llvm/lib/CodeGen/SelectionDAG/StatepointLowering.cpp

//===-- StatepointLowering.cpp - SDAGBuilder's statepoint code -----------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file includes support code use by SelectionDAGBuilder when lowering a
// statepoint sequence in SelectionDAG IR.
//
//===----------------------------------------------------------------------===//

#include "StatepointLowering.h"
#include "SelectionDAGBuilder.h"
#include "llvm/ADT/SmallSet.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/CodeGen/FunctionLoweringInfo.h"
#include "llvm/CodeGen/GCMetadata.h"
#include "llvm/CodeGen/GCStrategy.h"
#include "llvm/CodeGen/SelectionDAG.h"
#include "llvm/CodeGen/StackMaps.h"
#include "llvm/IR/CallingConv.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Intrinsics.h"
#include "llvm/IR/Statepoint.h"
#include "llvm/Target/TargetLowering.h"
#include <algorithm>
using namespace llvm;

#define DEBUG_TYPE "statepoint-lowering"

STATISTIC(NumSlotsAllocatedForStatepoints,
          "Number of stack slots allocated for statepoints");
STATISTIC(NumOfStatepoints, "Number of statepoint nodes encountered");
STATISTIC(StatepointMaxSlotsRequired,
          "Maximum number of stack slots required for a singe statepoint");

static void pushStackMapConstant(SmallVectorImpl<SDValue>& Ops,
                                 SelectionDAGBuilder &Builder, uint64_t Value) {
  SDLoc L = Builder.getCurSDLoc();
  Ops.push_back(Builder.DAG.getTargetConstant(StackMaps::ConstantOp, L,
                                              MVT::i64));
  Ops.push_back(Builder.DAG.getTargetConstant(Value, L, MVT::i64));
}

void StatepointLoweringState::startNewStatepoint(SelectionDAGBuilder &Builder) {
  // Consistency check
  assert(PendingGCRelocateCalls.empty() &&
         "Trying to visit statepoint before finished processing previous one");
  Locations.clear();
  NextSlotToAllocate = 0;
  // Need to resize this on each safepoint - we need the two to stay in
  // sync and the clear patterns of a SelectionDAGBuilder have no relation
  // to FunctionLoweringInfo.
  AllocatedStackSlots.resize(Builder.FuncInfo.StatepointStackSlots.size());
  for (size_t i = 0; i < AllocatedStackSlots.size(); i++) {
    AllocatedStackSlots[i] = false;
  }
}

void StatepointLoweringState::clear() {
  Locations.clear();
  AllocatedStackSlots.clear();
  assert(PendingGCRelocateCalls.empty() &&
         "cleared before statepoint sequence completed");
}

SDValue
StatepointLoweringState::allocateStackSlot(EVT ValueType,
                                           SelectionDAGBuilder &Builder) {

  NumSlotsAllocatedForStatepoints++;

  // The basic scheme here is to first look for a previously created stack slot
  // which is not in use (accounting for the fact arbitrary slots may already
  // be reserved), or to create a new stack slot and use it.

  // If this doesn't succeed in 40000 iterations, something is seriously wrong
  for (int i = 0; i < 40000; i++) {
    assert(Builder.FuncInfo.StatepointStackSlots.size() ==
               AllocatedStackSlots.size() &&
           "broken invariant");
    const size_t NumSlots = AllocatedStackSlots.size();
    assert(NextSlotToAllocate <= NumSlots && "broken invariant");

    if (NextSlotToAllocate >= NumSlots) {
      assert(NextSlotToAllocate == NumSlots);
      // record stats
      if (NumSlots + 1 > StatepointMaxSlotsRequired) {
        StatepointMaxSlotsRequired = NumSlots + 1;
      }

      SDValue SpillSlot = Builder.DAG.CreateStackTemporary(ValueType);
      const unsigned FI = cast<FrameIndexSDNode>(SpillSlot)->getIndex();
      Builder.FuncInfo.StatepointStackSlots.push_back(FI);
      AllocatedStackSlots.push_back(true);
      return SpillSlot;
    }
    if (!AllocatedStackSlots[NextSlotToAllocate]) {
      const int FI = Builder.FuncInfo.StatepointStackSlots[NextSlotToAllocate];
      AllocatedStackSlots[NextSlotToAllocate] = true;
      return Builder.DAG.getFrameIndex(FI, ValueType);
    }
    // Note: We deliberately choose to advance this only on the failing path.
    // Doing so on the succeeding path involves a bit of complexity that caused
    // a minor bug previously.  Unless performance shows this matters, please
    // keep this code as simple as possible.
    NextSlotToAllocate++;
  }
  llvm_unreachable("infinite loop?");
}

/// Utility function for reservePreviousStackSlotForValue. Tries to find
/// stack slot index to which we have spilled value for previous statepoints.
/// LookUpDepth specifies maximum DFS depth this function is allowed to look.
static Optional<int> findPreviousSpillSlot(const Value *Val,
                                           SelectionDAGBuilder &Builder,
                                           int LookUpDepth) {
  // Can not look any further - give up now
  if (LookUpDepth <= 0)
    return Optional<int>();

  // Spill location is known for gc relocates
  if (isGCRelocate(Val)) {
    GCRelocateOperands RelocOps(cast<Instruction>(Val));

    FunctionLoweringInfo::StatepointSpilledValueMapTy &SpillMap =
        Builder.FuncInfo.StatepointRelocatedValues[RelocOps.getStatepoint()];

    auto It = SpillMap.find(RelocOps.getDerivedPtr());
    if (It == SpillMap.end())
      return Optional<int>();

    return It->second;
  }

  // Look through bitcast instructions.
  if (const BitCastInst *Cast = dyn_cast<BitCastInst>(Val)) {
    return findPreviousSpillSlot(Cast->getOperand(0), Builder, LookUpDepth - 1);
  }

  // Look through phi nodes
  // All incoming values should have same known stack slot, otherwise result
  // is unknown.
  if (const PHINode *Phi = dyn_cast<PHINode>(Val)) {
    Optional<int> MergedResult = None;

    for (auto &IncomingValue : Phi->incoming_values()) {
      Optional<int> SpillSlot =
          findPreviousSpillSlot(IncomingValue, Builder, LookUpDepth - 1);
      if (!SpillSlot.hasValue())
        return Optional<int>();

      if (MergedResult.hasValue() && *MergedResult != *SpillSlot)
        return Optional<int>();

      MergedResult = SpillSlot;
    }
    return MergedResult;
  }

  // TODO: We can do better for PHI nodes. In cases like this:
  //   ptr = phi(relocated_pointer, not_relocated_pointer)
  //   statepoint(ptr)
  // We will return that stack slot for ptr is unknown. And later we might
  // assign different stack slots for ptr and relocated_pointer. This limits
  // llvm's ability to remove redundant stores.
  // Unfortunately it's hard to accomplish in current infrastructure.
  // We use this function to eliminate spill store completely, while
  // in example we still need to emit store, but instead of any location
  // we need to use special "preferred" location.

  // TODO: handle simple updates.  If a value is modified and the original
  // value is no longer live, it would be nice to put the modified value in the
  // same slot.  This allows folding of the memory accesses for some
  // instructions types (like an increment).
  //   statepoint (i)
  //   i1 = i+1
  //   statepoint (i1)
  // However we need to be careful for cases like this:
  //   statepoint(i)
  //   i1 = i+1
  //   statepoint(i, i1)
  // Here we want to reserve spill slot for 'i', but not for 'i+1'. If we just
  // put handling of simple modifications in this function like it's done
  // for bitcasts we might end up reserving i's slot for 'i+1' because order in
  // which we visit values is unspecified.

  // Don't know any information about this instruction
  return Optional<int>();
}

/// Try to find existing copies of the incoming values in stack slots used for
/// statepoint spilling.  If we can find a spill slot for the incoming value,
/// mark that slot as allocated, and reuse the same slot for this safepoint.
/// This helps to avoid series of loads and stores that only serve to reshuffle
/// values on the stack between calls.
static void reservePreviousStackSlotForValue(const Value *IncomingValue,
                                             SelectionDAGBuilder &Builder) {

  SDValue Incoming = Builder.getValue(IncomingValue);

  if (isa<ConstantSDNode>(Incoming) || isa<FrameIndexSDNode>(Incoming)) {
    // We won't need to spill this, so no need to check for previously
    // allocated stack slots
    return;
  }

  SDValue OldLocation = Builder.StatepointLowering.getLocation(Incoming);
  if (OldLocation.getNode())
    // duplicates in input
    return;

  const int LookUpDepth = 6;
  Optional<int> Index =
      findPreviousSpillSlot(IncomingValue, Builder, LookUpDepth);
  if (!Index.hasValue())
    return;

  auto Itr = std::find(Builder.FuncInfo.StatepointStackSlots.begin(),
                       Builder.FuncInfo.StatepointStackSlots.end(), *Index);
  assert(Itr != Builder.FuncInfo.StatepointStackSlots.end() &&
         "value spilled to the unknown stack slot");

  // This is one of our dedicated lowering slots
  const int Offset =
      std::distance(Builder.FuncInfo.StatepointStackSlots.begin(), Itr);
  if (Builder.StatepointLowering.isStackSlotAllocated(Offset)) {
    // stack slot already assigned to someone else, can't use it!
    // TODO: currently we reserve space for gc arguments after doing
    // normal allocation for deopt arguments.  We should reserve for
    // _all_ deopt and gc arguments, then start allocating.  This
    // will prevent some moves being inserted when vm state changes,
    // but gc state doesn't between two calls.
    return;
  }
  // Reserve this stack slot
  Builder.StatepointLowering.reserveStackSlot(Offset);

  // Cache this slot so we find it when going through the normal
  // assignment loop.
  SDValue Loc = Builder.DAG.getTargetFrameIndex(*Index, Incoming.getValueType());
  Builder.StatepointLowering.setLocation(Incoming, Loc);
}

/// Remove any duplicate (as SDValues) from the derived pointer pairs.  This
/// is not required for correctness.  It's purpose is to reduce the size of
/// StackMap section.  It has no effect on the number of spill slots required
/// or the actual lowering.
static void removeDuplicatesGCPtrs(SmallVectorImpl<const Value *> &Bases,
                                   SmallVectorImpl<const Value *> &Ptrs,
                                   SmallVectorImpl<const Value *> &Relocs,
                                   SelectionDAGBuilder &Builder) {

  // This is horribly inefficient, but I don't care right now
  SmallSet<SDValue, 64> Seen;

  SmallVector<const Value *, 64> NewBases, NewPtrs, NewRelocs;
  for (size_t i = 0; i < Ptrs.size(); i++) {
    SDValue SD = Builder.getValue(Ptrs[i]);
    // Only add non-duplicates
    if (Seen.count(SD) == 0) {
      NewBases.push_back(Bases[i]);
      NewPtrs.push_back(Ptrs[i]);
      NewRelocs.push_back(Relocs[i]);
    }
    Seen.insert(SD);
  }
  assert(Bases.size() >= NewBases.size());
  assert(Ptrs.size() >= NewPtrs.size());
  assert(Relocs.size() >= NewRelocs.size());
  Bases = NewBases;
  Ptrs = NewPtrs;
  Relocs = NewRelocs;
  assert(Ptrs.size() == Bases.size());
  assert(Ptrs.size() == Relocs.size());
}

/// Extract call from statepoint, lower it and return pointer to the
/// call node. Also update NodeMap so that getValue(statepoint) will
/// reference lowered call result
static SDNode *
lowerCallFromStatepoint(ImmutableStatepoint ISP, MachineBasicBlock *LandingPad,
                        SelectionDAGBuilder &Builder,
                        SmallVectorImpl<SDValue> &PendingExports) {

  ImmutableCallSite CS(ISP.getCallSite());

  SDValue ActualCallee;

  if (ISP.getNumPatchBytes() > 0) {
    // If we've been asked to emit a nop sequence instead of a call instruction
    // for this statepoint then don't lower the call target, but use a constant
    // `null` instead.  Not lowering the call target lets statepoint clients get
    // away without providing a physical address for the symbolic call target at
    // link time.

    const auto &TLI = Builder.DAG.getTargetLoweringInfo();
    const auto &DL = Builder.DAG.getDataLayout();

    unsigned AS = ISP.getCalledValue()->getType()->getPointerAddressSpace();
    ActualCallee = Builder.DAG.getConstant(0, Builder.getCurSDLoc(),
                                           TLI.getPointerTy(DL, AS));
  } else
    ActualCallee = Builder.getValue(ISP.getCalledValue());

  assert(CS.getCallingConv() != CallingConv::AnyReg &&
         "anyregcc is not supported on statepoints!");

  Type *DefTy = ISP.getActualReturnType();
  bool HasDef = !DefTy->isVoidTy();

  SDValue ReturnValue, CallEndVal;
  std::tie(ReturnValue, CallEndVal) = Builder.lowerCallOperands(
      ISP.getCallSite(), ImmutableStatepoint::CallArgsBeginPos,
      ISP.getNumCallArgs(), ActualCallee, DefTy, LandingPad,
      false /* IsPatchPoint */);

  SDNode *CallEnd = CallEndVal.getNode();

  // Get a call instruction from the call sequence chain.  Tail calls are not
  // allowed.  The following code is essentially reverse engineering X86's
  // LowerCallTo.
  //
  // We are expecting DAG to have the following form:
  //
  // ch = eh_label (only in case of invoke statepoint)
  //   ch, glue = callseq_start ch
  //   ch, glue = X86::Call ch, glue
  //   ch, glue = callseq_end ch, glue
  //   get_return_value ch, glue
  //
  // get_return_value can either be a CopyFromReg to grab the return value from
  // %RAX, or it can be a LOAD to load a value returned by reference via a stack
  // slot.

  if (HasDef && (CallEnd->getOpcode() == ISD::CopyFromReg ||
                 CallEnd->getOpcode() == ISD::LOAD))
    CallEnd = CallEnd->getOperand(0).getNode();

  assert(CallEnd->getOpcode() == ISD::CALLSEQ_END && "expected!");

  if (HasDef) {
    if (CS.isInvoke()) {
      // Result value will be used in different basic block for invokes
      // so we need to export it now. But statepoint call has a different type
      // than the actual call. It means that standard exporting mechanism will
      // create register of the wrong type. So instead we need to create
      // register with correct type and save value into it manually.
      // TODO: To eliminate this problem we can remove gc.result intrinsics
      //       completely and make statepoint call to return a tuple.
      unsigned Reg = Builder.FuncInfo.CreateRegs(ISP.getActualReturnType());
      RegsForValue RFV(
          *Builder.DAG.getContext(), Builder.DAG.getTargetLoweringInfo(),
          Builder.DAG.getDataLayout(), Reg, ISP.getActualReturnType());
      SDValue Chain = Builder.DAG.getEntryNode();

      RFV.getCopyToRegs(ReturnValue, Builder.DAG, Builder.getCurSDLoc(), Chain,
                        nullptr);
      PendingExports.push_back(Chain);
      Builder.FuncInfo.ValueMap[CS.getInstruction()] = Reg;
    } else {
      // The value of the statepoint itself will be the value of call itself.
      // We'll replace the actually call node shortly.  gc_result will grab
      // this value.
      Builder.setValue(CS.getInstruction(), ReturnValue);
    }
  } else {
    // The token value is never used from here on, just generate a poison value
    Builder.setValue(CS.getInstruction(),
                     Builder.DAG.getIntPtrConstant(-1, Builder.getCurSDLoc()));
  }

  return CallEnd->getOperand(0).getNode();
}

/// Callect all gc pointers coming into statepoint intrinsic, clean them up,
/// and return two arrays:
///   Bases - base pointers incoming to this statepoint
///   Ptrs - derived pointers incoming to this statepoint
///   Relocs - the gc_relocate corresponding to each base/ptr pair
/// Elements of this arrays should be in one-to-one correspondence with each
/// other i.e Bases[i], Ptrs[i] are from the same gcrelocate call
static void getIncomingStatepointGCValues(
    SmallVectorImpl<const Value *> &Bases, SmallVectorImpl<const Value *> &Ptrs,
    SmallVectorImpl<const Value *> &Relocs, ImmutableStatepoint StatepointSite,
    SelectionDAGBuilder &Builder) {
  for (GCRelocateOperands relocateOpers : StatepointSite.getRelocates()) {
    Relocs.push_back(relocateOpers.getUnderlyingCallSite().getInstruction());
    Bases.push_back(relocateOpers.getBasePtr());
    Ptrs.push_back(relocateOpers.getDerivedPtr());
  }

  // Remove any redundant llvm::Values which map to the same SDValue as another
  // input.  Also has the effect of removing duplicates in the original
  // llvm::Value input list as well.  This is a useful optimization for
  // reducing the size of the StackMap section.  It has no other impact.
  removeDuplicatesGCPtrs(Bases, Ptrs, Relocs, Builder);

  assert(Bases.size() == Ptrs.size() && Ptrs.size() == Relocs.size());
}

/// Spill a value incoming to the statepoint. It might be either part of
/// vmstate
/// or gcstate. In both cases unconditionally spill it on the stack unless it
/// is a null constant. Return pair with first element being frame index
/// containing saved value and second element with outgoing chain from the
/// emitted store
static std::pair<SDValue, SDValue>
spillIncomingStatepointValue(SDValue Incoming, SDValue Chain,
                             SelectionDAGBuilder &Builder) {
  SDValue Loc = Builder.StatepointLowering.getLocation(Incoming);

  // Emit new store if we didn't do it for this ptr before
  if (!Loc.getNode()) {
    Loc = Builder.StatepointLowering.allocateStackSlot(Incoming.getValueType(),
                                                       Builder);
    assert(isa<FrameIndexSDNode>(Loc));
    int Index = cast<FrameIndexSDNode>(Loc)->getIndex();
    // We use TargetFrameIndex so that isel will not select it into LEA
    Loc = Builder.DAG.getTargetFrameIndex(Index, Incoming.getValueType());

    // TODO: We can create TokenFactor node instead of
    //       chaining stores one after another, this may allow
    //       a bit more optimal scheduling for them
    Chain = Builder.DAG.getStore(Chain, Builder.getCurSDLoc(), Incoming, Loc,
                                 MachinePointerInfo::getFixedStack(Index),
                                 false, false, 0);

    Builder.StatepointLowering.setLocation(Incoming, Loc);
  }

  assert(Loc.getNode());
  return std::make_pair(Loc, Chain);
}

/// Lower a single value incoming to a statepoint node.  This value can be
/// either a deopt value or a gc value, the handling is the same.  We special
/// case constants and allocas, then fall back to spilling if required.
static void lowerIncomingStatepointValue(SDValue Incoming,
                                         SmallVectorImpl<SDValue> &Ops,
                                         SelectionDAGBuilder &Builder) {
  SDValue Chain = Builder.getRoot();

  if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(Incoming)) {
    // If the original value was a constant, make sure it gets recorded as
    // such in the stackmap.  This is required so that the consumer can
    // parse any internal format to the deopt state.  It also handles null
    // pointers and other constant pointers in GC states
    pushStackMapConstant(Ops, Builder, C->getSExtValue());
  } else if (FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Incoming)) {
    // This handles allocas as arguments to the statepoint (this is only
    // really meaningful for a deopt value.  For GC, we'd be trying to
    // relocate the address of the alloca itself?)
    Ops.push_back(Builder.DAG.getTargetFrameIndex(FI->getIndex(),
                                                  Incoming.getValueType()));
  } else {
    // Otherwise, locate a spill slot and explicitly spill it so it
    // can be found by the runtime later.  We currently do not support
    // tracking values through callee saved registers to their eventual
    // spill location.  This would be a useful optimization, but would
    // need to be optional since it requires a lot of complexity on the
    // runtime side which not all would support.
    std::pair<SDValue, SDValue> Res =
        spillIncomingStatepointValue(Incoming, Chain, Builder);
    Ops.push_back(Res.first);
    Chain = Res.second;
  }

  Builder.DAG.setRoot(Chain);
}

/// Lower deopt state and gc pointer arguments of the statepoint.  The actual
/// lowering is described in lowerIncomingStatepointValue.  This function is
/// responsible for lowering everything in the right position and playing some
/// tricks to avoid redundant stack manipulation where possible.  On
/// completion, 'Ops' will contain ready to use operands for machine code
/// statepoint. The chain nodes will have already been created and the DAG root
/// will be set to the last value spilled (if any were).
static void lowerStatepointMetaArgs(SmallVectorImpl<SDValue> &Ops,
                                    ImmutableStatepoint StatepointSite,
                                    SelectionDAGBuilder &Builder) {

  // Lower the deopt and gc arguments for this statepoint.  Layout will
  // be: deopt argument length, deopt arguments.., gc arguments...

  SmallVector<const Value *, 64> Bases, Ptrs, Relocations;
  getIncomingStatepointGCValues(Bases, Ptrs, Relocations, StatepointSite,
                                Builder);

#ifndef NDEBUG
  // Check that each of the gc pointer and bases we've gotten out of the
  // safepoint is something the strategy thinks might be a pointer into the GC
  // heap.  This is basically just here to help catch errors during statepoint
  // insertion. TODO: This should actually be in the Verifier, but we can't get
  // to the GCStrategy from there (yet).
  GCStrategy &S = Builder.GFI->getStrategy();
  for (const Value *V : Bases) {
    auto Opt = S.isGCManagedPointer(V);
    if (Opt.hasValue()) {
      assert(Opt.getValue() &&
             "non gc managed base pointer found in statepoint");
    }
  }
  for (const Value *V : Ptrs) {
    auto Opt = S.isGCManagedPointer(V);
    if (Opt.hasValue()) {
      assert(Opt.getValue() &&
             "non gc managed derived pointer found in statepoint");
    }
  }
  for (const Value *V : Relocations) {
    auto Opt = S.isGCManagedPointer(V);
    if (Opt.hasValue()) {
      assert(Opt.getValue() && "non gc managed pointer relocated");
    }
  }
#endif

  // Before we actually start lowering (and allocating spill slots for values),
  // reserve any stack slots which we judge to be profitable to reuse for a
  // particular value.  This is purely an optimization over the code below and
  // doesn't change semantics at all.  It is important for performance that we
  // reserve slots for both deopt and gc values before lowering either.
  for (const Value *V : StatepointSite.vm_state_args()) {
    reservePreviousStackSlotForValue(V, Builder);
  }
  for (unsigned i = 0; i < Bases.size(); ++i) {
    reservePreviousStackSlotForValue(Bases[i], Builder);
    reservePreviousStackSlotForValue(Ptrs[i], Builder);
  }

  // First, prefix the list with the number of unique values to be
  // lowered.  Note that this is the number of *Values* not the
  // number of SDValues required to lower them.
  const int NumVMSArgs = StatepointSite.getNumTotalVMSArgs();
  pushStackMapConstant(Ops, Builder, NumVMSArgs);

  assert(NumVMSArgs == std::distance(StatepointSite.vm_state_begin(),
                                     StatepointSite.vm_state_end()));

  // The vm state arguments are lowered in an opaque manner.  We do
  // not know what type of values are contained within.  We skip the
  // first one since that happens to be the total number we lowered
  // explicitly just above.  We could have left it in the loop and
  // not done it explicitly, but it's far easier to understand this
  // way.
  for (const Value *V : StatepointSite.vm_state_args()) {
    SDValue Incoming = Builder.getValue(V);
    lowerIncomingStatepointValue(Incoming, Ops, Builder);
  }

  // Finally, go ahead and lower all the gc arguments.  There's no prefixed
  // length for this one.  After lowering, we'll have the base and pointer
  // arrays interwoven with each (lowered) base pointer immediately followed by
  // it's (lowered) derived pointer.  i.e
  // (base[0], ptr[0], base[1], ptr[1], ...)
  for (unsigned i = 0; i < Bases.size(); ++i) {
    const Value *Base = Bases[i];
    lowerIncomingStatepointValue(Builder.getValue(Base), Ops, Builder);

    const Value *Ptr = Ptrs[i];
    lowerIncomingStatepointValue(Builder.getValue(Ptr), Ops, Builder);
  }

  // If there are any explicit spill slots passed to the statepoint, record
  // them, but otherwise do not do anything special.  These are user provided
  // allocas and give control over placement to the consumer.  In this case,
  // it is the contents of the slot which may get updated, not the pointer to
  // the alloca
  for (Value *V : StatepointSite.gc_args()) {
    SDValue Incoming = Builder.getValue(V);
    if (FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Incoming)) {
      // This handles allocas as arguments to the statepoint
      Ops.push_back(Builder.DAG.getTargetFrameIndex(FI->getIndex(),
                                                    Incoming.getValueType()));
    }
  }

  // Record computed locations for all lowered values.
  // This can not be embedded in lowering loops as we need to record *all*
  // values, while previous loops account only values with unique SDValues.
  const Instruction *StatepointInstr =
    StatepointSite.getCallSite().getInstruction();
  FunctionLoweringInfo::StatepointSpilledValueMapTy &SpillMap =
    Builder.FuncInfo.StatepointRelocatedValues[StatepointInstr];

  for (GCRelocateOperands RelocateOpers : StatepointSite.getRelocates()) {
    const Value *V = RelocateOpers.getDerivedPtr();
    SDValue SDV = Builder.getValue(V);
    SDValue Loc = Builder.StatepointLowering.getLocation(SDV);

    if (Loc.getNode()) {
      SpillMap[V] = cast<FrameIndexSDNode>(Loc)->getIndex();
    } else {
      // Record value as visited, but not spilled. This is case for allocas
      // and constants. For this values we can avoid emitting spill load while
      // visiting corresponding gc_relocate.
      // Actually we do not need to record them in this map at all.
      // We do this only to check that we are not relocating any unvisited
      // value.
      SpillMap[V] = None;

      // Default llvm mechanisms for exporting values which are used in
      // different basic blocks does not work for gc relocates.
      // Note that it would be incorrect to teach llvm that all relocates are
      // uses of the corresponding values so that it would automatically
      // export them. Relocates of the spilled values does not use original
      // value.
      if (StatepointSite.getCallSite().isInvoke())
        Builder.ExportFromCurrentBlock(V);
    }
  }
}

void SelectionDAGBuilder::visitStatepoint(const CallInst &CI) {
  // Check some preconditions for sanity
  assert(isStatepoint(&CI) &&
         "function called must be the statepoint function");

  LowerStatepoint(ImmutableStatepoint(&CI));
}

void SelectionDAGBuilder::LowerStatepoint(
    ImmutableStatepoint ISP, MachineBasicBlock *LandingPad /*=nullptr*/) {
  // The basic scheme here is that information about both the original call and
  // the safepoint is encoded in the CallInst.  We create a temporary call and
  // lower it, then reverse engineer the calling sequence.

  NumOfStatepoints++;
  // Clear state
  StatepointLowering.startNewStatepoint(*this);

  ImmutableCallSite CS(ISP.getCallSite());

#ifndef NDEBUG
  // Consistency check. Don't do this for invokes. It would be too
  // expensive to preserve this information across different basic blocks
  if (!CS.isInvoke()) {
    for (const User *U : CS->users()) {
      const CallInst *Call = cast<CallInst>(U);
      if (isGCRelocate(Call))
        StatepointLowering.scheduleRelocCall(*Call);
    }
  }
#endif

#ifndef NDEBUG
  // If this is a malformed statepoint, report it early to simplify debugging.
  // This should catch any IR level mistake that's made when constructing or
  // transforming statepoints.
  ISP.verify();

  // Check that the associated GCStrategy expects to encounter statepoints.
  assert(GFI->getStrategy().useStatepoints() &&
         "GCStrategy does not expect to encounter statepoints");
#endif

  // Lower statepoint vmstate and gcstate arguments
  SmallVector<SDValue, 10> LoweredMetaArgs;
  lowerStatepointMetaArgs(LoweredMetaArgs, ISP, *this);

  // Get call node, we will replace it later with statepoint
  SDNode *CallNode =
      lowerCallFromStatepoint(ISP, LandingPad, *this, PendingExports);

  // Construct the actual GC_TRANSITION_START, STATEPOINT, and GC_TRANSITION_END
  // nodes with all the appropriate arguments and return values.

  // Call Node: Chain, Target, {Args}, RegMask, [Glue]
  SDValue Chain = CallNode->getOperand(0);

  SDValue Glue;
  bool CallHasIncomingGlue = CallNode->getGluedNode();
  if (CallHasIncomingGlue) {
    // Glue is always last operand
    Glue = CallNode->getOperand(CallNode->getNumOperands() - 1);
  }

  // Build the GC_TRANSITION_START node if necessary.
  //
  // The operands to the GC_TRANSITION_{START,END} nodes are laid out in the
  // order in which they appear in the call to the statepoint intrinsic. If
  // any of the operands is a pointer-typed, that operand is immediately
  // followed by a SRCVALUE for the pointer that may be used during lowering
  // (e.g. to form MachinePointerInfo values for loads/stores).
  const bool IsGCTransition =
      (ISP.getFlags() & (uint64_t)StatepointFlags::GCTransition) ==
          (uint64_t)StatepointFlags::GCTransition;
  if (IsGCTransition) {
    SmallVector<SDValue, 8> TSOps;

    // Add chain
    TSOps.push_back(Chain);

    // Add GC transition arguments
    for (const Value *V : ISP.gc_transition_args()) {
      TSOps.push_back(getValue(V));
      if (V->getType()->isPointerTy())
        TSOps.push_back(DAG.getSrcValue(V));
    }

    // Add glue if necessary
    if (CallHasIncomingGlue)
      TSOps.push_back(Glue);

    SDVTList NodeTys = DAG.getVTList(MVT::Other, MVT::Glue);

    SDValue GCTransitionStart =
        DAG.getNode(ISD::GC_TRANSITION_START, getCurSDLoc(), NodeTys, TSOps);

    Chain = GCTransitionStart.getValue(0);
    Glue = GCTransitionStart.getValue(1);
  }

  // TODO: Currently, all of these operands are being marked as read/write in
  // PrologEpilougeInserter.cpp, we should special case the VMState arguments
  // and flags to be read-only.
  SmallVector<SDValue, 40> Ops;

  // Add the <id> and <numBytes> constants.
  Ops.push_back(DAG.getTargetConstant(ISP.getID(), getCurSDLoc(), MVT::i64));
  Ops.push_back(
      DAG.getTargetConstant(ISP.getNumPatchBytes(), getCurSDLoc(), MVT::i32));

  // Calculate and push starting position of vmstate arguments
  // Get number of arguments incoming directly into call node
  unsigned NumCallRegArgs =
      CallNode->getNumOperands() - (CallHasIncomingGlue ? 4 : 3);
  Ops.push_back(DAG.getTargetConstant(NumCallRegArgs, getCurSDLoc(), MVT::i32));

  // Add call target
  SDValue CallTarget = SDValue(CallNode->getOperand(1).getNode(), 0);
  Ops.push_back(CallTarget);

  // Add call arguments
  // Get position of register mask in the call
  SDNode::op_iterator RegMaskIt;
  if (CallHasIncomingGlue)
    RegMaskIt = CallNode->op_end() - 2;
  else
    RegMaskIt = CallNode->op_end() - 1;
  Ops.insert(Ops.end(), CallNode->op_begin() + 2, RegMaskIt);

  // Add a constant argument for the calling convention
  pushStackMapConstant(Ops, *this, CS.getCallingConv());

  // Add a constant argument for the flags
  uint64_t Flags = ISP.getFlags();
  assert(
      ((Flags & ~(uint64_t)StatepointFlags::MaskAll) == 0)
          && "unknown flag used");
  pushStackMapConstant(Ops, *this, Flags);

  // Insert all vmstate and gcstate arguments
  Ops.insert(Ops.end(), LoweredMetaArgs.begin(), LoweredMetaArgs.end());

  // Add register mask from call node
  Ops.push_back(*RegMaskIt);

  // Add chain
  Ops.push_back(Chain);

  // Same for the glue, but we add it only if original call had it
  if (Glue.getNode())
    Ops.push_back(Glue);

  // Compute return values.  Provide a glue output since we consume one as
  // input.  This allows someone else to chain off us as needed.
  SDVTList NodeTys = DAG.getVTList(MVT::Other, MVT::Glue);

  SDNode *StatepointMCNode =
      DAG.getMachineNode(TargetOpcode::STATEPOINT, getCurSDLoc(), NodeTys, Ops);

  SDNode *SinkNode = StatepointMCNode;

  // Build the GC_TRANSITION_END node if necessary.
  //
  // See the comment above regarding GC_TRANSITION_START for the layout of
  // the operands to the GC_TRANSITION_END node.
  if (IsGCTransition) {
    SmallVector<SDValue, 8> TEOps;

    // Add chain
    TEOps.push_back(SDValue(StatepointMCNode, 0));

    // Add GC transition arguments
    for (const Value *V : ISP.gc_transition_args()) {
      TEOps.push_back(getValue(V));
      if (V->getType()->isPointerTy())
        TEOps.push_back(DAG.getSrcValue(V));
    }

    // Add glue
    TEOps.push_back(SDValue(StatepointMCNode, 1));

    SDVTList NodeTys = DAG.getVTList(MVT::Other, MVT::Glue);

    SDValue GCTransitionStart =
        DAG.getNode(ISD::GC_TRANSITION_END, getCurSDLoc(), NodeTys, TEOps);

    SinkNode = GCTransitionStart.getNode();
  }

  // Replace original call
  DAG.ReplaceAllUsesWith(CallNode, SinkNode); // This may update Root
  // Remove original call node
  DAG.DeleteNode(CallNode);

  // DON'T set the root - under the assumption that it's already set past the
  // inserted node we created.

  // TODO: A better future implementation would be to emit a single variable
  // argument, variable return value STATEPOINT node here and then hookup the
  // return value of each gc.relocate to the respective output of the
  // previously emitted STATEPOINT value.  Unfortunately, this doesn't appear
  // to actually be possible today.
}

void SelectionDAGBuilder::visitGCResult(const CallInst &CI) {
  // The result value of the gc_result is simply the result of the actual
  // call.  We've already emitted this, so just grab the value.
  Instruction *I = cast<Instruction>(CI.getArgOperand(0));
  assert(isStatepoint(I) && "first argument must be a statepoint token");

  if (isa<InvokeInst>(I)) {
    // For invokes we should have stored call result in a virtual register.
    // We can not use default getValue() functionality to copy value from this
    // register because statepoint and actuall call return types can be
    // different, and getValue() will use CopyFromReg of the wrong type,
    // which is always i32 in our case.
    PointerType *CalleeType = cast<PointerType>(
        ImmutableStatepoint(I).getCalledValue()->getType());
    Type *RetTy =
        cast<FunctionType>(CalleeType->getElementType())->getReturnType();
    SDValue CopyFromReg = getCopyFromRegs(I, RetTy);

    assert(CopyFromReg.getNode());
    setValue(&CI, CopyFromReg);
  } else {
    setValue(&CI, getValue(I));
  }
}

void SelectionDAGBuilder::visitGCRelocate(const CallInst &CI) {
  GCRelocateOperands RelocateOpers(&CI);

#ifndef NDEBUG
  // Consistency check
  // We skip this check for invoke statepoints. It would be too expensive to
  // preserve validation info through different basic blocks.
  if (!RelocateOpers.isTiedToInvoke()) {
    StatepointLowering.relocCallVisited(CI);
  }
#endif

  const Value *DerivedPtr = RelocateOpers.getDerivedPtr();
  SDValue SD = getValue(DerivedPtr);

  FunctionLoweringInfo::StatepointSpilledValueMapTy &SpillMap =
    FuncInfo.StatepointRelocatedValues[RelocateOpers.getStatepoint()];

  // We should have recorded location for this pointer
  assert(SpillMap.count(DerivedPtr) && "Relocating not lowered gc value");
  Optional<int> DerivedPtrLocation = SpillMap[DerivedPtr];

  // We didn't need to spill these special cases (constants and allocas).
  // See the handling in spillIncomingValueForStatepoint for detail.
  if (!DerivedPtrLocation) {
    setValue(&CI, SD);
    return;
  }

  SDValue SpillSlot = DAG.getTargetFrameIndex(*DerivedPtrLocation,
                                              SD.getValueType());

  // Be conservative: flush all pending loads
  // TODO: Probably we can be less restrictive on this,
  // it may allow more scheduling opportunities.
  SDValue Chain = getRoot();

  SDValue SpillLoad =
    DAG.getLoad(SpillSlot.getValueType(), getCurSDLoc(), Chain, SpillSlot,
                MachinePointerInfo::getFixedStack(*DerivedPtrLocation),
                false, false, false, 0);

  // Again, be conservative, don't emit pending loads
  DAG.setRoot(SpillLoad.getValue(1));

  assert(SpillLoad.getNode());
  setValue(&CI, SpillLoad);
}