llvm/lib/Target/ARM/ARMTargetMachine.cpp

//===-- ARMTargetMachine.cpp - Define TargetMachine for ARM ---------------===//
//
// 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 "ARMTargetMachine.h"
#include "ARM.h"
#include "ARMLatencyMutations.h"
#include "ARMMachineFunctionInfo.h"
#include "ARMMacroFusion.h"
#include "ARMSubtarget.h"
#include "ARMTargetObjectFile.h"
#include "ARMTargetTransformInfo.h"
#include "MCTargetDesc/ARMMCTargetDesc.h"
#include "TargetInfo/ARMTargetInfo.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/CodeGen/ExecutionDomainFix.h"
#include "llvm/CodeGen/GlobalISel/CSEInfo.h"
#include "llvm/CodeGen/GlobalISel/CallLowering.h"
#include "llvm/CodeGen/GlobalISel/IRTranslator.h"
#include "llvm/CodeGen/GlobalISel/InstructionSelect.h"
#include "llvm/CodeGen/GlobalISel/Legalizer.h"
#include "llvm/CodeGen/GlobalISel/LegalizerInfo.h"
#include "llvm/CodeGen/GlobalISel/RegBankSelect.h"
#include "llvm/CodeGen/MIRParser/MIParser.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineScheduler.h"
#include "llvm/CodeGen/Passes.h"
#include "llvm/CodeGen/TargetPassConfig.h"
#include "llvm/IR/Attributes.h"
#include "llvm/IR/CallingConv.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/DiagnosticInfo.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/InstIterator.h"
#include "llvm/IR/InstrTypes.h"
#include "llvm/IR/Module.h"
#include "llvm/MC/TargetRegistry.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/Support/ErrorHandling.h"
#include "llvm/Target/TargetLoweringObjectFile.h"
#include "llvm/Target/TargetOptions.h"
#include "llvm/TargetParser/ARMTargetParser.h"
#include "llvm/TargetParser/Triple.h"
#include "llvm/Transforms/CFGuard.h"
#include "llvm/Transforms/IPO.h"
#include "llvm/Transforms/Scalar.h"
#include <cassert>
#include <memory>
#include <optional>
#include <string>

using namespace llvm;

static cl::opt<bool>
DisableA15SDOptimization("disable-a15-sd-optimization", cl::Hidden,
                   cl::desc("Inhibit optimization of S->D register accesses on A15"),
                   cl::init(false));

static cl::opt<bool>
EnableAtomicTidy("arm-atomic-cfg-tidy", cl::Hidden,
                 cl::desc("Run SimplifyCFG after expanding atomic operations"
                          " to make use of cmpxchg flow-based information"),
                 cl::init(true));

static cl::opt<bool>
EnableARMLoadStoreOpt("arm-load-store-opt", cl::Hidden,
                      cl::desc("Enable ARM load/store optimization pass"),
                      cl::init(true));

// FIXME: Unify control over GlobalMerge.
static cl::opt<cl::boolOrDefault>
EnableGlobalMerge("arm-global-merge", cl::Hidden,
                  cl::desc("Enable the global merge pass"));

namespace llvm {
  void initializeARMExecutionDomainFixPass(PassRegistry&);
}

extern "C" LLVM_ABI LLVM_EXTERNAL_VISIBILITY void LLVMInitializeARMTarget() {
  // Register the target.
  RegisterTargetMachine<ARMLETargetMachine> X(getTheARMLETarget());
  RegisterTargetMachine<ARMLETargetMachine> A(getTheThumbLETarget());
  RegisterTargetMachine<ARMBETargetMachine> Y(getTheARMBETarget());
  RegisterTargetMachine<ARMBETargetMachine> B(getTheThumbBETarget());

  PassRegistry &Registry = *PassRegistry::getPassRegistry();
  initializeGlobalISel(Registry);
  initializeARMAsmPrinterPass(Registry);
  initializeARMLoadStoreOptLegacyPass(Registry);
  initializeARMPreAllocLoadStoreOptLegacyPass(Registry);
  initializeARMParallelDSPPass(Registry);
  initializeARMBranchTargetsPass(Registry);
  initializeARMConstantIslandsPass(Registry);
  initializeARMExecutionDomainFixPass(Registry);
  initializeARMExpandPseudoPass(Registry);
  initializeThumb2SizeReducePass(Registry);
  initializeMVEVPTBlockPass(Registry);
  initializeMVETPAndVPTOptimisationsPass(Registry);
  initializeMVETailPredicationPass(Registry);
  initializeARMLowOverheadLoopsPass(Registry);
  initializeARMBlockPlacementPass(Registry);
  initializeMVEGatherScatterLoweringPass(Registry);
  initializeARMSLSHardeningPass(Registry);
  initializeMVELaneInterleavingPass(Registry);
  initializeARMFixCortexA57AES1742098Pass(Registry);
  initializeARMDAGToDAGISelLegacyPass(Registry);
  initializeMachineKCFILegacyPass(Registry);
}

static std::unique_ptr<TargetLoweringObjectFile> createTLOF(const Triple &TT) {
  if (TT.isOSBinFormatMachO())
    return std::make_unique<TargetLoweringObjectFileMachO>();
  if (TT.isOSWindows())
    return std::make_unique<TargetLoweringObjectFileCOFF>();
  return std::make_unique<ARMElfTargetObjectFile>();
}

static Reloc::Model getEffectiveRelocModel(const Triple &TT,
                                           std::optional<Reloc::Model> RM) {
  if (!RM)
    // Default relocation model on Darwin is PIC.
    return TT.isOSBinFormatMachO() ? Reloc::PIC_ : Reloc::Static;

  if (*RM == Reloc::ROPI || *RM == Reloc::RWPI || *RM == Reloc::ROPI_RWPI)
    assert(TT.isOSBinFormatELF() &&
           "ROPI/RWPI currently only supported for ELF");

  // DynamicNoPIC is only used on darwin.
  if (*RM == Reloc::DynamicNoPIC && !TT.isOSDarwin())
    return Reloc::Static;

  return *RM;
}

/// Create an ARM architecture model.
///
ARMBaseTargetMachine::ARMBaseTargetMachine(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)
    : CodeGenTargetMachineImpl(T, TT, CPU, FS, Options,
                               getEffectiveRelocModel(TT, RM),
                               getEffectiveCodeModel(CM, CodeModel::Small), OL),
      TargetABI(ARM::computeTargetABI(TT, Options.MCOptions.ABIName)),
      TLOF(createTLOF(getTargetTriple())), isLittle(TT.isLittleEndian()) {

  if (TT.isOSBinFormatMachO()) {
    this->Options.TrapUnreachable = true;
    this->Options.NoTrapAfterNoreturn = true;
  }

  // ARM supports the debug entry values.
  setSupportsDebugEntryValues(true);

  initAsmInfo();

  // ARM supports the MachineOutliner.
  setMachineOutliner(true);
  setSupportsDefaultOutlining(true);
}

ARMBaseTargetMachine::~ARMBaseTargetMachine() = default;

MachineFunctionInfo *ARMBaseTargetMachine::createMachineFunctionInfo(
    BumpPtrAllocator &Allocator, const Function &F,
    const TargetSubtargetInfo *STI) const {
  const auto *ARMSTI = static_cast<const ARMSubtarget *>(STI);
  if (!ARMSTI->hasFPRegs() || ARMSTI->isThumb1Only() ||
      ARMSTI->useSoftFloat()) {
    const StringRef FPRegsUnavailableMsg =
        ", but floating-point registers are unavailable";
    const ARMTargetLowering *TLI = ARMSTI->getTargetLowering();

    if (TLI->getEffectiveCallingConv(F.getCallingConv(), F.isVarArg()) ==
        CallingConv::ARM_AAPCS_VFP) {
      F.getContext().diagnose(DiagnosticInfoUnsupported(
          F, Twine("calling convention is hard-float") + FPRegsUnavailableMsg,
          DiagnosticLocation(F.getSubprogram())));
    } else {
      for (const Instruction &I : instructions(F)) {
        const auto *CB = dyn_cast<CallBase>(&I);
        if (!CB || CB->isInlineAsm() ||
            (CB->getCalledFunction() && CB->getCalledFunction()->isIntrinsic()))
          continue;
        if (TLI->getEffectiveCallingConv(CB->getCallingConv(),
                                         CB->getFunctionType()->isVarArg()) ==
            CallingConv::ARM_AAPCS_VFP) {
          const Function *Callee = CB->getCalledFunction();
          F.getContext().diagnose(DiagnosticInfoUnsupported(
              F,
              (Callee ? Twine("'") + F.getName() + "' calls '" +
                            Callee->getName() + "', which"
                      : Twine("'") + F.getName() +
                            "' makes an indirect call that") +
                  " expects a hard-float calling convention" +
                  FPRegsUnavailableMsg,
              CB->getDebugLoc()));
        }
      }
    }
  }
  return ARMFunctionInfo::create<ARMFunctionInfo>(Allocator, F, ARMSTI);
}

FloatABI::ABIType ARMBaseTargetMachine::getFloatABI(const Module &M) const {
  // An explicit "float-abi" module flag always wins, even for AAPCS16.
  if (auto *Val = dyn_cast_or_null<MDString>(M.getModuleFlag("float-abi")))
    return *FloatABI::parseABIType(Val->getString());

  // With no explicit ABI, an explicit -target-abi=aapcs16 forces hard float
  // even on triples whose default float ABI is soft (the triple default only
  // detects AAPCS16 when it is the triple's own default ABI).
  if (TargetABI == ARM::ARM_ABI_AAPCS16)
    return FloatABI::Hard;
  // Otherwise fall back to the ABI implied by the target triple.
  return M.getTargetTriple().getDefaultFloatABI();
}

ARM::ARMABI ARMBaseTargetMachine::getEffectiveABI(const Module &M) const {
  // Consistency of "target-abi" and -target-abi is validated elsewhere.
  if (const auto *MD = cast_or_null<MDString>(M.getModuleFlag("target-abi")))
    return ARM::computeTargetABI(TargetTriple, MD->getString());
  return TargetABI;
}

const ARMSubtarget *
ARMBaseTargetMachine::getSubtargetImpl(const Function &F) const {
  Attribute CPUAttr = F.getFnAttribute("target-cpu");
  Attribute FSAttr = F.getFnAttribute("target-features");

  std::string CPU =
      CPUAttr.isValid() ? CPUAttr.getValueAsString().str() : TargetCPU;
  std::string FS =
      FSAttr.isValid() ? FSAttr.getValueAsString().str() : TargetFS;

  // FIXME: This is related to the code below to reset the target options,
  // we need to know whether or not the soft float flag is set on the
  // function before we can generate a subtarget. We also need to use
  // it as a key for the subtarget since that can be the only difference
  // between two functions.
  bool SoftFloat = F.getFnAttribute("use-soft-float").getValueAsBool();
  // If the soft float attribute is set on the function turn on the soft float
  // subtarget feature.
  if (SoftFloat)
    FS += FS.empty() ? "+soft-float" : ",+soft-float";

  // Use the optminsize to identify the subtarget, but don't use it in the
  // feature string.
  std::string Key = CPU + FS;
  if (F.hasMinSize())
    Key += "+minsize";

  DenormalMode DM = F.getDenormalFPEnv().DefaultMode;
  if (DM != DenormalMode::getIEEE())
    Key += "denormal-fp-math=" + DM.str();

  FloatABI::ABIType FloatABI = getFloatABI(*F.getParent());
  // It is legal to have FloatABI::Hard for targets with SIMD registers
  // but no floating-point hardware (mve+nofp).
  Key += FloatABI == FloatABI::Hard ? "+hard-float-abi" : "+soft-float-abi";

  ARM::ARMABI ABI = getEffectiveABI(*F.getParent());
  Key += "+abi=" + std::to_string((int)ABI);

  auto &I = SubtargetMap[Key];
  if (!I) {
    I = std::make_unique<ARMSubtarget>(TargetTriple, CPU, FS, *this, isLittle,
                                       FloatABI, ABI, F.hasMinSize(), DM);

    if (!I->isThumb() && !I->hasARMOps())
      F.getContext().emitError("Function '" + F.getName() + "' uses ARM "
          "instructions, but the target does not support ARM mode execution.");
  }

  return I.get();
}

TargetTransformInfo
ARMBaseTargetMachine::getTargetTransformInfo(const Function &F) const {
  return TargetTransformInfo(std::make_unique<ARMTTIImpl>(this, F));
}

ScheduleDAGInstrs *
ARMBaseTargetMachine::createMachineScheduler(MachineSchedContext *C) const {
  ScheduleDAGMILive *DAG = createSchedLive(C);
  // add DAG Mutations here.
  const ARMSubtarget &ST = C->MF->getSubtarget<ARMSubtarget>();
  if (ST.hasFusion())
    DAG->addMutation(createARMMacroFusionDAGMutation());
  return DAG;
}

ScheduleDAGInstrs *
ARMBaseTargetMachine::createPostMachineScheduler(MachineSchedContext *C) const {
  ScheduleDAGMI *DAG = createSchedPostRA(C);
  // add DAG Mutations here.
  const ARMSubtarget &ST = C->MF->getSubtarget<ARMSubtarget>();
  if (ST.hasFusion())
    DAG->addMutation(createARMMacroFusionDAGMutation());
  if (auto Mutation = createARMLatencyMutations(ST, C->AA))
    DAG->addMutation(std::move(Mutation));
  return DAG;
}

ARMLETargetMachine::ARMLETargetMachine(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)
    : ARMBaseTargetMachine(T, TT, CPU, FS, Options, RM, CM, OL) {}

ARMBETargetMachine::ARMBETargetMachine(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)
    : ARMBaseTargetMachine(T, TT, CPU, FS, Options, RM, CM, OL) {}

namespace {

/// ARM Code Generator Pass Configuration Options.
class ARMPassConfig : public TargetPassConfig {
public:
  ARMPassConfig(ARMBaseTargetMachine &TM, PassManagerBase &PM)
      : TargetPassConfig(TM, PM) {}

  ARMBaseTargetMachine &getARMTargetMachine() const {
    return getTM<ARMBaseTargetMachine>();
  }

  void addIRPasses() override;
  void addCodeGenPrepare() override;
  bool addPreISel() override;
  bool addInstSelector() override;
  bool addIRTranslator() override;
  bool addLegalizeMachineIR() override;
  bool addRegBankSelect() override;
  bool addGlobalInstructionSelect() override;
  void addPreRegAlloc() override;
  void addPreSched2() override;
  void addPreEmitPass() override;
  void addPreEmitPass2() override;

  std::unique_ptr<CSEConfigBase> getCSEConfig() const override;
};

class ARMExecutionDomainFix : public ExecutionDomainFix {
public:
  static char ID;
  ARMExecutionDomainFix() : ExecutionDomainFix(ID, ARM::DPRRegClass) {}
  StringRef getPassName() const override {
    return "ARM Execution Domain Fix";
  }
};
char ARMExecutionDomainFix::ID;

} // end anonymous namespace

INITIALIZE_PASS_BEGIN(ARMExecutionDomainFix, "arm-execution-domain-fix",
                      "ARM Execution Domain Fix", false, false)
INITIALIZE_PASS_DEPENDENCY(ReachingDefInfoWrapperPass)
INITIALIZE_PASS_END(ARMExecutionDomainFix, "arm-execution-domain-fix",
                    "ARM Execution Domain Fix", false, false)

void ARMBaseTargetMachine::registerPassBuilderCallbacks(PassBuilder &PB) {
#define GET_PASS_REGISTRY "ARMPassRegistry.def"
#include "llvm/Passes/TargetPassRegistry.inc"
}

TargetPassConfig *ARMBaseTargetMachine::createPassConfig(PassManagerBase &PM) {
  return new ARMPassConfig(*this, PM);
}

std::unique_ptr<CSEConfigBase> ARMPassConfig::getCSEConfig() const {
  return getStandardCSEConfigForOpt(TM->getOptLevel());
}

void ARMPassConfig::addIRPasses() {
  addPass(createAtomicExpandLegacyPass());

  // 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 && EnableAtomicTidy)
    addPass(createCFGSimplificationPass(
        SimplifyCFGOptions().hoistCommonInsts(true).sinkCommonInsts(true),
        [this](const Function &F) {
          const auto &ST = this->TM->getSubtarget<ARMSubtarget>(F);
          return ST.hasAnyDataBarrier() && !ST.isThumb1Only();
        }));

  addPass(createMVEGatherScatterLoweringPass());
  addPass(createMVELaneInterleavingPass());

  TargetPassConfig::addIRPasses();

  // Run the parallel DSP pass.
  if (getOptLevel() == CodeGenOptLevel::Aggressive)
    addPass(createARMParallelDSPPass());

  // 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(createInterleavedAccessPass());

  // Add Control Flow Guard checks.
  if (TM->getTargetTriple().isOSWindows())
    addPass(createCFGuardPass());

  if (TM->Options.JMCInstrument)
    addPass(createJMCInstrumenterPass());
}

void ARMPassConfig::addCodeGenPrepare() {
  if (getOptLevel() != CodeGenOptLevel::None)
    addPass(createTypePromotionLegacyPass());
  TargetPassConfig::addCodeGenPrepare();
}

bool ARMPassConfig::addPreISel() {
  if ((TM->getOptLevel() != CodeGenOptLevel::None &&
       EnableGlobalMerge == cl::boolOrDefault::BOU_UNSET) ||
      EnableGlobalMerge == cl::boolOrDefault::BOU_TRUE) {
    // FIXME: This is using the thumb1 only constant value for
    // maximal global offset for merging globals. We may want
    // to look into using the old value for non-thumb1 code of
    // 4095 based on the TargetMachine, but this starts to become
    // tricky when doing code gen per function.
    bool OnlyOptimizeForSize =
        (TM->getOptLevel() < CodeGenOptLevel::Aggressive) &&
        (EnableGlobalMerge == cl::boolOrDefault::BOU_UNSET);
    // 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, 127, OnlyOptimizeForSize,
                                  MergeExternalByDefault));
  }

  if (TM->getOptLevel() != CodeGenOptLevel::None) {
    addPass(createHardwareLoopsLegacyPass());
    addPass(createMVETailPredicationPass());
    // FIXME: IR passes can delete address-taken basic blocks, deleting
    // corresponding blockaddresses. ARMConstantPoolConstant holds references to
    // address-taken basic blocks which can be invalidated if the function
    // containing the blockaddress has already been codegen'd and the basic
    // block is removed. Work around this by forcing all IR passes to run before
    // any ISel takes place. We should have a more principled way of handling
    // this. See D99707 for more details.
    addPass(createBarrierNoopPass());
  }

  return false;
}

bool ARMPassConfig::addInstSelector() {
  addPass(createARMISelDag(getARMTargetMachine(), getOptLevel()));
  return false;
}

bool ARMPassConfig::addIRTranslator() {
  addPass(new IRTranslatorLegacy(getOptLevel()));
  return false;
}

bool ARMPassConfig::addLegalizeMachineIR() {
  addPass(new LegalizerLegacy());
  return false;
}

bool ARMPassConfig::addRegBankSelect() {
  addPass(new RegBankSelectLegacy());
  return false;
}

bool ARMPassConfig::addGlobalInstructionSelect() {
  addPass(new InstructionSelectLegacy(getOptLevel()));
  return false;
}

void ARMPassConfig::addPreRegAlloc() {
  if (getOptLevel() != CodeGenOptLevel::None) {
    if (getOptLevel() == CodeGenOptLevel::Aggressive)
      addPass(&MachinePipelinerID);

    addPass(createMVETPAndVPTOptimisationsPass());

    addPass(createMLxExpansionPass());

    if (EnableARMLoadStoreOpt)
      addPass(createARMLoadStoreOptLegacyPass(/* pre-register alloc */ true));

    if (!DisableA15SDOptimization)
      addPass(createA15SDOptimizerPass());
  }
}

void ARMPassConfig::addPreSched2() {
  if (getOptLevel() != CodeGenOptLevel::None) {
    if (EnableARMLoadStoreOpt)
      addPass(createARMLoadStoreOptLegacyPass());

    addPass(new ARMExecutionDomainFix());
    addPass(createBreakFalseDepsLegacyPass());
  }

  // Expand some pseudo instructions into multiple instructions to allow
  // proper scheduling.
  addPass(createARMExpandPseudoPass());

  // Emit KCFI checks for indirect calls.
  addPass(createKCFIPass());

  if (getOptLevel() != CodeGenOptLevel::None) {
    // When optimising for size, always run the Thumb2SizeReduction pass before
    // IfConversion. Otherwise, check whether IT blocks are restricted
    // (e.g. in v8, IfConversion depends on Thumb instruction widths)
    addPass(createThumb2SizeReductionPass([this](const Function &F) {
      return this->TM->getSubtarget<ARMSubtarget>(F).hasMinSize() ||
             this->TM->getSubtarget<ARMSubtarget>(F).restrictIT();
    }));

    addPass(createIfConverter([](const MachineFunction &MF) {
      return !MF.getSubtarget<ARMSubtarget>().isThumb1Only();
    }));
  }
  addPass(createThumb2ITBlockPass());

  // Add both scheduling passes to give the subtarget an opportunity to pick
  // between them.
  if (getOptLevel() != CodeGenOptLevel::None) {
    addPass(&PostMachineSchedulerID);
    addPass(&PostRASchedulerID);
  }

  addPass(createMVEVPTBlockPass());
  addPass(createARMIndirectThunks());
  addPass(createARMSLSHardeningPass());
}

void ARMPassConfig::addPreEmitPass() {
  addPass(createThumb2SizeReductionPass());

  // Unpack bundles for:
  // - Thumb2: Constant island pass requires unbundled instructions
  // - KCFI: KCFI_CHECK pseudo instructions need to be unbundled for AsmPrinter
  addPass(createUnpackMachineBundlesLegacy([](const MachineFunction &MF) {
    return MF.getSubtarget<ARMSubtarget>().isThumb2() ||
           MF.getFunction().getParent()->getModuleFlag("kcfi");
  }));

  // Don't optimize barriers or block placement at -O0.
  if (getOptLevel() != CodeGenOptLevel::None) {
    addPass(createARMBlockPlacementPass());
    addPass(createARMOptimizeBarriersPass());
  }
}

void ARMPassConfig::addPreEmitPass2() {

  // Inserts fixup instructions before unsafe AES operations. Instructions may
  // be inserted at the start of blocks and at within blocks so this pass has to
  // come before those below.
  addPass(createARMFixCortexA57AES1742098Pass());
  // Inserts BTIs at the start of functions and indirectly-called basic blocks,
  // so passes cannot add to the start of basic blocks once this has run.
  addPass(createARMBranchTargetsPass());
  // Inserts Constant Islands. Block sizes cannot be increased after this point,
  // as this may push the branch ranges and load offsets of accessing constant
  // pools out of range..
  addPass(createARMConstantIslandPass());
  // Finalises Low-Overhead Loops. This replaces pseudo instructions with real
  // instructions, but the pseudos all have conservative sizes so that block
  // sizes will only be decreased by this pass.
  addPass(createARMLowOverheadLoopsPass());

  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());
  }
}

yaml::MachineFunctionInfo *
ARMBaseTargetMachine::createDefaultFuncInfoYAML() const {
  return new yaml::ARMFunctionInfo();
}

yaml::MachineFunctionInfo *
ARMBaseTargetMachine::convertFuncInfoToYAML(const MachineFunction &MF) const {
  const auto *MFI = MF.getInfo<ARMFunctionInfo>();
  return new yaml::ARMFunctionInfo(*MFI);
}

bool ARMBaseTargetMachine::parseMachineFunctionInfo(
    const yaml::MachineFunctionInfo &MFI, PerFunctionMIParsingState &PFS,
    SMDiagnostic &Error, SMRange &SourceRange) const {
  const auto &YamlMFI = static_cast<const yaml::ARMFunctionInfo &>(MFI);
  MachineFunction &MF = PFS.MF;
  MF.getInfo<ARMFunctionInfo>()->initializeBaseYamlFields(YamlMFI);
  return false;
}

void ARMBaseTargetMachine::reset() { SubtargetMap.clear(); }