bool TargetInstrInfo::getStackSlotRange(const TargetRegisterClass *RC,
unsigned SubIdx, unsigned &Size,
unsigned &Offset,
const MachineFunction &MF) const {
if (!SubIdx) {
Size = RC->getSize();
Offset = 0;
return true;
}
const TargetRegisterInfo *TRI = MF.getSubtarget().getRegisterInfo();
unsigned BitSize = TRI->getSubRegIdxSize(SubIdx);
// Convert bit size to byte size to be consistent with
// MCRegisterClass::getSize().
if (BitSize % 8)
return false;
int BitOffset = TRI->getSubRegIdxOffset(SubIdx);
if (BitOffset < 0 || BitOffset % 8)
return false;
Size = BitSize /= 8;
Offset = (unsigned)BitOffset / 8;
assert(RC->getSize() >= (Offset + Size) && "bad subregister range");
if (!MF.getDataLayout().isLittleEndian()) {
Offset = RC->getSize() - (Offset + Size);
}
return true;
}
void FunctionLoweringInfo::set(const Function &fn, MachineFunction &mf,
SelectionDAG *DAG) {
Fn = &fn;
MF = &mf;
TLI = MF->getSubtarget().getTargetLowering();
RegInfo = &MF->getRegInfo();
const TargetFrameLowering *TFI = MF->getSubtarget().getFrameLowering();
unsigned StackAlign = TFI->getStackAlignment();
// Check whether the function can return without sret-demotion.
SmallVector<ISD::OutputArg, 4> Outs;
GetReturnInfo(Fn->getReturnType(), Fn->getAttributes(), Outs, *TLI,
mf.getDataLayout());
CanLowerReturn = TLI->CanLowerReturn(Fn->getCallingConv(), *MF,
Fn->isVarArg(), Outs, Fn->getContext());
// If this personality uses funclets, we need to do a bit more work.
DenseMap<const AllocaInst *, TinyPtrVector<int *>> CatchObjects;
EHPersonality Personality = classifyEHPersonality(
Fn->hasPersonalityFn() ? Fn->getPersonalityFn() : nullptr);
if (isFuncletEHPersonality(Personality)) {
// Calculate state numbers if we haven't already.
WinEHFuncInfo &EHInfo = *MF->getWinEHFuncInfo();
if (Personality == EHPersonality::MSVC_CXX)
calculateWinCXXEHStateNumbers(&fn, EHInfo);
else if (isAsynchronousEHPersonality(Personality))
calculateSEHStateNumbers(&fn, EHInfo);
else if (Personality == EHPersonality::CoreCLR)
calculateClrEHStateNumbers(&fn, EHInfo);
// Map all BB references in the WinEH data to MBBs.
for (WinEHTryBlockMapEntry &TBME : EHInfo.TryBlockMap) {
for (WinEHHandlerType &H : TBME.HandlerArray) {
if (const AllocaInst *AI = H.CatchObj.Alloca)
CatchObjects.insert({AI, {}}).first->second.push_back(
&H.CatchObj.FrameIndex);
else
H.CatchObj.FrameIndex = INT_MAX;
}
}
}
// Initialize the mapping of values to registers. This is only set up for
// instruction values that are used outside of the block that defines
// them.
for (const BasicBlock &BB : *Fn) {
for (const Instruction &I : BB) {
if (const AllocaInst *AI = dyn_cast<AllocaInst>(&I)) {
Type *Ty = AI->getAllocatedType();
unsigned Align =
std::max((unsigned)MF->getDataLayout().getPrefTypeAlignment(Ty),
AI->getAlignment());
// Static allocas can be folded into the initial stack frame
// adjustment. For targets that don't realign the stack, don't
// do this if there is an extra alignment requirement.
if (AI->isStaticAlloca() &&
(TFI->isStackRealignable() || (Align <= StackAlign))) {
const ConstantInt *CUI = cast<ConstantInt>(AI->getArraySize());
uint64_t TySize = MF->getDataLayout().getTypeAllocSize(Ty);
TySize *= CUI->getZExtValue(); // Get total allocated size.
if (TySize == 0) TySize = 1; // Don't create zero-sized stack objects.
int FrameIndex = INT_MAX;
auto Iter = CatchObjects.find(AI);
if (Iter != CatchObjects.end() && TLI->needsFixedCatchObjects()) {
FrameIndex = MF->getFrameInfo().CreateFixedObject(
TySize, 0, /*Immutable=*/false, /*isAliased=*/true);
MF->getFrameInfo().setObjectAlignment(FrameIndex, Align);
} else {
FrameIndex =
MF->getFrameInfo().CreateStackObject(TySize, Align, false, AI);
}
StaticAllocaMap[AI] = FrameIndex;
// Update the catch handler information.
if (Iter != CatchObjects.end()) {
for (int *CatchObjPtr : Iter->second)
*CatchObjPtr = FrameIndex;
}
} else {
// FIXME: Overaligned static allocas should be grouped into
// a single dynamic allocation instead of using a separate
// stack allocation for each one.
if (Align <= StackAlign)
Align = 0;
// Inform the Frame Information that we have variable-sized objects.
MF->getFrameInfo().CreateVariableSizedObject(Align ? Align : 1, AI);
}
}
// Look for inline asm that clobbers the SP register.
if (isa<CallInst>(I) || isa<InvokeInst>(I)) {
ImmutableCallSite CS(&I);
if (isa<InlineAsm>(CS.getCalledValue())) {
unsigned SP = TLI->getStackPointerRegisterToSaveRestore();
const TargetRegisterInfo *TRI = MF->getSubtarget().getRegisterInfo();
std::vector<TargetLowering::AsmOperandInfo> Ops =
TLI->ParseConstraints(Fn->getParent()->getDataLayout(), TRI, CS);
for (TargetLowering::AsmOperandInfo &Op : Ops) {
if (Op.Type == InlineAsm::isClobber) {
// Clobbers don't have SDValue operands, hence SDValue().
TLI->ComputeConstraintToUse(Op, SDValue(), DAG);
std::pair<unsigned, const TargetRegisterClass *> PhysReg =
TLI->getRegForInlineAsmConstraint(TRI, Op.ConstraintCode,
Op.ConstraintVT);
if (PhysReg.first == SP)
MF->getFrameInfo().setHasOpaqueSPAdjustment(true);
}
}
}
}
// Look for calls to the @llvm.va_start intrinsic. We can omit some
// prologue boilerplate for variadic functions that don't examine their
// arguments.
if (const auto *II = dyn_cast<IntrinsicInst>(&I)) {
if (II->getIntrinsicID() == Intrinsic::vastart)
MF->getFrameInfo().setHasVAStart(true);
}
// If we have a musttail call in a variadic function, we need to ensure we
// forward implicit register parameters.
if (const auto *CI = dyn_cast<CallInst>(&I)) {
if (CI->isMustTailCall() && Fn->isVarArg())
MF->getFrameInfo().setHasMustTailInVarArgFunc(true);
}
// Mark values used outside their block as exported, by allocating
// a virtual register for them.
if (isUsedOutsideOfDefiningBlock(&I))
if (!isa<AllocaInst>(I) || !StaticAllocaMap.count(cast<AllocaInst>(&I)))
InitializeRegForValue(&I);
// Decide the preferred extend type for a value.
PreferredExtendType[&I] = getPreferredExtendForValue(&I);
}
}
// Create an initial MachineBasicBlock for each LLVM BasicBlock in F. This
// also creates the initial PHI MachineInstrs, though none of the input
// operands are populated.
for (const BasicBlock &BB : *Fn) {
// Don't create MachineBasicBlocks for imaginary EH pad blocks. These blocks
// are really data, and no instructions can live here.
if (BB.isEHPad()) {
const Instruction *PadInst = BB.getFirstNonPHI();
// If this is a non-landingpad EH pad, mark this function as using
// funclets.
// FIXME: SEH catchpads do not create funclets, so we could avoid setting
// this in such cases in order to improve frame layout.
if (!isa<LandingPadInst>(PadInst)) {
MF->setHasEHFunclets(true);
MF->getFrameInfo().setHasOpaqueSPAdjustment(true);
}
if (isa<CatchSwitchInst>(PadInst)) {
assert(&*BB.begin() == PadInst &&
"WinEHPrepare failed to remove PHIs from imaginary BBs");
continue;
}
if (isa<FuncletPadInst>(PadInst))
assert(&*BB.begin() == PadInst && "WinEHPrepare failed to demote PHIs");
}
MachineBasicBlock *MBB = mf.CreateMachineBasicBlock(&BB);
MBBMap[&BB] = MBB;
MF->push_back(MBB);
// Transfer the address-taken flag. This is necessary because there could
// be multiple MachineBasicBlocks corresponding to one BasicBlock, and only
// the first one should be marked.
if (BB.hasAddressTaken())
MBB->setHasAddressTaken();
// Mark landing pad blocks.
if (BB.isEHPad())
MBB->setIsEHPad();
// Create Machine PHI nodes for LLVM PHI nodes, lowering them as
// appropriate.
for (BasicBlock::const_iterator I = BB.begin();
const PHINode *PN = dyn_cast<PHINode>(I); ++I) {
if (PN->use_empty()) continue;
// Skip empty types
if (PN->getType()->isEmptyTy())
continue;
DebugLoc DL = PN->getDebugLoc();
unsigned PHIReg = ValueMap[PN];
assert(PHIReg && "PHI node does not have an assigned virtual register!");
SmallVector<EVT, 4> ValueVTs;
ComputeValueVTs(*TLI, MF->getDataLayout(), PN->getType(), ValueVTs);
for (EVT VT : ValueVTs) {
unsigned NumRegisters = TLI->getNumRegisters(Fn->getContext(), VT);
const TargetInstrInfo *TII = MF->getSubtarget().getInstrInfo();
for (unsigned i = 0; i != NumRegisters; ++i)
BuildMI(MBB, DL, TII->get(TargetOpcode::PHI), PHIReg + i);
PHIReg += NumRegisters;
}
}
}
if (!isFuncletEHPersonality(Personality))
return;
WinEHFuncInfo &EHInfo = *MF->getWinEHFuncInfo();
// Map all BB references in the WinEH data to MBBs.
for (WinEHTryBlockMapEntry &TBME : EHInfo.TryBlockMap) {
for (WinEHHandlerType &H : TBME.HandlerArray) {
if (H.Handler)
H.Handler = MBBMap[H.Handler.get<const BasicBlock *>()];
}
}
for (CxxUnwindMapEntry &UME : EHInfo.CxxUnwindMap)
if (UME.Cleanup)
UME.Cleanup = MBBMap[UME.Cleanup.get<const BasicBlock *>()];
for (SEHUnwindMapEntry &UME : EHInfo.SEHUnwindMap) {
const BasicBlock *BB = UME.Handler.get<const BasicBlock *>();
UME.Handler = MBBMap[BB];
}
for (ClrEHUnwindMapEntry &CME : EHInfo.ClrEHUnwindMap) {
const BasicBlock *BB = CME.Handler.get<const BasicBlock *>();
CME.Handler = MBBMap[BB];
}
}
void AArch64FrameLowering::emitPrologue(MachineFunction &MF,
MachineBasicBlock &MBB) const {
MachineBasicBlock::iterator MBBI = MBB.begin();
const MachineFrameInfo *MFI = MF.getFrameInfo();
const Function *Fn = MF.getFunction();
const AArch64Subtarget &Subtarget = MF.getSubtarget<AArch64Subtarget>();
const AArch64RegisterInfo *RegInfo = Subtarget.getRegisterInfo();
const TargetInstrInfo *TII = Subtarget.getInstrInfo();
MachineModuleInfo &MMI = MF.getMMI();
AArch64FunctionInfo *AFI = MF.getInfo<AArch64FunctionInfo>();
bool needsFrameMoves = MMI.hasDebugInfo() || Fn->needsUnwindTableEntry();
bool HasFP = hasFP(MF);
// Debug location must be unknown since the first debug location is used
// to determine the end of the prologue.
DebugLoc DL;
// All calls are tail calls in GHC calling conv, and functions have no
// prologue/epilogue.
if (MF.getFunction()->getCallingConv() == CallingConv::GHC)
return;
int NumBytes = (int)MFI->getStackSize();
if (!AFI->hasStackFrame()) {
assert(!HasFP && "unexpected function without stack frame but with FP");
// All of the stack allocation is for locals.
AFI->setLocalStackSize(NumBytes);
// Label used to tie together the PROLOG_LABEL and the MachineMoves.
MCSymbol *FrameLabel = MMI.getContext().createTempSymbol();
// REDZONE: If the stack size is less than 128 bytes, we don't need
// to actually allocate.
if (NumBytes && !canUseRedZone(MF)) {
emitFrameOffset(MBB, MBBI, DL, AArch64::SP, AArch64::SP, -NumBytes, TII,
MachineInstr::FrameSetup);
// Encode the stack size of the leaf function.
unsigned CFIIndex = MMI.addFrameInst(
MCCFIInstruction::createDefCfaOffset(FrameLabel, -NumBytes));
BuildMI(MBB, MBBI, DL, TII->get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex)
.setMIFlags(MachineInstr::FrameSetup);
} else if (NumBytes) {
++NumRedZoneFunctions;
}
return;
}
// Only set up FP if we actually need to.
int FPOffset = 0;
if (HasFP)
FPOffset = getFPOffsetInPrologue(MBBI);
// Move past the saves of the callee-saved registers.
while (isCSSave(MBBI)) {
++MBBI;
NumBytes -= 16;
}
assert(NumBytes >= 0 && "Negative stack allocation size!?");
if (HasFP) {
// Issue sub fp, sp, FPOffset or
// mov fp,sp when FPOffset is zero.
// Note: All stores of callee-saved registers are marked as "FrameSetup".
// This code marks the instruction(s) that set the FP also.
emitFrameOffset(MBB, MBBI, DL, AArch64::FP, AArch64::SP, FPOffset, TII,
MachineInstr::FrameSetup);
}
// All of the remaining stack allocations are for locals.
AFI->setLocalStackSize(NumBytes);
// Allocate space for the rest of the frame.
const unsigned Alignment = MFI->getMaxAlignment();
const bool NeedsRealignment = RegInfo->needsStackRealignment(MF);
unsigned scratchSPReg = AArch64::SP;
if (NumBytes && NeedsRealignment) {
// Use the first callee-saved register as a scratch register.
scratchSPReg = AArch64::X9;
}
// If we're a leaf function, try using the red zone.
if (NumBytes && !canUseRedZone(MF))
// FIXME: in the case of dynamic re-alignment, NumBytes doesn't have
// the correct value here, as NumBytes also includes padding bytes,
// which shouldn't be counted here.
emitFrameOffset(MBB, MBBI, DL, scratchSPReg, AArch64::SP, -NumBytes, TII,
MachineInstr::FrameSetup);
if (NumBytes && NeedsRealignment) {
const unsigned NrBitsToZero = countTrailingZeros(Alignment);
assert(NrBitsToZero > 1);
assert(scratchSPReg != AArch64::SP);
// SUB X9, SP, NumBytes
// -- X9 is temporary register, so shouldn't contain any live data here,
// -- free to use. This is already produced by emitFrameOffset above.
// AND SP, X9, 0b11111...0000
// The logical immediates have a non-trivial encoding. The following
// formula computes the encoded immediate with all ones but
// NrBitsToZero zero bits as least significant bits.
uint32_t andMaskEncoded =
(1 <<12) // = N
| ((64-NrBitsToZero) << 6) // immr
| ((64-NrBitsToZero-1) << 0) // imms
;
BuildMI(MBB, MBBI, DL, TII->get(AArch64::ANDXri), AArch64::SP)
.addReg(scratchSPReg, RegState::Kill)
.addImm(andMaskEncoded);
}
// If we need a base pointer, set it up here. It's whatever the value of the
// stack pointer is at this point. Any variable size objects will be allocated
// after this, so we can still use the base pointer to reference locals.
//
// FIXME: Clarify FrameSetup flags here.
// Note: Use emitFrameOffset() like above for FP if the FrameSetup flag is
// needed.
if (RegInfo->hasBasePointer(MF)) {
TII->copyPhysReg(MBB, MBBI, DL, RegInfo->getBaseRegister(), AArch64::SP,
false);
}
if (needsFrameMoves) {
const DataLayout &TD = MF.getDataLayout();
const int StackGrowth = -TD.getPointerSize(0);
unsigned FramePtr = RegInfo->getFrameRegister(MF);
// An example of the prologue:
//
// .globl __foo
// .align 2
// __foo:
// Ltmp0:
// .cfi_startproc
// .cfi_personality 155, ___gxx_personality_v0
// Leh_func_begin:
// .cfi_lsda 16, Lexception33
//
// stp xa,bx, [sp, -#offset]!
// ...
// stp x28, x27, [sp, #offset-32]
// stp fp, lr, [sp, #offset-16]
// add fp, sp, #offset - 16
// sub sp, sp, #1360
//
// The Stack:
// +-------------------------------------------+
// 10000 | ........ | ........ | ........ | ........ |
// 10004 | ........ | ........ | ........ | ........ |
// +-------------------------------------------+
// 10008 | ........ | ........ | ........ | ........ |
// 1000c | ........ | ........ | ........ | ........ |
// +===========================================+
// 10010 | X28 Register |
// 10014 | X28 Register |
// +-------------------------------------------+
// 10018 | X27 Register |
// 1001c | X27 Register |
// +===========================================+
// 10020 | Frame Pointer |
// 10024 | Frame Pointer |
// +-------------------------------------------+
// 10028 | Link Register |
// 1002c | Link Register |
// +===========================================+
// 10030 | ........ | ........ | ........ | ........ |
// 10034 | ........ | ........ | ........ | ........ |
// +-------------------------------------------+
// 10038 | ........ | ........ | ........ | ........ |
// 1003c | ........ | ........ | ........ | ........ |
// +-------------------------------------------+
//
// [sp] = 10030 :: >>initial value<<
// sp = 10020 :: stp fp, lr, [sp, #-16]!
// fp = sp == 10020 :: mov fp, sp
// [sp] == 10020 :: stp x28, x27, [sp, #-16]!
// sp == 10010 :: >>final value<<
//
// The frame pointer (w29) points to address 10020. If we use an offset of
// '16' from 'w29', we get the CFI offsets of -8 for w30, -16 for w29, -24
// for w27, and -32 for w28:
//
// Ltmp1:
// .cfi_def_cfa w29, 16
// Ltmp2:
// .cfi_offset w30, -8
// Ltmp3:
// .cfi_offset w29, -16
// Ltmp4:
// .cfi_offset w27, -24
// Ltmp5:
// .cfi_offset w28, -32
if (HasFP) {
// Define the current CFA rule to use the provided FP.
unsigned Reg = RegInfo->getDwarfRegNum(FramePtr, true);
unsigned CFIIndex = MMI.addFrameInst(
MCCFIInstruction::createDefCfa(nullptr, Reg, 2 * StackGrowth));
BuildMI(MBB, MBBI, DL, TII->get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex)
.setMIFlags(MachineInstr::FrameSetup);
// Record the location of the stored LR
unsigned LR = RegInfo->getDwarfRegNum(AArch64::LR, true);
CFIIndex = MMI.addFrameInst(
MCCFIInstruction::createOffset(nullptr, LR, StackGrowth));
BuildMI(MBB, MBBI, DL, TII->get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex)
.setMIFlags(MachineInstr::FrameSetup);
// Record the location of the stored FP
CFIIndex = MMI.addFrameInst(
MCCFIInstruction::createOffset(nullptr, Reg, 2 * StackGrowth));
BuildMI(MBB, MBBI, DL, TII->get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex)
.setMIFlags(MachineInstr::FrameSetup);
} else {
// Encode the stack size of the leaf function.
unsigned CFIIndex = MMI.addFrameInst(
MCCFIInstruction::createDefCfaOffset(nullptr, -MFI->getStackSize()));
BuildMI(MBB, MBBI, DL, TII->get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex)
.setMIFlags(MachineInstr::FrameSetup);
}
// Now emit the moves for whatever callee saved regs we have.
emitCalleeSavedFrameMoves(MBB, MBBI, FramePtr);
}
}
MachinePointerInfo MachinePointerInfo::getUnknownStack(MachineFunction &MF) {
return MachinePointerInfo(MF.getDataLayout().getAllocaAddrSpace());
}
bool GcInfoRecorder::runOnMachineFunction(MachineFunction &MF) {
const Function *F = MF.getFunction();
if (!GcInfo::isGcFunction(F)) {
return false;
}
LLILCJitContext *Context = LLILCJit::TheJit->getLLILCJitContext();
GcFuncInfo *GcFuncInfo = Context->GcInfo->getGcInfo(F);
ValueMap<const AllocaInst *, AllocaInfo> &AllocaMap = GcFuncInfo->AllocaMap;
#if !defined(NDEBUG)
bool EmitLogs = Context->Options->LogGcInfo;
if (EmitLogs) {
dbgs() << "GcInfoRecorder: " << MF.getFunction()->getName() << "\n";
}
#endif // !NDEBUG
const MachineFrameInfo *FrameInfo = MF.getFrameInfo();
int ObjectIndexBegin = FrameInfo->getObjectIndexBegin();
int ObjectIndexEnd = FrameInfo->getObjectIndexEnd();
// FrameInfo reports the allocation offsets in terms of the
// incoming (caller's) StackPointer. Convert these in terms of the
// current (callee's) StackPointer.
uint64_t StackPointerSize = MF.getDataLayout().getPointerSize();
uint64_t SpOffset = FrameInfo->getStackSize() + StackPointerSize;
for (int Idx = ObjectIndexBegin; Idx < ObjectIndexEnd; Idx++) {
const AllocaInst *Alloca = FrameInfo->getObjectAllocation(Idx);
if (Alloca == nullptr) {
continue;
}
if (GcFuncInfo->hasRecord(Alloca)) {
int32_t SlotOffset = SpOffset + FrameInfo->getObjectOffset(Idx);
AllocaInfo &AllocaInfo = AllocaMap[Alloca];
assert(SlotOffset >= 0);
assert(AllocaInfo.Offset == GcInfo::InvalidPointerOffset &&
"Two slots for the same alloca!");
AllocaInfo.Offset = SlotOffset;
#if !defined(NDEBUG)
if (AllocaInfo.isGcAggregate()) {
assert(isa<StructType>(Alloca->getAllocatedType()) &&
"Unhandled Type of GcAggregate");
} else if (AllocaInfo.isGcPointer()) {
assert(GcInfo::isGcPointer(Alloca->getAllocatedType()));
} else {
assert(!GcInfo::isGcAllocation(Alloca));
}
if (EmitLogs) {
dbgs() << AllocaInfo.getAllocTypeString() << " @ sp+" << SlotOffset
<< " [";
Alloca->printAsOperand(dbgs(), false);
dbgs() << "]\n";
}
#endif // !NDEBUG
} else {
// All GC-aggregate Allocas must be registered before this phase.
// This ensures that the allocations are properly initialized,
// and marked as frame-escaped if necessary.
//
// TODO: The following check should be:
// assert(!GcInfo::isGcAllocation(Alloca) &&
// "Gc Allocation Record missing");
#if !defined(NDEBUG)
if (GcInfo::isGcAllocation(Alloca)) {
// However, some Gc-pointer slots created by WinEHPrepare phase
// go unrecorded currently because the AllocaInfos are recorded
// by the Reader post-pass.
// TODO: Report the Spill slots created by WinEHPrepare
// https://github.com/dotnet/llilc/issues/901
assert(Alloca->hasName());
assert(Alloca->getName().find(".wineh.spillslot") != StringRef::npos);
// WinEH shouldn't spill GC-aggregates
assert(!GcInfo::isGcAggregate(Alloca->getAllocatedType()));
// The unreported slots are live across safepoints in the
// EH path, so the execution is correct unless we take the
// exception path.
assert(!(Context->Options->ExecuteHandlers &&
Context->Options->DoInsertStatepoints) &&
"Untested: Use at your own risk");
}
#endif // !NDEBUG
}
}
return false; // success
}
void FunctionLoweringInfo::set(const Function &fn, MachineFunction &mf,
SelectionDAG *DAG) {
Fn = &fn;
MF = &mf;
TLI = MF->getSubtarget().getTargetLowering();
RegInfo = &MF->getRegInfo();
MachineModuleInfo &MMI = MF->getMMI();
// Check whether the function can return without sret-demotion.
SmallVector<ISD::OutputArg, 4> Outs;
GetReturnInfo(Fn->getReturnType(), Fn->getAttributes(), Outs, *TLI,
mf.getDataLayout());
CanLowerReturn = TLI->CanLowerReturn(Fn->getCallingConv(), *MF,
Fn->isVarArg(), Outs, Fn->getContext());
// Initialize the mapping of values to registers. This is only set up for
// instruction values that are used outside of the block that defines
// them.
Function::const_iterator BB = Fn->begin(), EB = Fn->end();
for (; BB != EB; ++BB)
for (BasicBlock::const_iterator I = BB->begin(), E = BB->end();
I != E; ++I) {
if (const AllocaInst *AI = dyn_cast<AllocaInst>(I)) {
// Static allocas can be folded into the initial stack frame adjustment.
if (AI->isStaticAlloca()) {
const ConstantInt *CUI = cast<ConstantInt>(AI->getArraySize());
Type *Ty = AI->getAllocatedType();
uint64_t TySize = MF->getDataLayout().getTypeAllocSize(Ty);
unsigned Align =
std::max((unsigned)MF->getDataLayout().getPrefTypeAlignment(Ty),
AI->getAlignment());
TySize *= CUI->getZExtValue(); // Get total allocated size.
if (TySize == 0) TySize = 1; // Don't create zero-sized stack objects.
StaticAllocaMap[AI] =
MF->getFrameInfo()->CreateStackObject(TySize, Align, false, AI);
} else {
unsigned Align =
std::max((unsigned)MF->getDataLayout().getPrefTypeAlignment(
AI->getAllocatedType()),
AI->getAlignment());
unsigned StackAlign =
MF->getSubtarget().getFrameLowering()->getStackAlignment();
if (Align <= StackAlign)
Align = 0;
// Inform the Frame Information that we have variable-sized objects.
MF->getFrameInfo()->CreateVariableSizedObject(Align ? Align : 1, AI);
}
}
// Look for inline asm that clobbers the SP register.
if (isa<CallInst>(I) || isa<InvokeInst>(I)) {
ImmutableCallSite CS(I);
if (isa<InlineAsm>(CS.getCalledValue())) {
unsigned SP = TLI->getStackPointerRegisterToSaveRestore();
const TargetRegisterInfo *TRI = MF->getSubtarget().getRegisterInfo();
std::vector<TargetLowering::AsmOperandInfo> Ops =
TLI->ParseConstraints(Fn->getParent()->getDataLayout(), TRI, CS);
for (size_t I = 0, E = Ops.size(); I != E; ++I) {
TargetLowering::AsmOperandInfo &Op = Ops[I];
if (Op.Type == InlineAsm::isClobber) {
// Clobbers don't have SDValue operands, hence SDValue().
TLI->ComputeConstraintToUse(Op, SDValue(), DAG);
std::pair<unsigned, const TargetRegisterClass *> PhysReg =
TLI->getRegForInlineAsmConstraint(TRI, Op.ConstraintCode,
Op.ConstraintVT);
if (PhysReg.first == SP)
MF->getFrameInfo()->setHasOpaqueSPAdjustment(true);
}
}
}
}
// Look for calls to the @llvm.va_start intrinsic. We can omit some
// prologue boilerplate for variadic functions that don't examine their
// arguments.
if (const auto *II = dyn_cast<IntrinsicInst>(I)) {
if (II->getIntrinsicID() == Intrinsic::vastart)
MF->getFrameInfo()->setHasVAStart(true);
}
// If we have a musttail call in a variadic funciton, we need to ensure we
// forward implicit register parameters.
if (const auto *CI = dyn_cast<CallInst>(I)) {
if (CI->isMustTailCall() && Fn->isVarArg())
MF->getFrameInfo()->setHasMustTailInVarArgFunc(true);
}
// Mark values used outside their block as exported, by allocating
// a virtual register for them.
if (isUsedOutsideOfDefiningBlock(I))
if (!isa<AllocaInst>(I) ||
!StaticAllocaMap.count(cast<AllocaInst>(I)))
InitializeRegForValue(I);
// Collect llvm.dbg.declare information. This is done now instead of
// during the initial isel pass through the IR so that it is done
// in a predictable order.
if (const DbgDeclareInst *DI = dyn_cast<DbgDeclareInst>(I)) {
assert(DI->getVariable() && "Missing variable");
assert(DI->getDebugLoc() && "Missing location");
if (MMI.hasDebugInfo()) {
// Don't handle byval struct arguments or VLAs, for example.
// Non-byval arguments are handled here (they refer to the stack
// temporary alloca at this point).
const Value *Address = DI->getAddress();
if (Address) {
if (const BitCastInst *BCI = dyn_cast<BitCastInst>(Address))
Address = BCI->getOperand(0);
if (const AllocaInst *AI = dyn_cast<AllocaInst>(Address)) {
DenseMap<const AllocaInst *, int>::iterator SI =
StaticAllocaMap.find(AI);
if (SI != StaticAllocaMap.end()) { // Check for VLAs.
int FI = SI->second;
MMI.setVariableDbgInfo(DI->getVariable(), DI->getExpression(),
FI, DI->getDebugLoc());
}
}
}
}
}
// Decide the preferred extend type for a value.
PreferredExtendType[I] = getPreferredExtendForValue(I);
}
// Create an initial MachineBasicBlock for each LLVM BasicBlock in F. This
// also creates the initial PHI MachineInstrs, though none of the input
// operands are populated.
for (BB = Fn->begin(); BB != EB; ++BB) {
// Don't create MachineBasicBlocks for imaginary EH pad blocks. These blocks
// are really data, and no instructions can live here.
if (BB->isEHPad()) {
const Instruction *I = BB->getFirstNonPHI();
// FIXME: Don't mark SEH functions without __finally blocks as having
// funclets.
if (!isa<LandingPadInst>(I))
MMI.setHasEHFunclets(true);
if (isa<CatchEndPadInst>(I) || isa<CleanupEndPadInst>(I)) {
assert(&*BB->begin() == I &&
"WinEHPrepare failed to remove PHIs from imaginary BBs");
continue;
}
if (isa<CatchPadInst>(I) || isa<CleanupPadInst>(I))
assert(&*BB->begin() == I && "WinEHPrepare failed to demote PHIs");
}
MachineBasicBlock *MBB = mf.CreateMachineBasicBlock(BB);
MBBMap[BB] = MBB;
MF->push_back(MBB);
// Transfer the address-taken flag. This is necessary because there could
// be multiple MachineBasicBlocks corresponding to one BasicBlock, and only
// the first one should be marked.
if (BB->hasAddressTaken())
MBB->setHasAddressTaken();
// Create Machine PHI nodes for LLVM PHI nodes, lowering them as
// appropriate.
for (BasicBlock::const_iterator I = BB->begin();
const PHINode *PN = dyn_cast<PHINode>(I); ++I) {
if (PN->use_empty()) continue;
// Skip empty types
if (PN->getType()->isEmptyTy())
continue;
DebugLoc DL = PN->getDebugLoc();
unsigned PHIReg = ValueMap[PN];
assert(PHIReg && "PHI node does not have an assigned virtual register!");
SmallVector<EVT, 4> ValueVTs;
ComputeValueVTs(*TLI, MF->getDataLayout(), PN->getType(), ValueVTs);
for (unsigned vti = 0, vte = ValueVTs.size(); vti != vte; ++vti) {
EVT VT = ValueVTs[vti];
unsigned NumRegisters = TLI->getNumRegisters(Fn->getContext(), VT);
const TargetInstrInfo *TII = MF->getSubtarget().getInstrInfo();
for (unsigned i = 0; i != NumRegisters; ++i)
BuildMI(MBB, DL, TII->get(TargetOpcode::PHI), PHIReg + i);
PHIReg += NumRegisters;
}
}
}
// Mark landing pad blocks.
SmallVector<const LandingPadInst *, 4> LPads;
for (BB = Fn->begin(); BB != EB; ++BB) {
const Instruction *FNP = BB->getFirstNonPHI();
if (BB->isEHPad() && MBBMap.count(BB))
MBBMap[BB]->setIsEHPad();
if (const auto *LPI = dyn_cast<LandingPadInst>(FNP))
LPads.push_back(LPI);
}
// If this personality uses funclets, we need to do a bit more work.
if (!Fn->hasPersonalityFn())
return;
EHPersonality Personality = classifyEHPersonality(Fn->getPersonalityFn());
if (!isFuncletEHPersonality(Personality))
return;
if (Personality == EHPersonality::MSVC_Win64SEH ||
Personality == EHPersonality::MSVC_X86SEH) {
addSEHHandlersForLPads(LPads);
}
// Calculate state numbers if we haven't already.
WinEHFuncInfo &EHInfo = MMI.getWinEHFuncInfo(&fn);
const Function *WinEHParentFn = MMI.getWinEHParent(&fn);
if (Personality == EHPersonality::MSVC_CXX)
calculateWinCXXEHStateNumbers(WinEHParentFn, EHInfo);
else if (isAsynchronousEHPersonality(Personality))
calculateSEHStateNumbers(WinEHParentFn, EHInfo);
else if (Personality == EHPersonality::CoreCLR)
calculateClrEHStateNumbers(WinEHParentFn, EHInfo);
calculateCatchReturnSuccessorColors(WinEHParentFn, EHInfo);
// Map all BB references in the WinEH data to MBBs.
for (WinEHTryBlockMapEntry &TBME : EHInfo.TryBlockMap) {
for (WinEHHandlerType &H : TBME.HandlerArray) {
if (H.CatchObjRecoverIdx == -2 && H.CatchObj.Alloca) {
assert(StaticAllocaMap.count(H.CatchObj.Alloca));
H.CatchObj.FrameIndex = StaticAllocaMap[H.CatchObj.Alloca];
} else {
H.CatchObj.FrameIndex = INT_MAX;
}
if (const auto *BB = dyn_cast<BasicBlock>(H.Handler.get<const Value *>()))
H.Handler = MBBMap[BB];
}
}
for (WinEHUnwindMapEntry &UME : EHInfo.UnwindMap)
if (UME.Cleanup)
if (const auto *BB = dyn_cast<BasicBlock>(UME.Cleanup.get<const Value *>()))
UME.Cleanup = MBBMap[BB];
for (SEHUnwindMapEntry &UME : EHInfo.SEHUnwindMap) {
const BasicBlock *BB = UME.Handler.get<const BasicBlock *>();
UME.Handler = MBBMap[BB];
}
for (ClrEHUnwindMapEntry &CME : EHInfo.ClrEHUnwindMap) {
const BasicBlock *BB = CME.Handler.get<const BasicBlock *>();
CME.Handler = MBBMap[BB];
}
// If there's an explicit EH registration node on the stack, record its
// frame index.
if (EHInfo.EHRegNode && EHInfo.EHRegNode->getParent()->getParent() == Fn) {
assert(StaticAllocaMap.count(EHInfo.EHRegNode));
EHInfo.EHRegNodeFrameIndex = StaticAllocaMap[EHInfo.EHRegNode];
}
// Copy the state numbers to LandingPadInfo for the current function, which
// could be a handler or the parent. This should happen for 32-bit SEH and
// C++ EH.
if (Personality == EHPersonality::MSVC_CXX ||
Personality == EHPersonality::MSVC_X86SEH) {
for (const LandingPadInst *LP : LPads) {
MachineBasicBlock *LPadMBB = MBBMap[LP->getParent()];
MMI.addWinEHState(LPadMBB, EHInfo.EHPadStateMap[LP]);
}
}
}
