// SurveyFunction - This performs the initial survey of the specified function,
// checking out whether or not it uses any of its incoming arguments or whether
// any callers use the return value. This fills in the LiveValues set and Uses
// map.
//
// We consider arguments of non-internal functions to be intrinsically alive as
// well as arguments to functions which have their "address taken".
//
void DAE::SurveyFunction(const Function &F) {
unsigned RetCount = NumRetVals(&F);
// Assume all return values are dead
typedef SmallVector<Liveness, 5> RetVals;
RetVals RetValLiveness(RetCount, MaybeLive);
typedef SmallVector<UseVector, 5> RetUses;
// These vectors map each return value to the uses that make it MaybeLive, so
// we can add those to the Uses map if the return value really turns out to be
// MaybeLive. Initialized to a list of RetCount empty lists.
RetUses MaybeLiveRetUses(RetCount);
for (Function::const_iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
if (const ReturnInst *RI = dyn_cast<ReturnInst>(BB->getTerminator()))
if (RI->getNumOperands() != 0 && RI->getOperand(0)->getType()
!= F.getFunctionType()->getReturnType()) {
// We don't support old style multiple return values.
MarkLive(F);
return;
}
if (!F.hasLocalLinkage() && (!ShouldHackArguments() || F.isIntrinsic())) {
MarkLive(F);
return;
}
DEBUG(dbgs() << "DAE - Inspecting callers for fn: " << F.getName() << "\n");
// Keep track of the number of live retvals, so we can skip checks once all
// of them turn out to be live.
unsigned NumLiveRetVals = 0;
Type *STy = dyn_cast<StructType>(F.getReturnType());
// Loop all uses of the function.
for (Value::const_use_iterator I = F.use_begin(), E = F.use_end();
I != E; ++I) {
// If the function is PASSED IN as an argument, its address has been
// taken.
ImmutableCallSite CS(*I);
if (!CS || !CS.isCallee(I)) {
MarkLive(F);
return;
}
// If this use is anything other than a call site, the function is alive.
const Instruction *TheCall = CS.getInstruction();
if (!TheCall) { // Not a direct call site?
MarkLive(F);
return;
}
// If we end up here, we are looking at a direct call to our function.
// Now, check how our return value(s) is/are used in this caller. Don't
// bother checking return values if all of them are live already.
if (NumLiveRetVals != RetCount) {
if (STy) {
// Check all uses of the return value.
for (Value::const_use_iterator I = TheCall->use_begin(),
E = TheCall->use_end(); I != E; ++I) {
const ExtractValueInst *Ext = dyn_cast<ExtractValueInst>(*I);
if (Ext && Ext->hasIndices()) {
// This use uses a part of our return value, survey the uses of
// that part and store the results for this index only.
unsigned Idx = *Ext->idx_begin();
if (RetValLiveness[Idx] != Live) {
RetValLiveness[Idx] = SurveyUses(Ext, MaybeLiveRetUses[Idx]);
if (RetValLiveness[Idx] == Live)
NumLiveRetVals++;
}
} else {
// Used by something else than extractvalue. Mark all return
// values as live.
for (unsigned i = 0; i != RetCount; ++i )
RetValLiveness[i] = Live;
NumLiveRetVals = RetCount;
break;
}
}
} else {
// Single return value
RetValLiveness[0] = SurveyUses(TheCall, MaybeLiveRetUses[0]);
if (RetValLiveness[0] == Live)
NumLiveRetVals = RetCount;
}
}
}
// Now we've inspected all callers, record the liveness of our return values.
for (unsigned i = 0; i != RetCount; ++i)
MarkValue(CreateRet(&F, i), RetValLiveness[i], MaybeLiveRetUses[i]);
DEBUG(dbgs() << "DAE - Inspecting args for fn: " << F.getName() << "\n");
// Now, check all of our arguments.
unsigned i = 0;
UseVector MaybeLiveArgUses;
for (Function::const_arg_iterator AI = F.arg_begin(),
E = F.arg_end(); AI != E; ++AI, ++i) {
// See what the effect of this use is (recording any uses that cause
// MaybeLive in MaybeLiveArgUses).
Liveness Result = SurveyUses(AI, MaybeLiveArgUses);
// Mark the result.
MarkValue(CreateArg(&F, i), Result, MaybeLiveArgUses);
// Clear the vector again for the next iteration.
MaybeLiveArgUses.clear();
}
}
// SurveyFunction - This performs the initial survey of the specified function,
// checking out whether or not it uses any of its incoming arguments or whether
// any callers use the return value. This fills in the LiveValues set and Uses
// map.
//
// We consider arguments of non-internal functions to be intrinsically alive as
// well as arguments to functions which have their "address taken".
//
void DAE::SurveyFunction(const Function &F) {
// Functions with inalloca parameters are expecting args in a particular
// register and memory layout.
if (F.getAttributes().hasAttrSomewhere(Attribute::InAlloca)) {
MarkLive(F);
return;
}
// Don't touch naked functions. The assembly might be using an argument, or
// otherwise rely on the frame layout in a way that this analysis will not
// see.
if (F.hasFnAttribute(Attribute::Naked)) {
MarkLive(F);
return;
}
unsigned RetCount = NumRetVals(&F);
// Assume all return values are dead
typedef SmallVector<Liveness, 5> RetVals;
RetVals RetValLiveness(RetCount, MaybeLive);
typedef SmallVector<UseVector, 5> RetUses;
// These vectors map each return value to the uses that make it MaybeLive, so
// we can add those to the Uses map if the return value really turns out to be
// MaybeLive. Initialized to a list of RetCount empty lists.
RetUses MaybeLiveRetUses(RetCount);
for (Function::const_iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
if (const ReturnInst *RI = dyn_cast<ReturnInst>(BB->getTerminator()))
if (RI->getNumOperands() != 0 && RI->getOperand(0)->getType()
!= F.getFunctionType()->getReturnType()) {
// We don't support old style multiple return values.
MarkLive(F);
return;
}
if (!F.hasLocalLinkage() && (!ShouldHackArguments() || F.isIntrinsic())) {
MarkLive(F);
return;
}
DEBUG(dbgs() << "DAE - Inspecting callers for fn: " << F.getName() << "\n");
// Keep track of the number of live retvals, so we can skip checks once all
// of them turn out to be live.
unsigned NumLiveRetVals = 0;
// Loop all uses of the function.
for (const Use &U : F.uses()) {
// If the function is PASSED IN as an argument, its address has been
// taken.
ImmutableCallSite CS(U.getUser());
if (!CS || !CS.isCallee(&U)) {
MarkLive(F);
return;
}
// If this use is anything other than a call site, the function is alive.
const Instruction *TheCall = CS.getInstruction();
if (!TheCall) { // Not a direct call site?
MarkLive(F);
return;
}
// If we end up here, we are looking at a direct call to our function.
// Now, check how our return value(s) is/are used in this caller. Don't
// bother checking return values if all of them are live already.
if (NumLiveRetVals == RetCount)
continue;
// Check all uses of the return value.
for (const Use &U : TheCall->uses()) {
if (ExtractValueInst *Ext = dyn_cast<ExtractValueInst>(U.getUser())) {
// This use uses a part of our return value, survey the uses of
// that part and store the results for this index only.
unsigned Idx = *Ext->idx_begin();
if (RetValLiveness[Idx] != Live) {
RetValLiveness[Idx] = SurveyUses(Ext, MaybeLiveRetUses[Idx]);
if (RetValLiveness[Idx] == Live)
NumLiveRetVals++;
}
} else {
// Used by something else than extractvalue. Survey, but assume that the
// result applies to all sub-values.
UseVector MaybeLiveAggregateUses;
if (SurveyUse(&U, MaybeLiveAggregateUses) == Live) {
NumLiveRetVals = RetCount;
RetValLiveness.assign(RetCount, Live);
break;
} else {
for (unsigned i = 0; i != RetCount; ++i) {
if (RetValLiveness[i] != Live)
MaybeLiveRetUses[i].append(MaybeLiveAggregateUses.begin(),
MaybeLiveAggregateUses.end());
}
}
}
}
}
// Now we've inspected all callers, record the liveness of our return values.
for (unsigned i = 0; i != RetCount; ++i)
MarkValue(CreateRet(&F, i), RetValLiveness[i], MaybeLiveRetUses[i]);
DEBUG(dbgs() << "DAE - Inspecting args for fn: " << F.getName() << "\n");
// Now, check all of our arguments.
unsigned i = 0;
UseVector MaybeLiveArgUses;
for (Function::const_arg_iterator AI = F.arg_begin(),
E = F.arg_end(); AI != E; ++AI, ++i) {
Liveness Result;
if (F.getFunctionType()->isVarArg()) {
// Variadic functions will already have a va_arg function expanded inside
// them, making them potentially very sensitive to ABI changes resulting
// from removing arguments entirely, so don't. For example AArch64 handles
// register and stack HFAs very differently, and this is reflected in the
// IR which has already been generated.
Result = Live;
} else {
// See what the effect of this use is (recording any uses that cause
// MaybeLive in MaybeLiveArgUses).
Result = SurveyUses(&*AI, MaybeLiveArgUses);
}
// Mark the result.
MarkValue(CreateArg(&F, i), Result, MaybeLiveArgUses);
// Clear the vector again for the next iteration.
MaybeLiveArgUses.clear();
}
}
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);
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 = TLI->getDataLayout()->getTypeAllocSize(Ty);
unsigned Align =
std::max((unsigned)TLI->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)TLI->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(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()->setHasInlineAsmWithSPAdjust(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) {
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, 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) {
if (const auto *Invoke = dyn_cast<InvokeInst>(BB->getTerminator()))
MBBMap[Invoke->getSuccessor(1)]->setIsLandingPad();
if (BB->isLandingPad())
LPads.push_back(BB->getLandingPadInst());
}
// If this is an MSVC EH personality, we need to do a bit more work.
EHPersonality Personality = EHPersonality::Unknown;
if (!LPads.empty())
Personality = classifyEHPersonality(LPads.back()->getPersonalityFn());
if (!isMSVCEHPersonality(Personality))
return;
if (Personality == EHPersonality::MSVC_Win64SEH ||
Personality == EHPersonality::MSVC_X86SEH) {
addSEHHandlersForLPads(LPads);
}
WinEHFuncInfo &EHInfo = MMI.getWinEHFuncInfo(&fn);
if (Personality == EHPersonality::MSVC_CXX) {
const Function *WinEHParentFn = MMI.getWinEHParent(&fn);
calculateWinCXXEHStateNumbers(WinEHParentFn, EHInfo);
}
// 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.LandingPadStateMap[LP]);
}
}
}
void FunctionLoweringInfo::set(const Function &fn, MachineFunction &mf,
SelectionDAG *DAG) {
const TargetLowering *TLI = TM.getTargetLowering();
Fn = &fn;
MF = &mf;
RegInfo = &MF->getRegInfo();
// Check whether the function can return without sret-demotion.
SmallVector<ISD::OutputArg, 4> Outs;
GetReturnInfo(Fn->getReturnType(), Fn->getAttributes(), Outs, *TLI);
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 (BasicBlock::const_iterator I = BB->begin(), E = BB->end(); I != E; ++I)
if (const AllocaInst *AI = dyn_cast<AllocaInst>(I)) {
// Don't fold inalloca allocas or other dynamic allocas into the initial
// stack frame allocation, even if they are in the entry block.
if (!AI->isStaticAlloca())
continue;
if (const ConstantInt *CUI = dyn_cast<ConstantInt>(AI->getArraySize())) {
Type *Ty = AI->getAllocatedType();
uint64_t TySize = TLI->getDataLayout()->getTypeAllocSize(Ty);
unsigned Align =
std::max((unsigned)TLI->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);
}
}
for (; BB != EB; ++BB)
for (BasicBlock::const_iterator I = BB->begin(), E = BB->end();
I != E; ++I) {
// Look for dynamic allocas.
if (const AllocaInst *AI = dyn_cast<AllocaInst>(I)) {
if (!AI->isStaticAlloca()) {
unsigned Align = std::max(
(unsigned)TLI->getDataLayout()->getPrefTypeAlignment(
AI->getAllocatedType()),
AI->getAlignment());
unsigned StackAlign = TM.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();
std::vector<TargetLowering::AsmOperandInfo> Ops =
TLI->ParseConstraints(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(Op.ConstraintCode,
Op.ConstraintVT);
if (PhysReg.first == SP)
MF->getFrameInfo()->setHasInlineAsmWithSPAdjust(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)) {
MachineModuleInfo &MMI = MF->getMMI();
DIVariable DIVar(DI->getVariable());
assert((!DIVar || DIVar.isVariable()) &&
"Variable in DbgDeclareInst should be either null or a DIVariable.");
if (MMI.hasDebugInfo() &&
DIVar &&
!DI->getDebugLoc().isUnknown()) {
// 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(),
FI, DI->getDebugLoc());
}
}
}
}
}
}
// 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) {
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, 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->getTarget().getInstrInfo();
for (unsigned i = 0; i != NumRegisters; ++i)
BuildMI(MBB, DL, TII->get(TargetOpcode::PHI), PHIReg + i);
PHIReg += NumRegisters;
}
}
}
// Mark landing pad blocks.
for (BB = Fn->begin(); BB != EB; ++BB)
if (const InvokeInst *Invoke = dyn_cast<InvokeInst>(BB->getTerminator()))
MBBMap[Invoke->getSuccessor(1)]->setIsLandingPad();
}
/// InlineFunction - This function inlines the called function into the basic
/// block of the caller. This returns false if it is not possible to inline
/// this call. The program is still in a well defined state if this occurs
/// though.
///
/// Note that this only does one level of inlining. For example, if the
/// instruction 'call B' is inlined, and 'B' calls 'C', then the call to 'C' now
/// exists in the instruction stream. Similarly this will inline a recursive
/// function by one level.
bool llvm::InlineFunction(CallSite CS, InlineFunctionInfo &IFI,
bool InsertLifetime) {
Instruction *TheCall = CS.getInstruction();
assert(TheCall->getParent() && TheCall->getParent()->getParent() &&
"Instruction not in function!");
// If IFI has any state in it, zap it before we fill it in.
IFI.reset();
const Function *CalledFunc = CS.getCalledFunction();
if (CalledFunc == 0 || // Can't inline external function or indirect
CalledFunc->isDeclaration() || // call, or call to a vararg function!
CalledFunc->getFunctionType()->isVarArg()) return false;
// If the call to the callee is not a tail call, we must clear the 'tail'
// flags on any calls that we inline.
bool MustClearTailCallFlags =
!(isa<CallInst>(TheCall) && cast<CallInst>(TheCall)->isTailCall());
// If the call to the callee cannot throw, set the 'nounwind' flag on any
// calls that we inline.
bool MarkNoUnwind = CS.doesNotThrow();
BasicBlock *OrigBB = TheCall->getParent();
Function *Caller = OrigBB->getParent();
// GC poses two hazards to inlining, which only occur when the callee has GC:
// 1. If the caller has no GC, then the callee's GC must be propagated to the
// caller.
// 2. If the caller has a differing GC, it is invalid to inline.
if (CalledFunc->hasGC()) {
if (!Caller->hasGC())
Caller->setGC(CalledFunc->getGC());
else if (CalledFunc->getGC() != Caller->getGC())
return false;
}
// Get the personality function from the callee if it contains a landing pad.
Value *CalleePersonality = 0;
for (Function::const_iterator I = CalledFunc->begin(), E = CalledFunc->end();
I != E; ++I)
if (const InvokeInst *II = dyn_cast<InvokeInst>(I->getTerminator())) {
const BasicBlock *BB = II->getUnwindDest();
const LandingPadInst *LP = BB->getLandingPadInst();
CalleePersonality = LP->getPersonalityFn();
break;
}
// Find the personality function used by the landing pads of the caller. If it
// exists, then check to see that it matches the personality function used in
// the callee.
if (CalleePersonality) {
for (Function::const_iterator I = Caller->begin(), E = Caller->end();
I != E; ++I)
if (const InvokeInst *II = dyn_cast<InvokeInst>(I->getTerminator())) {
const BasicBlock *BB = II->getUnwindDest();
const LandingPadInst *LP = BB->getLandingPadInst();
// If the personality functions match, then we can perform the
// inlining. Otherwise, we can't inline.
// TODO: This isn't 100% true. Some personality functions are proper
// supersets of others and can be used in place of the other.
if (LP->getPersonalityFn() != CalleePersonality)
return false;
break;
}
}
// Get an iterator to the last basic block in the function, which will have
// the new function inlined after it.
Function::iterator LastBlock = &Caller->back();
// Make sure to capture all of the return instructions from the cloned
// function.
SmallVector<ReturnInst*, 8> Returns;
ClonedCodeInfo InlinedFunctionInfo;
Function::iterator FirstNewBlock;
{ // Scope to destroy VMap after cloning.
ValueToValueMapTy VMap;
assert(CalledFunc->arg_size() == CS.arg_size() &&
"No varargs calls can be inlined!");
// Calculate the vector of arguments to pass into the function cloner, which
// matches up the formal to the actual argument values.
CallSite::arg_iterator AI = CS.arg_begin();
unsigned ArgNo = 0;
for (Function::const_arg_iterator I = CalledFunc->arg_begin(),
E = CalledFunc->arg_end(); I != E; ++I, ++AI, ++ArgNo) {
Value *ActualArg = *AI;
const Argument *Arg = I;
// When byval arguments actually inlined, we need to make the copy implied
// by them explicit. However, we don't do this if the callee is readonly
// or readnone, because the copy would be unneeded: the callee doesn't
// modify the struct.
if (CS.isByValArgument(ArgNo)) {
ActualArg = HandleByValArgument(ActualArg, Arg, TheCall, CalledFunc, IFI,
CalledFunc->getParamAlignment(ArgNo+1));
// Calls that we inline may use the new alloca, so we need to clear
// their 'tail' flags if HandleByValArgument introduced a new alloca and
// the callee has calls.
MustClearTailCallFlags |= ActualArg != *AI;
}
VMap[I] = ActualArg;
}
// We want the inliner to prune the code as it copies. We would LOVE to
// have no dead or constant instructions leftover after inlining occurs
// (which can happen, e.g., because an argument was constant), but we'll be
// happy with whatever the cloner can do.
CloneAndPruneFunctionInto(Caller, CalledFunc, VMap,
/*ModuleLevelChanges=*/false, Returns, ".i",
&InlinedFunctionInfo, IFI.TD, TheCall);
// Remember the first block that is newly cloned over.
FirstNewBlock = LastBlock; ++FirstNewBlock;
// Update the callgraph if requested.
if (IFI.CG)
UpdateCallGraphAfterInlining(CS, FirstNewBlock, VMap, IFI);
// Update inlined instructions' line number information.
fixupLineNumbers(Caller, FirstNewBlock, TheCall);
}
// If there are any alloca instructions in the block that used to be the entry
// block for the callee, move them to the entry block of the caller. First
// calculate which instruction they should be inserted before. We insert the
// instructions at the end of the current alloca list.
{
BasicBlock::iterator InsertPoint = Caller->begin()->begin();
for (BasicBlock::iterator I = FirstNewBlock->begin(),
E = FirstNewBlock->end(); I != E; ) {
AllocaInst *AI = dyn_cast<AllocaInst>(I++);
if (AI == 0) continue;
// If the alloca is now dead, remove it. This often occurs due to code
// specialization.
if (AI->use_empty()) {
AI->eraseFromParent();
continue;
}
if (!isa<Constant>(AI->getArraySize()))
continue;
// Keep track of the static allocas that we inline into the caller.
IFI.StaticAllocas.push_back(AI);
// Scan for the block of allocas that we can move over, and move them
// all at once.
while (isa<AllocaInst>(I) &&
isa<Constant>(cast<AllocaInst>(I)->getArraySize())) {
IFI.StaticAllocas.push_back(cast<AllocaInst>(I));
++I;
}
// Transfer all of the allocas over in a block. Using splice means
// that the instructions aren't removed from the symbol table, then
// reinserted.
Caller->getEntryBlock().getInstList().splice(InsertPoint,
FirstNewBlock->getInstList(),
AI, I);
}
}
// Leave lifetime markers for the static alloca's, scoping them to the
// function we just inlined.
if (InsertLifetime && !IFI.StaticAllocas.empty()) {
IRBuilder<> builder(FirstNewBlock->begin());
for (unsigned ai = 0, ae = IFI.StaticAllocas.size(); ai != ae; ++ai) {
AllocaInst *AI = IFI.StaticAllocas[ai];
// If the alloca is already scoped to something smaller than the whole
// function then there's no need to add redundant, less accurate markers.
if (hasLifetimeMarkers(AI))
continue;
// Try to determine the size of the allocation.
ConstantInt *AllocaSize = 0;
if (ConstantInt *AIArraySize =
dyn_cast<ConstantInt>(AI->getArraySize())) {
if (IFI.TD) {
Type *AllocaType = AI->getAllocatedType();
uint64_t AllocaTypeSize = IFI.TD->getTypeAllocSize(AllocaType);
uint64_t AllocaArraySize = AIArraySize->getLimitedValue();
assert(AllocaArraySize > 0 && "array size of AllocaInst is zero");
// Check that array size doesn't saturate uint64_t and doesn't
// overflow when it's multiplied by type size.
if (AllocaArraySize != ~0ULL &&
UINT64_MAX / AllocaArraySize >= AllocaTypeSize) {
AllocaSize = ConstantInt::get(Type::getInt64Ty(AI->getContext()),
AllocaArraySize * AllocaTypeSize);
}
}
}
builder.CreateLifetimeStart(AI, AllocaSize);
for (unsigned ri = 0, re = Returns.size(); ri != re; ++ri) {
IRBuilder<> builder(Returns[ri]);
builder.CreateLifetimeEnd(AI, AllocaSize);
}
}
}
// If the inlined code contained dynamic alloca instructions, wrap the inlined
// code with llvm.stacksave/llvm.stackrestore intrinsics.
if (InlinedFunctionInfo.ContainsDynamicAllocas) {
Module *M = Caller->getParent();
// Get the two intrinsics we care about.
Function *StackSave = Intrinsic::getDeclaration(M, Intrinsic::stacksave);
Function *StackRestore=Intrinsic::getDeclaration(M,Intrinsic::stackrestore);
// Insert the llvm.stacksave.
CallInst *SavedPtr = IRBuilder<>(FirstNewBlock, FirstNewBlock->begin())
.CreateCall(StackSave, "savedstack");
// Insert a call to llvm.stackrestore before any return instructions in the
// inlined function.
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
IRBuilder<>(Returns[i]).CreateCall(StackRestore, SavedPtr);
}
}
// If we are inlining tail call instruction through a call site that isn't
// marked 'tail', we must remove the tail marker for any calls in the inlined
// code. Also, calls inlined through a 'nounwind' call site should be marked
// 'nounwind'.
if (InlinedFunctionInfo.ContainsCalls &&
(MustClearTailCallFlags || MarkNoUnwind)) {
for (Function::iterator BB = FirstNewBlock, E = Caller->end();
BB != E; ++BB)
for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
if (CallInst *CI = dyn_cast<CallInst>(I)) {
if (MustClearTailCallFlags)
CI->setTailCall(false);
if (MarkNoUnwind)
CI->setDoesNotThrow();
}
}
// If we are inlining for an invoke instruction, we must make sure to rewrite
// any call instructions into invoke instructions.
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall))
HandleInlinedInvoke(II, FirstNewBlock, InlinedFunctionInfo);
// If we cloned in _exactly one_ basic block, and if that block ends in a
// return instruction, we splice the body of the inlined callee directly into
// the calling basic block.
if (Returns.size() == 1 && std::distance(FirstNewBlock, Caller->end()) == 1) {
// Move all of the instructions right before the call.
OrigBB->getInstList().splice(TheCall, FirstNewBlock->getInstList(),
FirstNewBlock->begin(), FirstNewBlock->end());
// Remove the cloned basic block.
Caller->getBasicBlockList().pop_back();
// If the call site was an invoke instruction, add a branch to the normal
// destination.
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall)) {
BranchInst *NewBr = BranchInst::Create(II->getNormalDest(), TheCall);
NewBr->setDebugLoc(Returns[0]->getDebugLoc());
}
// If the return instruction returned a value, replace uses of the call with
// uses of the returned value.
if (!TheCall->use_empty()) {
ReturnInst *R = Returns[0];
if (TheCall == R->getReturnValue())
TheCall->replaceAllUsesWith(UndefValue::get(TheCall->getType()));
else
TheCall->replaceAllUsesWith(R->getReturnValue());
}
// Since we are now done with the Call/Invoke, we can delete it.
TheCall->eraseFromParent();
// Since we are now done with the return instruction, delete it also.
Returns[0]->eraseFromParent();
// We are now done with the inlining.
return true;
}
// Otherwise, we have the normal case, of more than one block to inline or
// multiple return sites.
// We want to clone the entire callee function into the hole between the
// "starter" and "ender" blocks. How we accomplish this depends on whether
// this is an invoke instruction or a call instruction.
BasicBlock *AfterCallBB;
BranchInst *CreatedBranchToNormalDest = NULL;
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall)) {
// Add an unconditional branch to make this look like the CallInst case...
CreatedBranchToNormalDest = BranchInst::Create(II->getNormalDest(), TheCall);
// Split the basic block. This guarantees that no PHI nodes will have to be
// updated due to new incoming edges, and make the invoke case more
// symmetric to the call case.
AfterCallBB = OrigBB->splitBasicBlock(CreatedBranchToNormalDest,
CalledFunc->getName()+".exit");
} else { // It's a call
// If this is a call instruction, we need to split the basic block that
// the call lives in.
//
AfterCallBB = OrigBB->splitBasicBlock(TheCall,
CalledFunc->getName()+".exit");
}
// Change the branch that used to go to AfterCallBB to branch to the first
// basic block of the inlined function.
//
TerminatorInst *Br = OrigBB->getTerminator();
assert(Br && Br->getOpcode() == Instruction::Br &&
"splitBasicBlock broken!");
Br->setOperand(0, FirstNewBlock);
// Now that the function is correct, make it a little bit nicer. In
// particular, move the basic blocks inserted from the end of the function
// into the space made by splitting the source basic block.
Caller->getBasicBlockList().splice(AfterCallBB, Caller->getBasicBlockList(),
FirstNewBlock, Caller->end());
// Handle all of the return instructions that we just cloned in, and eliminate
// any users of the original call/invoke instruction.
Type *RTy = CalledFunc->getReturnType();
PHINode *PHI = 0;
if (Returns.size() > 1) {
// The PHI node should go at the front of the new basic block to merge all
// possible incoming values.
if (!TheCall->use_empty()) {
PHI = PHINode::Create(RTy, Returns.size(), TheCall->getName(),
AfterCallBB->begin());
// Anything that used the result of the function call should now use the
// PHI node as their operand.
TheCall->replaceAllUsesWith(PHI);
}
// Loop over all of the return instructions adding entries to the PHI node
// as appropriate.
if (PHI) {
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
ReturnInst *RI = Returns[i];
assert(RI->getReturnValue()->getType() == PHI->getType() &&
"Ret value not consistent in function!");
PHI->addIncoming(RI->getReturnValue(), RI->getParent());
}
}
// Add a branch to the merge points and remove return instructions.
DebugLoc Loc;
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
ReturnInst *RI = Returns[i];
BranchInst* BI = BranchInst::Create(AfterCallBB, RI);
Loc = RI->getDebugLoc();
BI->setDebugLoc(Loc);
RI->eraseFromParent();
}
// We need to set the debug location to *somewhere* inside the
// inlined function. The line number may be nonsensical, but the
// instruction will at least be associated with the right
// function.
if (CreatedBranchToNormalDest)
CreatedBranchToNormalDest->setDebugLoc(Loc);
} else if (!Returns.empty()) {
// Otherwise, if there is exactly one return value, just replace anything
// using the return value of the call with the computed value.
if (!TheCall->use_empty()) {
if (TheCall == Returns[0]->getReturnValue())
TheCall->replaceAllUsesWith(UndefValue::get(TheCall->getType()));
else
TheCall->replaceAllUsesWith(Returns[0]->getReturnValue());
}
// Update PHI nodes that use the ReturnBB to use the AfterCallBB.
BasicBlock *ReturnBB = Returns[0]->getParent();
ReturnBB->replaceAllUsesWith(AfterCallBB);
// Splice the code from the return block into the block that it will return
// to, which contains the code that was after the call.
AfterCallBB->getInstList().splice(AfterCallBB->begin(),
ReturnBB->getInstList());
if (CreatedBranchToNormalDest)
CreatedBranchToNormalDest->setDebugLoc(Returns[0]->getDebugLoc());
// Delete the return instruction now and empty ReturnBB now.
Returns[0]->eraseFromParent();
ReturnBB->eraseFromParent();
} else if (!TheCall->use_empty()) {
// No returns, but something is using the return value of the call. Just
// nuke the result.
TheCall->replaceAllUsesWith(UndefValue::get(TheCall->getType()));
}
// Since we are now done with the Call/Invoke, we can delete it.
TheCall->eraseFromParent();
// We should always be able to fold the entry block of the function into the
// single predecessor of the block...
assert(cast<BranchInst>(Br)->isUnconditional() && "splitBasicBlock broken!");
BasicBlock *CalleeEntry = cast<BranchInst>(Br)->getSuccessor(0);
// Splice the code entry block into calling block, right before the
// unconditional branch.
CalleeEntry->replaceAllUsesWith(OrigBB); // Update PHI nodes
OrigBB->getInstList().splice(Br, CalleeEntry->getInstList());
// Remove the unconditional branch.
OrigBB->getInstList().erase(Br);
// Now we can remove the CalleeEntry block, which is now empty.
Caller->getBasicBlockList().erase(CalleeEntry);
// If we inserted a phi node, check to see if it has a single value (e.g. all
// the entries are the same or undef). If so, remove the PHI so it doesn't
// block other optimizations.
if (PHI) {
if (Value *V = SimplifyInstruction(PHI, IFI.TD)) {
PHI->replaceAllUsesWith(V);
PHI->eraseFromParent();
}
}
return true;
}
void FunctionLoweringInfo::set(const Function &fn, MachineFunction &mf) {
Fn = &fn;
MF = &mf;
RegInfo = &MF->getRegInfo();
// Check whether the function can return without sret-demotion.
SmallVector<ISD::OutputArg, 4> Outs;
GetReturnInfo(Fn->getReturnType(),
Fn->getAttributes().getRetAttributes(), Outs, TLI);
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 (BasicBlock::const_iterator I = BB->begin(), E = BB->end(); I != E; ++I)
if (const AllocaInst *AI = dyn_cast<AllocaInst>(I))
if (const ConstantInt *CUI = dyn_cast<ConstantInt>(AI->getArraySize())) {
const Type *Ty = AI->getAllocatedType();
uint64_t TySize = TLI.getTargetData()->getTypeAllocSize(Ty);
unsigned Align =
std::max((unsigned)TLI.getTargetData()->getPrefTypeAlignment(Ty),
AI->getAlignment());
TySize *= CUI->getZExtValue(); // Get total allocated size.
if (TySize == 0) TySize = 1; // Don't create zero-sized stack objects.
// The object may need to be placed onto the stack near the stack
// protector if one exists. Determine here if this object is a suitable
// candidate. I.e., it would trigger the creation of a stack protector.
bool MayNeedSP =
(AI->isArrayAllocation() ||
(TySize > 8 && isa<ArrayType>(Ty) &&
cast<ArrayType>(Ty)->getElementType()->isIntegerTy(8)));
StaticAllocaMap[AI] =
MF->getFrameInfo()->CreateStackObject(TySize, Align, false, MayNeedSP);
}
for (; BB != EB; ++BB)
for (BasicBlock::const_iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
// 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)) {
MachineModuleInfo &MMI = MF->getMMI();
if (MMI.hasDebugInfo() &&
DIVariable(DI->getVariable()).Verify() &&
!DI->getDebugLoc().isUnknown()) {
// 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(),
FI, DI->getDebugLoc());
}
}
}
}
}
}
// 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) {
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, 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->getTarget().getInstrInfo();
for (unsigned i = 0; i != NumRegisters; ++i)
BuildMI(MBB, DL, TII->get(TargetOpcode::PHI), PHIReg + i);
PHIReg += NumRegisters;
}
}
}
// Mark landing pad blocks.
for (BB = Fn->begin(); BB != EB; ++BB)
if (const InvokeInst *Invoke = dyn_cast<InvokeInst>(BB->getTerminator()))
MBBMap[Invoke->getSuccessor(1)]->setIsLandingPad();
}
void FunctionLoweringInfo::set(const Function &fn, MachineFunction &mf) {
Fn = &fn;
MF = &mf;
RegInfo = &MF->getRegInfo();
// Create a vreg for each argument register that is not dead and is used
// outside of the entry block for the function.
for (Function::const_arg_iterator AI = Fn->arg_begin(), E = Fn->arg_end();
AI != E; ++AI)
if (!isOnlyUsedInEntryBlock(AI, EnableFastISel))
InitializeRegForValue(AI);
// 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 (BasicBlock::const_iterator I = BB->begin(), E = BB->end(); I != E; ++I)
if (const AllocaInst *AI = dyn_cast<AllocaInst>(I))
if (const ConstantInt *CUI = dyn_cast<ConstantInt>(AI->getArraySize())) {
const Type *Ty = AI->getAllocatedType();
uint64_t TySize = TLI.getTargetData()->getTypeAllocSize(Ty);
unsigned Align =
std::max((unsigned)TLI.getTargetData()->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);
}
for (; BB != EB; ++BB)
for (BasicBlock::const_iterator I = BB->begin(), E = BB->end(); I != E; ++I)
if (isUsedOutsideOfDefiningBlock(I))
if (!isa<AllocaInst>(I) ||
!StaticAllocaMap.count(cast<AllocaInst>(I)))
InitializeRegForValue(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) {
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;
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, 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->getTarget().getInstrInfo();
for (unsigned i = 0; i != NumRegisters; ++i)
BuildMI(MBB, DL, TII->get(TargetOpcode::PHI), PHIReg + i);
PHIReg += NumRegisters;
}
}
}
// Mark landing pad blocks.
for (BB = Fn->begin(); BB != EB; ++BB)
if (const InvokeInst *Invoke = dyn_cast<InvokeInst>(BB->getTerminator()))
MBBMap[Invoke->getSuccessor(1)]->setIsLandingPad();
}
