RISCVTargetMachine.cpp

Go to the documentation of this file.

//===-- RISCVTargetMachine.cpp - Define TargetMachine for RISC-V ----------===//
//
// 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
//
//===----------------------------------------------------------------------===//
//
// Implements the info about RISC-V target spec.
//
//===----------------------------------------------------------------------===//
 
#include "RISCVTargetMachine.h"
#include "RISCV.h"
#include "RISCVMachineFunctionInfo.h"
#include "RISCVMachineScheduler.h"
#include "RISCVTargetObjectFile.h"
#include "RISCVTargetTransformInfo.h"
#include "TargetInfo/RISCVTargetInfo.h"
#include "llvm/Analysis/TargetTransformInfo.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/RegBankSelect.h"
#include "llvm/CodeGen/MIRParser/MIParser.h"
#include "llvm/CodeGen/MIRYamlMapping.h"
#include "llvm/CodeGen/MachineScheduler.h"
#include "llvm/CodeGen/MacroFusion.h"
#include "llvm/CodeGen/Passes.h"
#include "llvm/CodeGen/RegAllocRegistry.h"
#include "llvm/CodeGen/TargetLoweringObjectFileImpl.h"
#include "llvm/CodeGen/TargetPassConfig.h"
#include "llvm/InitializePasses.h"
#include "llvm/MC/TargetRegistry.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Target/TargetOptions.h"
#include "llvm/Transforms/IPO.h"
#include "llvm/Transforms/Scalar.h"
#include <optional>
using namespace llvm;
 
static cl::opt<bool> EnableRedundantCopyElimination(
    "riscv-enable-copyelim",
    cl::desc("Enable the redundant copy elimination pass"), cl::init(true),
    cl::Hidden);
 
// FIXME: Unify control over GlobalMerge.
static cl::opt<cl::boolOrDefault>
    EnableGlobalMerge("riscv-enable-global-merge", cl::Hidden,
                      cl::desc("Enable the global merge pass"));
 
static cl::opt<bool>
    EnableMachineCombiner("riscv-enable-machine-combiner",
                          cl::desc("Enable the machine combiner pass"),
                          cl::init(true), cl::Hidden);
 
static cl::opt<unsigned> RVVVectorBitsMaxOpt(
    "riscv-v-vector-bits-max",
    cl::desc("Assume V extension vector registers are at most this big, "
             "with zero meaning no maximum size is assumed."),
    cl::init(0), cl::Hidden);
 
static cl::opt<int> RVVVectorBitsMinOpt(
    "riscv-v-vector-bits-min",
    cl::desc("Assume V extension vector registers are at least this big, "
             "with zero meaning no minimum size is assumed. A value of -1 "
             "means use Zvl*b extension. This is primarily used to enable "
             "autovectorization with fixed width vectors."),
    cl::init(-1), cl::Hidden);
 
static cl::opt<bool> EnableRISCVCopyPropagation(
    "riscv-enable-copy-propagation",
    cl::desc("Enable the copy propagation with RISC-V copy instr"),
    cl::init(true), cl::Hidden);
 
static cl::opt<bool> EnableRISCVDeadRegisterElimination(
    "riscv-enable-dead-defs", cl::Hidden,
    cl::desc("Enable the pass that removes dead"
             " definitions and replaces stores to"
             " them with stores to x0"),
    cl::init(true));
 
static cl::opt<bool>
    EnableSinkFold("riscv-enable-sink-fold",
                   cl::desc("Enable sinking and folding of instruction copies"),
                   cl::init(true), cl::Hidden);
 
static cl::opt<bool>
    EnableLoopDataPrefetch("riscv-enable-loop-data-prefetch", cl::Hidden,
                           cl::desc("Enable the loop data prefetch pass"),
                           cl::init(true));
 
static cl::opt<bool> DisableVectorMaskMutation(
    "riscv-disable-vector-mask-mutation",
    cl::desc("Disable the vector mask scheduling mutation"), cl::init(false),
    cl::Hidden);
 
static cl::opt<bool>
    EnableMachinePipeliner("riscv-enable-pipeliner",
                           cl::desc("Enable Machine Pipeliner for RISC-V"),
                           cl::init(false), cl::Hidden);
 
static cl::opt<bool> EnableCFIInstrInserter(
    "riscv-enable-cfi-instr-inserter",
    cl::desc("Enable CFI Instruction Inserter for RISC-V"), cl::init(false),
    cl::Hidden);
 
static cl::opt<bool>
    EnableSelectOpt("riscv-select-opt", cl::Hidden,
                    cl::desc("Enable select to branch optimizations"),
                    cl::init(true));
 
extern "C" LLVM_ABI LLVM_EXTERNAL_VISIBILITY void LLVMInitializeRISCVTarget() {
  RegisterTargetMachine<RISCVTargetMachine> X(getTheRISCV32Target());
  RegisterTargetMachine<RISCVTargetMachine> Y(getTheRISCV64Target());
  RegisterTargetMachine<RISCVTargetMachine> A(getTheRISCV32beTarget());
  RegisterTargetMachine<RISCVTargetMachine> B(getTheRISCV64beTarget());
  auto *PR = PassRegistry::getPassRegistry();
  initializeGlobalISel(*PR);
  initializeRISCVO0PreLegalizerCombinerPass(*PR);
  initializeRISCVPreLegalizerCombinerPass(*PR);
  initializeRISCVPostLegalizerCombinerPass(*PR);
  initializeMachineKCFILegacyPass(*PR);
  initializeRISCVDeadRegisterDefinitionsPass(*PR);
  initializeRISCVLateBranchOptPass(*PR);
  initializeRISCVMakeCompressibleOptPass(*PR);
  initializeRISCVQCRelaxMarkingPass(*PR);
  initializeRISCVGatherScatterLoweringLegacyPass(*PR);
  initializeRISCVCodeGenPrepareLegacyPass(*PR);
  initializeRISCVZacasABIFixLegacyPass(*PR);
  initializeRISCVExpandPseudoPostRALegacyPass(*PR);
  initializeRISCVMergeBaseOffsetOptPass(*PR);
  initializeRISCVOptWInstrsLegacyPass(*PR);
  initializeRISCVFoldMemOffsetLegacyPass(*PR);
  initializeRISCVExpandPseudoPreRALegacyPass(*PR);
  initializeRISCVExpandPseudoPreEmitLegacyPass(*PR);
  initializeRISCVVectorPeepholeLegacyPass(*PR);
  initializeRISCVVLOptimizerLegacyPass(*PR);
  initializeRISCVVMV0EliminationPass(*PR);
  initializeRISCVInsertVSETVLIPass(*PR);
  initializeRISCVInsertReadWriteCSRPass(*PR);
  initializeRISCVInsertWriteVXRMPass(*PR);
  initializeRISCVDAGToDAGISelLegacyPass(*PR);
  initializeRISCVMoveMergePass(*PR);
  initializeRISCVPushPopOptPass(*PR);
  initializeRISCVIndirectBranchTrackingPass(*PR);
  initializeRISCVLoadStoreOptPass(*PR);
  initializeRISCVPreAllocZilsdOptPass(*PR);
  initializeRISCVExpandPseudoAtomicsLegacyPass(*PR);
  initializeRISCVRedundantCopyEliminationPass(*PR);
  initializeRISCVAsmPrinterPass(*PR);
  initializeRISCVPromoteConstantPass(*PR);
}
 
static Reloc::Model getEffectiveRelocModel(const Triple &TT,
                                           std::optional<Reloc::Model> RM) {
  if (TT.isOSBinFormatMachO())
    return RM.value_or(Reloc::PIC_);
 
  return RM.value_or(Reloc::Static);
}
 
static std::unique_ptr<TargetLoweringObjectFile> createTLOF(const Triple &TT) {
  if (TT.isOSBinFormatMachO())
    return std::make_unique<RISCVMachOTargetObjectFile>();
  return std::make_unique<RISCVELFTargetObjectFile>();
}
 
RISCVTargetMachine::RISCVTargetMachine(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)
    : CodeGenTargetMachineImpl(
          T, TT.computeDataLayout(Options.MCOptions.getABIName()), TT, CPU, FS,
          Options, getEffectiveRelocModel(TT, RM),
          getEffectiveCodeModel(CM, CodeModel::Small), OL),
      TLOF(createTLOF(TT)) {
  initAsmInfo();
 
  // RISC-V supports the MachineOutliner.
  setMachineOutliner(true);
  setSupportsDefaultOutlining(true);
 
  // RISC-V supports the debug entry values.
  setSupportsDebugEntryValues(true);
 
  if (TT.isOSFuchsia() && !TT.isArch64Bit())
    report_fatal_error("Fuchsia is only supported for 64-bit");
 
  setCFIFixup(!EnableCFIInstrInserter);
}
 
const RISCVSubtarget *
RISCVTargetMachine::getSubtargetImpl(const Function &F) const {
  Attribute CPUAttr = F.getFnAttribute("target-cpu");
  Attribute TuneAttr = F.getFnAttribute("tune-cpu");
  Attribute FSAttr = F.getFnAttribute("target-features");
 
  std::string CPU =
      CPUAttr.isValid() ? CPUAttr.getValueAsString().str() : TargetCPU;
  std::string TuneCPU =
      TuneAttr.isValid() ? TuneAttr.getValueAsString().str() : CPU;
  std::string FS =
      FSAttr.isValid() ? FSAttr.getValueAsString().str() : TargetFS;
 
  unsigned RVVBitsMin = RVVVectorBitsMinOpt;
  unsigned RVVBitsMax = RVVVectorBitsMaxOpt;
 
  Attribute VScaleRangeAttr = F.getFnAttribute(Attribute::VScaleRange);
  if (VScaleRangeAttr.isValid()) {
    if (!RVVVectorBitsMinOpt.getNumOccurrences())
      RVVBitsMin = VScaleRangeAttr.getVScaleRangeMin() * RISCV::RVVBitsPerBlock;
    std::optional<unsigned> VScaleMax = VScaleRangeAttr.getVScaleRangeMax();
    if (VScaleMax.has_value() && !RVVVectorBitsMaxOpt.getNumOccurrences())
      RVVBitsMax = *VScaleMax * RISCV::RVVBitsPerBlock;
  }
 
  if (RVVBitsMin != -1U) {
    // FIXME: Change to >= 32 when VLEN = 32 is supported.
    assert((RVVBitsMin == 0 || (RVVBitsMin >= 64 && RVVBitsMin <= 65536 &&
                                isPowerOf2_32(RVVBitsMin))) &&
           "V or Zve* extension requires vector length to be in the range of "
           "64 to 65536 and a power 2!");
    assert((RVVBitsMax >= RVVBitsMin || RVVBitsMax == 0) &&
           "Minimum V extension vector length should not be larger than its "
           "maximum!");
  }
  assert((RVVBitsMax == 0 || (RVVBitsMax >= 64 && RVVBitsMax <= 65536 &&
                              isPowerOf2_32(RVVBitsMax))) &&
         "V or Zve* extension requires vector length to be in the range of "
         "64 to 65536 and a power 2!");
 
  if (RVVBitsMin != -1U) {
    if (RVVBitsMax != 0) {
      RVVBitsMin = std::min(RVVBitsMin, RVVBitsMax);
      RVVBitsMax = std::max(RVVBitsMin, RVVBitsMax);
    }
 
    RVVBitsMin = llvm::bit_floor(
        (RVVBitsMin < 64 || RVVBitsMin > 65536) ? 0 : RVVBitsMin);
  }
  RVVBitsMax =
      llvm::bit_floor((RVVBitsMax < 64 || RVVBitsMax > 65536) ? 0 : RVVBitsMax);
 
  SmallString<512> Key;
  raw_svector_ostream(Key) << "RVVMin" << RVVBitsMin << "RVVMax" << RVVBitsMax
                           << CPU << TuneCPU << FS;
  auto &I = SubtargetMap[Key];
  if (!I) {
    StringRef ABIName = getTargetABIName(*F.getParent());
    I = std::make_unique<RISCVSubtarget>(
        TargetTriple, CPU, TuneCPU, FS, ABIName, RVVBitsMin, RVVBitsMax, *this);
  }
  return I.get();
}
 
MachineFunctionInfo *RISCVTargetMachine::createMachineFunctionInfo(
    BumpPtrAllocator &Allocator, const Function &F,
    const TargetSubtargetInfo *STI) const {
  return RISCVMachineFunctionInfo::create<RISCVMachineFunctionInfo>(
      Allocator, F, static_cast<const RISCVSubtarget *>(STI));
}
 
TargetTransformInfo
RISCVTargetMachine::getTargetTransformInfo(const Function &F) const {
  return TargetTransformInfo(std::make_unique<RISCVTTIImpl>(this, F));
}
 
// A RISC-V hart has a single byte-addressable address space of 2^XLEN bytes
// for all memory accesses, so it is reasonable to assume that an
// implementation has no-op address space casts. If an implementation makes a
// change to this, they can override it here.
bool RISCVTargetMachine::isNoopAddrSpaceCast(unsigned SrcAS,
                                             unsigned DstAS) const {
  return true;
}
 
ScheduleDAGInstrs *
RISCVTargetMachine::createMachineScheduler(MachineSchedContext *C) const {
  const RISCVSubtarget &ST = C->MF->getSubtarget<RISCVSubtarget>();
  ScheduleDAGMILive *DAG = createSchedLive<RISCVPreRAMachineSchedStrategy>(C);
 
  // Add MacroFusion mutation first with a higher priority than later clustering
  const auto &MacroFusions = ST.getMacroFusions();
  if (!MacroFusions.empty())
    DAG->addMutation(createMacroFusionDAGMutation(MacroFusions));
 
  if (ST.enableMISchedLoadClustering())
    DAG->addMutation(createLoadClusterDAGMutation(
        DAG->TII, DAG->TRI, /*ReorderWhileClustering=*/true));
 
  if (ST.enableMISchedStoreClustering())
    DAG->addMutation(createStoreClusterDAGMutation(
        DAG->TII, DAG->TRI, /*ReorderWhileClustering=*/true));
 
  if (!DisableVectorMaskMutation && ST.hasVInstructions())
    DAG->addMutation(createRISCVVectorMaskDAGMutation(DAG->TRI));
 
  return DAG;
}
 
ScheduleDAGInstrs *
RISCVTargetMachine::createPostMachineScheduler(MachineSchedContext *C) const {
  const RISCVSubtarget &ST = C->MF->getSubtarget<RISCVSubtarget>();
  ScheduleDAGMI *DAG = createSchedPostRA(C);
 
  // Add MacroFusion mutation first with a higher priority than later clustering
  const auto &MacroFusions = ST.getMacroFusions();
  if (!MacroFusions.empty())
    DAG->addMutation(createMacroFusionDAGMutation(MacroFusions));
 
  if (ST.enablePostMISchedLoadClustering())
    DAG->addMutation(createLoadClusterDAGMutation(
        DAG->TII, DAG->TRI, /*ReorderWhileClustering=*/true));
 
  if (ST.enablePostMISchedStoreClustering())
    DAG->addMutation(createStoreClusterDAGMutation(
        DAG->TII, DAG->TRI, /*ReorderWhileClustering=*/true));
 
  return DAG;
}
 
namespace {
 
class RVVRegisterRegAlloc : public RegisterRegAllocBase<RVVRegisterRegAlloc> {
public:
  RVVRegisterRegAlloc(const char *N, const char *D, FunctionPassCtor C)
      : RegisterRegAllocBase(N, D, C) {}
};
 
static bool onlyAllocateRVVReg(const TargetRegisterInfo &TRI,
                               const MachineRegisterInfo &MRI,
                               const Register Reg) {
  const TargetRegisterClass *RC = MRI.getRegClass(Reg);
  return RISCVRegisterInfo::isRVVRegClass(RC);
}
 
static FunctionPass *useDefaultRegisterAllocator() { return nullptr; }
 
static llvm::once_flag InitializeDefaultRVVRegisterAllocatorFlag;
 
/// -riscv-rvv-regalloc=<fast|basic|greedy> command line option.
/// This option could designate the rvv register allocator only.
/// For example: -riscv-rvv-regalloc=basic
static cl::opt<RVVRegisterRegAlloc::FunctionPassCtor, false,
               RegisterPassParser<RVVRegisterRegAlloc>>
    RVVRegAlloc("riscv-rvv-regalloc", cl::Hidden,
                cl::init(&useDefaultRegisterAllocator),
                cl::desc("Register allocator to use for RVV register."));
 
static void initializeDefaultRVVRegisterAllocatorOnce() {
  RegisterRegAlloc::FunctionPassCtor Ctor = RVVRegisterRegAlloc::getDefault();
 
  if (!Ctor) {
    Ctor = RVVRegAlloc;
    RVVRegisterRegAlloc::setDefault(RVVRegAlloc);
  }
}
 
static FunctionPass *createBasicRVVRegisterAllocator() {
  return createBasicRegisterAllocator(onlyAllocateRVVReg);
}
 
static FunctionPass *createGreedyRVVRegisterAllocator() {
  return createGreedyRegisterAllocator(onlyAllocateRVVReg);
}
 
static FunctionPass *createFastRVVRegisterAllocator() {
  return createFastRegisterAllocator(onlyAllocateRVVReg, false);
}
 
static RVVRegisterRegAlloc basicRegAllocRVVReg("basic",
                                               "basic register allocator",
                                               createBasicRVVRegisterAllocator);
static RVVRegisterRegAlloc
    greedyRegAllocRVVReg("greedy", "greedy register allocator",
                         createGreedyRVVRegisterAllocator);
 
static RVVRegisterRegAlloc fastRegAllocRVVReg("fast", "fast register allocator",
                                              createFastRVVRegisterAllocator);
 
class RISCVPassConfig : public TargetPassConfig {
public:
  RISCVPassConfig(RISCVTargetMachine &TM, PassManagerBase &PM)
      : TargetPassConfig(TM, PM) {
    if (TM.getOptLevel() != CodeGenOptLevel::None)
      substitutePass(&PostRASchedulerID, &PostMachineSchedulerID);
    setEnableSinkAndFold(EnableSinkFold);
    EnableLoopTermFold = true;
  }
 
  RISCVTargetMachine &getRISCVTargetMachine() const {
    return getTM<RISCVTargetMachine>();
  }
 
  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 addPreEmitPass() override;
  void addPreEmitPass2() override;
  void addPreSched2() override;
  void addMachineSSAOptimization() override;
  FunctionPass *createRVVRegAllocPass(bool Optimized);
  bool addRegAssignAndRewriteFast() override;
  bool addRegAssignAndRewriteOptimized() override;
  void addPreRegAlloc() override;
  void addPostRegAlloc() override;
  void addFastRegAlloc() override;
  bool addILPOpts() override;
 
  std::unique_ptr<CSEConfigBase> getCSEConfig() const override;
};
} // namespace
 
TargetPassConfig *RISCVTargetMachine::createPassConfig(PassManagerBase &PM) {
  return new RISCVPassConfig(*this, PM);
}
 
std::unique_ptr<CSEConfigBase> RISCVPassConfig::getCSEConfig() const {
  return getStandardCSEConfigForOpt(TM->getOptLevel());
}
 
FunctionPass *RISCVPassConfig::createRVVRegAllocPass(bool Optimized) {
  // Initialize the global default.
  llvm::call_once(InitializeDefaultRVVRegisterAllocatorFlag,
                  initializeDefaultRVVRegisterAllocatorOnce);
 
  RegisterRegAlloc::FunctionPassCtor Ctor = RVVRegisterRegAlloc::getDefault();
  if (Ctor != useDefaultRegisterAllocator)
    return Ctor();
 
  if (Optimized)
    return createGreedyRVVRegisterAllocator();
 
  return createFastRVVRegisterAllocator();
}
 
bool RISCVPassConfig::addRegAssignAndRewriteFast() {
  addPass(createRVVRegAllocPass(false));
  addPass(createRISCVInsertVSETVLIPass());
  if (TM->getOptLevel() != CodeGenOptLevel::None &&
      EnableRISCVDeadRegisterElimination)
    addPass(createRISCVDeadRegisterDefinitionsPass());
  return TargetPassConfig::addRegAssignAndRewriteFast();
}
 
bool RISCVPassConfig::addRegAssignAndRewriteOptimized() {
  addPass(createRVVRegAllocPass(true));
  addPass(createVirtRegRewriter(false));
  addPass(createRISCVInsertVSETVLIPass());
  if (TM->getOptLevel() != CodeGenOptLevel::None &&
      EnableRISCVDeadRegisterElimination)
    addPass(createRISCVDeadRegisterDefinitionsPass());
  return TargetPassConfig::addRegAssignAndRewriteOptimized();
}
 
void RISCVPassConfig::addIRPasses() {
  addPass(createAtomicExpandLegacyPass());
  addPass(createRISCVZacasABIFixLegacyPass());
 
  if (getOptLevel() != CodeGenOptLevel::None) {
    if (EnableLoopDataPrefetch)
      addPass(createLoopDataPrefetchPass());
 
    addPass(createRISCVGatherScatterLoweringLegacyPass());
    addPass(createInterleavedAccessPass());
    addPass(createRISCVCodeGenPrepareLegacyPass());
  }
 
  TargetPassConfig::addIRPasses();
 
  if (getOptLevel() == CodeGenOptLevel::Aggressive && EnableSelectOpt)
    addPass(createSelectOptimizePass());
}
 
bool RISCVPassConfig::addPreISel() {
  if (TM->getOptLevel() != CodeGenOptLevel::None)
    addPass(createRISCVPromoteConstantPass());
  if (TM->getOptLevel() != CodeGenOptLevel::None) {
    // Add a barrier before instruction selection so that we will not get
    // deleted block address after enabling default outlining. See D99707 for
    // more details.
    addPass(createBarrierNoopPass());
  }
 
  if ((TM->getOptLevel() != CodeGenOptLevel::None &&
       EnableGlobalMerge == cl::boolOrDefault::BOU_UNSET) ||
      EnableGlobalMerge == cl::boolOrDefault::BOU_TRUE) {
    // FIXME: Like AArch64, we disable extern global merging by default due to
    // concerns it might regress some workloads. Unlike AArch64, we don't
    // currently support enabling the pass in an "OnlyOptimizeForSize" mode.
    // Investigating and addressing both items are TODO.
    addPass(createGlobalMergePass(TM, /* MaxOffset */ 2047,
                                  /* OnlyOptimizeForSize */ false,
                                  /* MergeExternalByDefault */ true));
  }
 
  return false;
}
 
void RISCVPassConfig::addCodeGenPrepare() {
  if (getOptLevel() != CodeGenOptLevel::None)
    addPass(createTypePromotionLegacyPass());
  TargetPassConfig::addCodeGenPrepare();
}
 
bool RISCVPassConfig::addInstSelector() {
  addPass(createRISCVISelDagLegacyPass(getRISCVTargetMachine(), getOptLevel()));
 
  return false;
}
 
bool RISCVPassConfig::addIRTranslator() {
  addPass(new IRTranslatorLegacy(getOptLevel()));
  return false;
}
 
void RISCVPassConfig::addPreLegalizeMachineIR() {
  if (getOptLevel() == CodeGenOptLevel::None) {
    addPass(createRISCVO0PreLegalizerCombiner());
  } else {
    addPass(createRISCVPreLegalizerCombiner());
  }
}
 
bool RISCVPassConfig::addLegalizeMachineIR() {
  addPass(new LegalizerLegacy());
  return false;
}
 
void RISCVPassConfig::addPreRegBankSelect() {
  if (getOptLevel() != CodeGenOptLevel::None)
    addPass(createRISCVPostLegalizerCombiner());
}
 
bool RISCVPassConfig::addRegBankSelect() {
  addPass(new RegBankSelectLegacy());
  return false;
}
 
bool RISCVPassConfig::addGlobalInstructionSelect() {
  addPass(new InstructionSelectLegacy(getOptLevel()));
  return false;
}
 
void RISCVPassConfig::addPreSched2() {
  addPass(createRISCVExpandPseudoPostRALegacyPass());
 
  // Emit KCFI checks for indirect calls.
  addPass(createKCFIPass());
  if (TM->getOptLevel() != CodeGenOptLevel::None)
    addPass(createRISCVLoadStoreOptPass());
}
 
void RISCVPassConfig::addPreEmitPass() {
  // TODO: It would potentially be better to schedule copy propagation after
  // expanding pseudos (in addPreEmitPass2). However, performing copy
  // propagation after the machine outliner (which runs after addPreEmitPass)
  // currently leads to incorrect code-gen, where copies to registers within
  // outlined functions are removed erroneously.
  if (TM->getOptLevel() >= CodeGenOptLevel::Default &&
      EnableRISCVCopyPropagation)
    addPass(createMachineCopyPropagationPass(true));
  if (TM->getOptLevel() >= CodeGenOptLevel::Default)
    addPass(createRISCVLateBranchOptPass());
  // The IndirectBranchTrackingPass inserts lpad and could have changed the
  // basic block alignment. It must be done before Branch Relaxation to
  // prevent the adjusted offset exceeding the branch range.
  addPass(createRISCVIndirectBranchTrackingPass());
  addPass(&BranchRelaxationPassID);
  addPass(createRISCVMakeCompressibleOptPass());
}
 
void RISCVPassConfig::addPreEmitPass2() {
  if (TM->getOptLevel() != CodeGenOptLevel::None) {
    addPass(createRISCVMoveMergePass());
    // Schedule PushPop Optimization before expansion of Pseudo instruction,
    // ensuring return instruction is detected correctly.
    addPass(createRISCVPushPopOptimizationPass());
  }
  addPass(createRISCVExpandPseudoPreEmitLegacyPass());
 
  // Add QC Relaxation Markers as late as possible, and only for RV32
  if (TM->getOptLevel() != CodeGenOptLevel::None &&
      TM->getTargetTriple().isRISCV32())
    addPass(createRISCVQCRelaxMarkingPass());
 
  // Schedule the expansion of AMOs at the last possible moment, avoiding the
  // possibility for other passes to break the requirements for forward
  // progress in the LR/SC block.
  addPass(createRISCVExpandPseudoAtomicsLegacyPass());
 
  // KCFI indirect call checks are lowered to a bundle.
  addPass(createUnpackMachineBundlesLegacy([&](const MachineFunction &MF) {
    return MF.getFunction().getParent()->getModuleFlag("kcfi");
  }));
 
  if (EnableCFIInstrInserter)
    addPass(createCFIInstrInserterLegacy());
}
 
void RISCVPassConfig::addMachineSSAOptimization() {
  // It's beneficial to reduce the VL to enable more
  // Machine SSA optimizations.
  if (TM->getOptLevel() != CodeGenOptLevel::None) {
    // RISCVVLOptimizer can make loop invariant instructions like vmv.v.i
    // loop variant by propagating a VL defined inside the loop. Run LICM and
    // hoist them early. Don't do this at -O0 to avoid the compile-time
    // overhead. Not reducing the VL of loop invariant pseudos results in more
    // vsetvli toggles, and still requires the MachineLoopInfo analysis to be
    // run.
    addPass(&EarlyMachineLICMID);
    addPass(createRISCVVLOptimizerLegacyPass());
  }
 
  addPass(createRISCVVectorPeepholeLegacyPass());
  addPass(createRISCVFoldMemOffsetLegacyPass());
 
  TargetPassConfig::addMachineSSAOptimization();
 
  if (TM->getTargetTriple().isRISCV64()) {
    addPass(createRISCVOptWInstrsLegacyPass());
  }
}
 
void RISCVPassConfig::addPreRegAlloc() {
  addPass(createRISCVExpandPseudoPreRALegacyPass());
  if (TM->getOptLevel() != CodeGenOptLevel::None) {
    addPass(createRISCVMergeBaseOffsetOptPass());
    // Add Zilsd pre-allocation load/store optimization
    addPass(createRISCVPreAllocZilsdOptPass());
  }
 
  addPass(createRISCVInsertReadWriteCSRPass());
  addPass(createRISCVInsertWriteVXRMPass());
  addPass(createRISCVLandingPadSetupPass());
 
  if (TM->getOptLevel() != CodeGenOptLevel::None && EnableMachinePipeliner)
    addPass(&MachinePipelinerID);
 
  addPass(createRISCVVMV0EliminationPass());
}
 
void RISCVPassConfig::addFastRegAlloc() {
  addPass(&InitUndefID);
  TargetPassConfig::addFastRegAlloc();
}
 
 
void RISCVPassConfig::addPostRegAlloc() {
  if (TM->getOptLevel() != CodeGenOptLevel::None &&
      EnableRedundantCopyElimination)
    addPass(createRISCVRedundantCopyEliminationPass());
}
 
bool RISCVPassConfig::addILPOpts() {
  if (EnableMachineCombiner)
    addPass(&MachineCombinerID);
 
  return true;
}
 
yaml::MachineFunctionInfo *
RISCVTargetMachine::createDefaultFuncInfoYAML() const {
  return new yaml::RISCVMachineFunctionInfo();
}
 
yaml::MachineFunctionInfo *
RISCVTargetMachine::convertFuncInfoToYAML(const MachineFunction &MF) const {
  const auto *MFI = MF.getInfo<RISCVMachineFunctionInfo>();
  return new yaml::RISCVMachineFunctionInfo(*MFI);
}
 
bool RISCVTargetMachine::parseMachineFunctionInfo(
    const yaml::MachineFunctionInfo &MFI, PerFunctionMIParsingState &PFS,
    SMDiagnostic &Error, SMRange &SourceRange) const {
  const auto &YamlMFI =
      static_cast<const yaml::RISCVMachineFunctionInfo &>(MFI);
  PFS.MF.getInfo<RISCVMachineFunctionInfo>()->initializeBaseYamlFields(YamlMFI);
  return false;
}