGenOptLevel::None && CLOpts.enable_collect_loh &&
      TM->getTargetTriple().isOSBinFormatMachO())
    addPass(createAArch64CollectLOHPass());
 
  // Apply code layout optimizations. Run late so detection reflects the
  // final MI stream.
  if (getOptLevel() != CodeGenOptLevel::None)
    addPass(createAArch64CodeLayoutOptPass());
}
 
void AArch64PassConfig::addPostBBSections() {
  addPass(createAArch64SLSHardeningLegacyPass());
  addPass(createAArch64PointerAuthPass());
  if (CLOpts.enable_branch_targets)
    addPass(createAArch64BranchTargetsPass());
  // Relax conditional branch instructions if they're otherwise out of
  // range of their destination.
  if (CLOpts.enable_branch_relax)
    addPass(&BranchRelaxationPassID);
 
  if (TM->getOptLevel() != CodeGenOptLevel::None &&
      CLOpts.enable_compress_jump_tables)
    addPass(createAArch64CompressJumpTablesPass());
}
 
void AArch64PassConfig::addPreEmitPass2() {
  // Insert pseudo probe annotation for callsite profiling
  addPass(createPseudoProbeInserter());
 
  // SVE bundles move prefixes with destructive operations. BLR_RVMARKER pseudo
  // instructions are lowered to bundles as well.
  addPass(createUnpackMachineBundlesLegacy(nullptr));
}
 
bool AArch64PassConfig::addRegAssignAndRewriteOptimized() {
  addPass(createAArch64PostCoalescerPass());
  return TargetPassConfig::addRegAssignAndRewriteOptimized();
}
 
MachineFunctionInfo *AArch64TargetMachine::createMachineFunctionInfo(
    BumpPtrAllocator &Allocator, const Function &F,
    const TargetSubtargetInfo *STI) const {
  return AArch64FunctionInfo::create<AArch64FunctionInfo>(
      Allocator, F, static_cast<const AArch64Subtarget *>(STI));
}
 
yaml::MachineFunctionInfo *
AArch64TargetMachine::createDefaultFuncInfoYAML() const {
  return new yaml::AArch64FunctionInfo();
}
 
yaml::MachineFunctionInfo *
AArch64TargetMachine::convertFuncInfoToYAML(const MachineFunction &MF) const {
  const auto *MFI = MF.getInfo<AArch64FunctionInfo>();
  return new yaml::AArch64FunctionInfo(*MFI);
}
 
bool AArch64TargetMachine::parseMachineFunctionInfo(
    const yaml::MachineFunctionInfo &MFI, PerFunctionMIParsingState &PFS,
    SMDiagnostic &Error, SMRange &SourceRange) const {
  const auto &YamlMFI = static_cast<const yaml::AArch64FunctionInfo &>(MFI);
  MachineFunction &MF = PFS.MF;
  MF.getInfo<AArch64FunctionInfo>()->initializeBaseYamlFields(YamlMFI);
  return false;
}