bool GlobalMerge::doInitialization(Module &M) {
DenseMap<unsigned, SmallVector<GlobalVariable*, 16> > Globals, ConstGlobals,
BSSGlobals;
const DataLayout *TD = TLI->getDataLayout();
unsigned MaxOffset = TLI->getMaximalGlobalOffset();
bool Changed = false;
// Grab all non-const globals.
for (Module::global_iterator I = M.global_begin(),
E = M.global_end(); I != E; ++I) {
// Merge is safe for "normal" internal globals only
if (!I->hasLocalLinkage() || I->isThreadLocal() || I->hasSection())
continue;
PointerType *PT = dyn_cast<PointerType>(I->getType());
assert(PT && "Global variable is not a pointer!");
unsigned AddressSpace = PT->getAddressSpace();
// Ignore fancy-aligned globals for now.
unsigned Alignment = TD->getPreferredAlignment(I);
Type *Ty = I->getType()->getElementType();
if (Alignment > TD->getABITypeAlignment(Ty))
continue;
// Ignore all 'special' globals.
if (I->getName().startswith("llvm.") ||
I->getName().startswith(".llvm."))
continue;
if (TD->getTypeAllocSize(Ty) < MaxOffset) {
if (TargetLoweringObjectFile::getKindForGlobal(I, TLI->getTargetMachine())
.isBSSLocal())
BSSGlobals[AddressSpace].push_back(I);
else if (I->isConstant())
ConstGlobals[AddressSpace].push_back(I);
else
Globals[AddressSpace].push_back(I);
}
}
for (DenseMap<unsigned, SmallVector<GlobalVariable*, 16> >::iterator
I = Globals.begin(), E = Globals.end(); I != E; ++I)
if (I->second.size() > 1)
Changed |= doMerge(I->second, M, false, I->first);
for (DenseMap<unsigned, SmallVector<GlobalVariable*, 16> >::iterator
I = BSSGlobals.begin(), E = BSSGlobals.end(); I != E; ++I)
if (I->second.size() > 1)
Changed |= doMerge(I->second, M, false, I->first);
// FIXME: This currently breaks the EH processing due to way how the
// typeinfo detection works. We might want to detect the TIs and ignore
// them in the future.
// if (ConstGlobals.size() > 1)
// Changed |= doMerge(ConstGlobals, M, true);
return Changed;
}
/// InstCombineLoadCast - Fold 'load (cast P)' -> cast (load P)' when possible.
static Instruction *InstCombineLoadCast(InstCombiner &IC, LoadInst &LI,
const DataLayout *DL) {
User *CI = cast<User>(LI.getOperand(0));
Value *CastOp = CI->getOperand(0);
PointerType *DestTy = cast<PointerType>(CI->getType());
Type *DestPTy = DestTy->getElementType();
if (PointerType *SrcTy = dyn_cast<PointerType>(CastOp->getType())) {
// If the address spaces don't match, don't eliminate the cast.
if (DestTy->getAddressSpace() != SrcTy->getAddressSpace())
return 0;
Type *SrcPTy = SrcTy->getElementType();
if (DestPTy->isIntegerTy() || DestPTy->isPointerTy() ||
DestPTy->isVectorTy()) {
// If the source is an array, the code below will not succeed. Check to
// see if a trivial 'gep P, 0, 0' will help matters. Only do this for
// constants.
if (ArrayType *ASrcTy = dyn_cast<ArrayType>(SrcPTy))
if (Constant *CSrc = dyn_cast<Constant>(CastOp))
if (ASrcTy->getNumElements() != 0) {
Type *IdxTy = DL
? DL->getIntPtrType(SrcTy)
: Type::getInt64Ty(SrcTy->getContext());
Value *Idx = Constant::getNullValue(IdxTy);
Value *Idxs[2] = { Idx, Idx };
CastOp = ConstantExpr::getGetElementPtr(CSrc, Idxs);
SrcTy = cast<PointerType>(CastOp->getType());
SrcPTy = SrcTy->getElementType();
}
if (IC.getDataLayout() &&
(SrcPTy->isIntegerTy() || SrcPTy->isPointerTy() ||
SrcPTy->isVectorTy()) &&
// Do not allow turning this into a load of an integer, which is then
// casted to a pointer, this pessimizes pointer analysis a lot.
(SrcPTy->isPtrOrPtrVectorTy() ==
LI.getType()->isPtrOrPtrVectorTy()) &&
IC.getDataLayout()->getTypeSizeInBits(SrcPTy) ==
IC.getDataLayout()->getTypeSizeInBits(DestPTy)) {
// Okay, we are casting from one integer or pointer type to another of
// the same size. Instead of casting the pointer before the load, cast
// the result of the loaded value.
LoadInst *NewLoad =
IC.Builder->CreateLoad(CastOp, LI.isVolatile(), CI->getName());
NewLoad->setAlignment(LI.getAlignment());
NewLoad->setAtomic(LI.getOrdering(), LI.getSynchScope());
// Now cast the result of the load.
return new BitCastInst(NewLoad, LI.getType());
}
}
}
return 0;
}
bool AMDGPUPromoteAlloca::runOnFunction(Function &F) {
if (!TM || skipFunction(F))
return false;
FunctionType *FTy = F.getFunctionType();
// If the function has any arguments in the local address space, then it's
// possible these arguments require the entire local memory space, so
// we cannot use local memory in the pass.
for (Type *ParamTy : FTy->params()) {
PointerType *PtrTy = dyn_cast<PointerType>(ParamTy);
if (PtrTy && PtrTy->getAddressSpace() == AMDGPUAS::LOCAL_ADDRESS) {
LocalMemAvailable = 0;
DEBUG(dbgs() << "Function has local memory argument. Promoting to "
"local memory disabled.\n");
return false;
}
}
const AMDGPUSubtarget &ST = TM->getSubtarget<AMDGPUSubtarget>(F);
LocalMemAvailable = ST.getLocalMemorySize();
if (LocalMemAvailable == 0)
return false;
// Check how much local memory is being used by global objects
for (GlobalVariable &GV : Mod->globals()) {
if (GV.getType()->getAddressSpace() != AMDGPUAS::LOCAL_ADDRESS)
continue;
for (User *U : GV.users()) {
Instruction *Use = dyn_cast<Instruction>(U);
if (!Use)
continue;
if (Use->getParent()->getParent() == &F) {
LocalMemAvailable -=
Mod->getDataLayout().getTypeAllocSize(GV.getValueType());
break;
}
}
}
LocalMemAvailable = std::max(0, LocalMemAvailable);
DEBUG(dbgs() << LocalMemAvailable << " bytes free in local memory.\n");
BasicBlock &EntryBB = *F.begin();
for (auto I = EntryBB.begin(), E = EntryBB.end(); I != E; ) {
AllocaInst *AI = dyn_cast<AllocaInst>(I);
++I;
if (AI)
handleAlloca(*AI);
}
return true;
}
// Decides whether V is an addrspacecast and shortcutting V in load/store is
// valid and beneficial.
static bool isEliminableAddrSpaceCast(Value *V) {
// Returns false if V is not even an addrspacecast.
Operator *Cast = dyn_cast<Operator>(V);
if (Cast == nullptr || Cast->getOpcode() != Instruction::AddrSpaceCast)
return false;
Value *Src = Cast->getOperand(0);
PointerType *SrcTy = cast<PointerType>(Src->getType());
PointerType *DestTy = cast<PointerType>(Cast->getType());
// TODO: For now, we only handle the case where the addrspacecast only changes
// the address space but not the type. If the type also changes, we could
// still get rid of the addrspacecast by adding an extra bitcast, but we
// rarely see such scenarios.
if (SrcTy->getElementType() != DestTy->getElementType())
return false;
// Checks whether the addrspacecast is from a non-generic address space to the
// generic address space.
return (SrcTy->getAddressSpace() != AddressSpace::ADDRESS_SPACE_GENERIC &&
DestTy->getAddressSpace() == AddressSpace::ADDRESS_SPACE_GENERIC);
}
Value *IRBuilderBase::getCastedInt8PtrValue(Value *Ptr) {
PointerType *PT = cast<PointerType>(Ptr->getType());
if (PT->getElementType()->isIntegerTy(8))
return Ptr;
// Otherwise, we need to insert a bitcast.
PT = getInt8PtrTy(PT->getAddressSpace());
BitCastInst *BCI = new BitCastInst(Ptr, PT, "");
BB->getInstList().insert(InsertPt, BCI);
SetInstDebugLocation(BCI);
return BCI;
}
bool AMDGPURewriteOutArguments::isOutArgumentCandidate(Argument &Arg) const {
const unsigned MaxOutArgSizeBytes = 4 * MaxNumRetRegs;
PointerType *ArgTy = dyn_cast<PointerType>(Arg.getType());
// TODO: It might be useful for any out arguments, not just privates.
if (!ArgTy || (ArgTy->getAddressSpace() != DL->getAllocaAddrSpace() &&
!AnyAddressSpace) ||
Arg.hasByValAttr() || Arg.hasStructRetAttr() ||
DL->getTypeStoreSize(ArgTy->getPointerElementType()) > MaxOutArgSizeBytes) {
return false;
}
return checkArgumentUses(Arg);
}
bool AMDGPUPromoteAlloca::runOnFunction(Function &F) {
const FunctionType *FTy = F.getFunctionType();
LocalMemAvailable = ST.getLocalMemorySize();
// If the function has any arguments in the local address space, then it's
// possible these arguments require the entire local memory space, so
// we cannot use local memory in the pass.
for (unsigned i = 0, e = FTy->getNumParams(); i != e; ++i) {
const Type *ParamTy = FTy->getParamType(i);
if (ParamTy->isPointerTy() &&
ParamTy->getPointerAddressSpace() == AMDGPUAS::LOCAL_ADDRESS) {
LocalMemAvailable = 0;
DEBUG(dbgs() << "Function has local memory argument. Promoting to "
"local memory disabled.\n");
break;
}
}
if (LocalMemAvailable > 0) {
// Check how much local memory is being used by global objects
for (Module::global_iterator I = Mod->global_begin(),
E = Mod->global_end(); I != E; ++I) {
GlobalVariable *GV = I;
PointerType *GVTy = GV->getType();
if (GVTy->getAddressSpace() != AMDGPUAS::LOCAL_ADDRESS)
continue;
for (Value::use_iterator U = GV->use_begin(),
UE = GV->use_end(); U != UE; ++U) {
Instruction *Use = dyn_cast<Instruction>(*U);
if (!Use)
continue;
if (Use->getParent()->getParent() == &F)
LocalMemAvailable -=
Mod->getDataLayout()->getTypeAllocSize(GVTy->getElementType());
}
}
}
LocalMemAvailable = std::max(0, LocalMemAvailable);
DEBUG(dbgs() << LocalMemAvailable << "bytes free in local memory.\n");
visit(F);
return false;
}
bool AMDGPUPromoteAlloca::runOnFunction(Function &F) {
if (!TM || F.hasFnAttribute(Attribute::OptimizeNone))
return false;
FunctionType *FTy = F.getFunctionType();
// If the function has any arguments in the local address space, then it's
// possible these arguments require the entire local memory space, so
// we cannot use local memory in the pass.
for (Type *ParamTy : FTy->params()) {
PointerType *PtrTy = dyn_cast<PointerType>(ParamTy);
if (PtrTy && PtrTy->getAddressSpace() == AMDGPUAS::LOCAL_ADDRESS) {
LocalMemAvailable = 0;
DEBUG(dbgs() << "Function has local memory argument. Promoting to "
"local memory disabled.\n");
return false;
}
}
const AMDGPUSubtarget &ST = TM->getSubtarget<AMDGPUSubtarget>(F);
LocalMemAvailable = ST.getLocalMemorySize();
if (LocalMemAvailable == 0)
return false;
// Check how much local memory is being used by global objects
for (GlobalVariable &GV : Mod->globals()) {
if (GV.getType()->getAddressSpace() != AMDGPUAS::LOCAL_ADDRESS)
continue;
for (Use &U : GV.uses()) {
Instruction *Use = dyn_cast<Instruction>(U);
if (!Use)
continue;
if (Use->getParent()->getParent() == &F)
LocalMemAvailable -=
Mod->getDataLayout().getTypeAllocSize(GV.getValueType());
}
}
LocalMemAvailable = std::max(0, LocalMemAvailable);
DEBUG(dbgs() << LocalMemAvailable << " bytes free in local memory.\n");
visit(F);
return true;
}
Value *GenericToNVVM::getOrInsertCVTA(Module *M, Function *F,
GlobalVariable *GV,
IRBuilder<> &Builder) {
PointerType *GVType = GV->getType();
Value *CVTA = nullptr;
// See if the address space conversion requires the operand to be bitcast
// to i8 addrspace(n)* first.
EVT ExtendedGVType = EVT::getEVT(GVType->getElementType(), true);
if (!ExtendedGVType.isInteger() && !ExtendedGVType.isFloatingPoint()) {
// A bitcast to i8 addrspace(n)* on the operand is needed.
LLVMContext &Context = M->getContext();
unsigned int AddrSpace = GVType->getAddressSpace();
Type *DestTy = PointerType::get(Type::getInt8Ty(Context), AddrSpace);
CVTA = Builder.CreateBitCast(GV, DestTy, "cvta");
// Insert the address space conversion.
Type *ResultType =
PointerType::get(Type::getInt8Ty(Context), llvm::ADDRESS_SPACE_GENERIC);
SmallVector<Type *, 2> ParamTypes;
ParamTypes.push_back(ResultType);
ParamTypes.push_back(DestTy);
Function *CVTAFunction = Intrinsic::getDeclaration(
M, Intrinsic::nvvm_ptr_global_to_gen, ParamTypes);
CVTA = Builder.CreateCall(CVTAFunction, CVTA, "cvta");
// Another bitcast from i8 * to <the element type of GVType> * is
// required.
DestTy =
PointerType::get(GVType->getElementType(), llvm::ADDRESS_SPACE_GENERIC);
CVTA = Builder.CreateBitCast(CVTA, DestTy, "cvta");
} else {
// A simple CVTA is enough.
SmallVector<Type *, 2> ParamTypes;
ParamTypes.push_back(PointerType::get(GVType->getElementType(),
llvm::ADDRESS_SPACE_GENERIC));
ParamTypes.push_back(GVType);
Function *CVTAFunction = Intrinsic::getDeclaration(
M, Intrinsic::nvvm_ptr_global_to_gen, ParamTypes);
CVTA = Builder.CreateCall(CVTAFunction, GV, "cvta");
}
return CVTA;
}
static Value *getFieldAddress(IRBuilder<> &Builder, Value *Base,
uint32_t Offset, Type *FieldTy) {
// The base value should be an i8* or i8[]*.
assert(Base->getType()->isPointerTy());
assert(Base->getType()->getPointerElementType()->isIntegerTy(8) ||
Base->getType()
->getPointerElementType()
->getArrayElementType()
->isIntegerTy(8));
Type *Int32Ty = Type::getInt32Ty(Builder.getContext());
Value *Indices[] = {ConstantInt::get(Int32Ty, 0),
ConstantInt::get(Int32Ty, Offset)};
Value *Address = Builder.CreateInBoundsGEP(Base, Indices);
PointerType *AddressTy = cast<PointerType>(Address->getType());
if (AddressTy->getElementType() != FieldTy) {
AddressTy = PointerType::get(FieldTy, AddressTy->getAddressSpace());
Address = Builder.CreatePointerCast(Address, AddressTy);
}
return Address;
}
bool GlobalMerge::doInitialization(Module &M) {
if (!EnableGlobalMerge)
return false;
IsMachO = Triple(M.getTargetTriple()).isOSBinFormatMachO();
auto &DL = M.getDataLayout();
DenseMap<unsigned, SmallVector<GlobalVariable *, 16>> Globals, ConstGlobals,
BSSGlobals;
bool Changed = false;
setMustKeepGlobalVariables(M);
// Grab all non-const globals.
for (auto &GV : M.globals()) {
// Merge is safe for "normal" internal or external globals only
if (GV.isDeclaration() || GV.isThreadLocal() ||
GV.hasSection() || GV.hasImplicitSection())
continue;
// It's not safe to merge globals that may be preempted
if (TM && !TM->shouldAssumeDSOLocal(M, &GV))
continue;
if (!(MergeExternalGlobals && GV.hasExternalLinkage()) &&
!GV.hasInternalLinkage())
continue;
PointerType *PT = dyn_cast<PointerType>(GV.getType());
assert(PT && "Global variable is not a pointer!");
unsigned AddressSpace = PT->getAddressSpace();
// Ignore fancy-aligned globals for now.
unsigned Alignment = DL.getPreferredAlignment(&GV);
Type *Ty = GV.getValueType();
if (Alignment > DL.getABITypeAlignment(Ty))
continue;
// Ignore all 'special' globals.
if (GV.getName().startswith("llvm.") ||
GV.getName().startswith(".llvm."))
continue;
// Ignore all "required" globals:
if (isMustKeepGlobalVariable(&GV))
continue;
if (DL.getTypeAllocSize(Ty) < MaxOffset) {
if (TM &&
TargetLoweringObjectFile::getKindForGlobal(&GV, *TM).isBSSLocal())
BSSGlobals[AddressSpace].push_back(&GV);
else if (GV.isConstant())
ConstGlobals[AddressSpace].push_back(&GV);
else
Globals[AddressSpace].push_back(&GV);
}
}
for (auto &P : Globals)
if (P.second.size() > 1)
Changed |= doMerge(P.second, M, false, P.first);
for (auto &P : BSSGlobals)
if (P.second.size() > 1)
Changed |= doMerge(P.second, M, false, P.first);
if (EnableGlobalMergeOnConst)
for (auto &P : ConstGlobals)
if (P.second.size() > 1)
Changed |= doMerge(P.second, M, true, P.first);
return Changed;
}
bool AMDGPURewriteOutArguments::runOnFunction(Function &F) {
if (skipFunction(F))
return false;
// TODO: Could probably handle variadic functions.
if (F.isVarArg() || F.hasStructRetAttr() ||
AMDGPU::isEntryFunctionCC(F.getCallingConv()))
return false;
MDA = &getAnalysis<MemoryDependenceWrapperPass>().getMemDep();
unsigned ReturnNumRegs = 0;
SmallSet<int, 4> OutArgIndexes;
SmallVector<Type *, 4> ReturnTypes;
Type *RetTy = F.getReturnType();
if (!RetTy->isVoidTy()) {
ReturnNumRegs = DL->getTypeStoreSize(RetTy) / 4;
if (ReturnNumRegs >= MaxNumRetRegs)
return false;
ReturnTypes.push_back(RetTy);
}
SmallVector<Argument *, 4> OutArgs;
for (Argument &Arg : F.args()) {
if (isOutArgumentCandidate(Arg)) {
LLVM_DEBUG(dbgs() << "Found possible out argument " << Arg
<< " in function " << F.getName() << '\n');
OutArgs.push_back(&Arg);
}
}
if (OutArgs.empty())
return false;
using ReplacementVec = SmallVector<std::pair<Argument *, Value *>, 4>;
DenseMap<ReturnInst *, ReplacementVec> Replacements;
SmallVector<ReturnInst *, 4> Returns;
for (BasicBlock &BB : F) {
if (ReturnInst *RI = dyn_cast<ReturnInst>(&BB.back()))
Returns.push_back(RI);
}
if (Returns.empty())
return false;
bool Changing;
do {
Changing = false;
// Keep retrying if we are able to successfully eliminate an argument. This
// helps with cases with multiple arguments which may alias, such as in a
// sincos implemntation. If we have 2 stores to arguments, on the first
// attempt the MDA query will succeed for the second store but not the
// first. On the second iteration we've removed that out clobbering argument
// (by effectively moving it into another function) and will find the second
// argument is OK to move.
for (Argument *OutArg : OutArgs) {
bool ThisReplaceable = true;
SmallVector<std::pair<ReturnInst *, StoreInst *>, 4> ReplaceableStores;
Type *ArgTy = OutArg->getType()->getPointerElementType();
// Skip this argument if converting it will push us over the register
// count to return limit.
// TODO: This is an approximation. When legalized this could be more. We
// can ask TLI for exactly how many.
unsigned ArgNumRegs = DL->getTypeStoreSize(ArgTy) / 4;
if (ArgNumRegs + ReturnNumRegs > MaxNumRetRegs)
continue;
// An argument is convertible only if all exit blocks are able to replace
// it.
for (ReturnInst *RI : Returns) {
BasicBlock *BB = RI->getParent();
MemDepResult Q = MDA->getPointerDependencyFrom(MemoryLocation(OutArg),
true, BB->end(), BB, RI);
StoreInst *SI = nullptr;
if (Q.isDef())
SI = dyn_cast<StoreInst>(Q.getInst());
if (SI) {
LLVM_DEBUG(dbgs() << "Found out argument store: " << *SI << '\n');
ReplaceableStores.emplace_back(RI, SI);
} else {
ThisReplaceable = false;
break;
}
}
if (!ThisReplaceable)
continue; // Try the next argument candidate.
for (std::pair<ReturnInst *, StoreInst *> Store : ReplaceableStores) {
Value *ReplVal = Store.second->getValueOperand();
auto &ValVec = Replacements[Store.first];
if (llvm::find_if(ValVec,
[OutArg](const std::pair<Argument *, Value *> &Entry) {
return Entry.first == OutArg;}) != ValVec.end()) {
LLVM_DEBUG(dbgs()
<< "Saw multiple out arg stores" << *OutArg << '\n');
// It is possible to see stores to the same argument multiple times,
// but we expect these would have been optimized out already.
ThisReplaceable = false;
break;
}
ValVec.emplace_back(OutArg, ReplVal);
Store.second->eraseFromParent();
}
if (ThisReplaceable) {
ReturnTypes.push_back(ArgTy);
OutArgIndexes.insert(OutArg->getArgNo());
++NumOutArgumentsReplaced;
Changing = true;
}
}
} while (Changing);
if (Replacements.empty())
return false;
LLVMContext &Ctx = F.getParent()->getContext();
StructType *NewRetTy = StructType::create(Ctx, ReturnTypes, F.getName());
FunctionType *NewFuncTy = FunctionType::get(NewRetTy,
F.getFunctionType()->params(),
F.isVarArg());
LLVM_DEBUG(dbgs() << "Computed new return type: " << *NewRetTy << '\n');
Function *NewFunc = Function::Create(NewFuncTy, Function::PrivateLinkage,
F.getName() + ".body");
F.getParent()->getFunctionList().insert(F.getIterator(), NewFunc);
NewFunc->copyAttributesFrom(&F);
NewFunc->setComdat(F.getComdat());
// We want to preserve the function and param attributes, but need to strip
// off any return attributes, e.g. zeroext doesn't make sense with a struct.
NewFunc->stealArgumentListFrom(F);
AttrBuilder RetAttrs;
RetAttrs.addAttribute(Attribute::SExt);
RetAttrs.addAttribute(Attribute::ZExt);
RetAttrs.addAttribute(Attribute::NoAlias);
NewFunc->removeAttributes(AttributeList::ReturnIndex, RetAttrs);
// TODO: How to preserve metadata?
// Move the body of the function into the new rewritten function, and replace
// this function with a stub.
NewFunc->getBasicBlockList().splice(NewFunc->begin(), F.getBasicBlockList());
for (std::pair<ReturnInst *, ReplacementVec> &Replacement : Replacements) {
ReturnInst *RI = Replacement.first;
IRBuilder<> B(RI);
B.SetCurrentDebugLocation(RI->getDebugLoc());
int RetIdx = 0;
Value *NewRetVal = UndefValue::get(NewRetTy);
Value *RetVal = RI->getReturnValue();
if (RetVal)
NewRetVal = B.CreateInsertValue(NewRetVal, RetVal, RetIdx++);
for (std::pair<Argument *, Value *> ReturnPoint : Replacement.second) {
Argument *Arg = ReturnPoint.first;
Value *Val = ReturnPoint.second;
Type *EltTy = Arg->getType()->getPointerElementType();
if (Val->getType() != EltTy) {
Type *EffectiveEltTy = EltTy;
if (StructType *CT = dyn_cast<StructType>(EltTy)) {
assert(CT->getNumElements() == 1);
EffectiveEltTy = CT->getElementType(0);
}
if (DL->getTypeSizeInBits(EffectiveEltTy) !=
DL->getTypeSizeInBits(Val->getType())) {
assert(isVec3ToVec4Shuffle(EffectiveEltTy, Val->getType()));
Val = B.CreateShuffleVector(Val, UndefValue::get(Val->getType()),
{ 0, 1, 2 });
}
Val = B.CreateBitCast(Val, EffectiveEltTy);
// Re-create single element composite.
if (EltTy != EffectiveEltTy)
Val = B.CreateInsertValue(UndefValue::get(EltTy), Val, 0);
}
NewRetVal = B.CreateInsertValue(NewRetVal, Val, RetIdx++);
}
if (RetVal)
RI->setOperand(0, NewRetVal);
else {
B.CreateRet(NewRetVal);
RI->eraseFromParent();
}
}
SmallVector<Value *, 16> StubCallArgs;
for (Argument &Arg : F.args()) {
if (OutArgIndexes.count(Arg.getArgNo())) {
// It's easier to preserve the type of the argument list. We rely on
// DeadArgumentElimination to take care of these.
StubCallArgs.push_back(UndefValue::get(Arg.getType()));
} else {
StubCallArgs.push_back(&Arg);
}
}
BasicBlock *StubBB = BasicBlock::Create(Ctx, "", &F);
IRBuilder<> B(StubBB);
CallInst *StubCall = B.CreateCall(NewFunc, StubCallArgs);
int RetIdx = RetTy->isVoidTy() ? 0 : 1;
for (Argument &Arg : F.args()) {
if (!OutArgIndexes.count(Arg.getArgNo()))
continue;
PointerType *ArgType = cast<PointerType>(Arg.getType());
auto *EltTy = ArgType->getElementType();
unsigned Align = Arg.getParamAlignment();
if (Align == 0)
Align = DL->getABITypeAlignment(EltTy);
Value *Val = B.CreateExtractValue(StubCall, RetIdx++);
Type *PtrTy = Val->getType()->getPointerTo(ArgType->getAddressSpace());
// We can peek through bitcasts, so the type may not match.
Value *PtrVal = B.CreateBitCast(&Arg, PtrTy);
B.CreateAlignedStore(Val, PtrVal, Align);
}
if (!RetTy->isVoidTy()) {
B.CreateRet(B.CreateExtractValue(StubCall, 0));
} else {
B.CreateRetVoid();
}
// The function is now a stub we want to inline.
F.addFnAttr(Attribute::AlwaysInline);
++NumOutArgumentFunctionsReplaced;
return true;
}
bool GlobalMerge::doInitialization(Module &M) {
if (!EnableGlobalMerge)
return false;
auto &DL = M.getDataLayout();
DenseMap<unsigned, SmallVector<GlobalVariable*, 16> > Globals, ConstGlobals,
BSSGlobals;
bool Changed = false;
setMustKeepGlobalVariables(M);
// Grab all non-const globals.
for (Module::global_iterator I = M.global_begin(),
E = M.global_end(); I != E; ++I) {
// Merge is safe for "normal" internal or external globals only
if (I->isDeclaration() || I->isThreadLocal() || I->hasSection())
continue;
if (!(EnableGlobalMergeOnExternal && I->hasExternalLinkage()) &&
!I->hasInternalLinkage())
continue;
PointerType *PT = dyn_cast<PointerType>(I->getType());
assert(PT && "Global variable is not a pointer!");
unsigned AddressSpace = PT->getAddressSpace();
// Ignore fancy-aligned globals for now.
unsigned Alignment = DL.getPreferredAlignment(I);
Type *Ty = I->getType()->getElementType();
if (Alignment > DL.getABITypeAlignment(Ty))
continue;
// Ignore all 'special' globals.
if (I->getName().startswith("llvm.") ||
I->getName().startswith(".llvm."))
continue;
// Ignore all "required" globals:
if (isMustKeepGlobalVariable(I))
continue;
if (DL.getTypeAllocSize(Ty) < MaxOffset) {
if (TargetLoweringObjectFile::getKindForGlobal(I, *TM).isBSSLocal())
BSSGlobals[AddressSpace].push_back(I);
else if (I->isConstant())
ConstGlobals[AddressSpace].push_back(I);
else
Globals[AddressSpace].push_back(I);
}
}
for (DenseMap<unsigned, SmallVector<GlobalVariable*, 16> >::iterator
I = Globals.begin(), E = Globals.end(); I != E; ++I)
if (I->second.size() > 1)
Changed |= doMerge(I->second, M, false, I->first);
for (DenseMap<unsigned, SmallVector<GlobalVariable*, 16> >::iterator
I = BSSGlobals.begin(), E = BSSGlobals.end(); I != E; ++I)
if (I->second.size() > 1)
Changed |= doMerge(I->second, M, false, I->first);
if (EnableGlobalMergeOnConst)
for (DenseMap<unsigned, SmallVector<GlobalVariable*, 16> >::iterator
I = ConstGlobals.begin(), E = ConstGlobals.end(); I != E; ++I)
if (I->second.size() > 1)
Changed |= doMerge(I->second, M, true, I->first);
return Changed;
}
bool AMDGPUPromoteAlloca::hasSufficientLocalMem(const Function &F) {
FunctionType *FTy = F.getFunctionType();
const AMDGPUSubtarget &ST = TM->getSubtarget<AMDGPUSubtarget>(F);
// If the function has any arguments in the local address space, then it's
// possible these arguments require the entire local memory space, so
// we cannot use local memory in the pass.
for (Type *ParamTy : FTy->params()) {
PointerType *PtrTy = dyn_cast<PointerType>(ParamTy);
if (PtrTy && PtrTy->getAddressSpace() == AS.LOCAL_ADDRESS) {
LocalMemLimit = 0;
DEBUG(dbgs() << "Function has local memory argument. Promoting to "
"local memory disabled.\n");
return false;
}
}
LocalMemLimit = ST.getLocalMemorySize();
if (LocalMemLimit == 0)
return false;
const DataLayout &DL = Mod->getDataLayout();
// Check how much local memory is being used by global objects
CurrentLocalMemUsage = 0;
for (GlobalVariable &GV : Mod->globals()) {
if (GV.getType()->getAddressSpace() != AS.LOCAL_ADDRESS)
continue;
for (const User *U : GV.users()) {
const Instruction *Use = dyn_cast<Instruction>(U);
if (!Use)
continue;
if (Use->getParent()->getParent() == &F) {
unsigned Align = GV.getAlignment();
if (Align == 0)
Align = DL.getABITypeAlignment(GV.getValueType());
// FIXME: Try to account for padding here. The padding is currently
// determined from the inverse order of uses in the function. I'm not
// sure if the use list order is in any way connected to this, so the
// total reported size is likely incorrect.
uint64_t AllocSize = DL.getTypeAllocSize(GV.getValueType());
CurrentLocalMemUsage = alignTo(CurrentLocalMemUsage, Align);
CurrentLocalMemUsage += AllocSize;
break;
}
}
}
unsigned MaxOccupancy = ST.getOccupancyWithLocalMemSize(CurrentLocalMemUsage,
F);
// Restrict local memory usage so that we don't drastically reduce occupancy,
// unless it is already significantly reduced.
// TODO: Have some sort of hint or other heuristics to guess occupancy based
// on other factors..
unsigned OccupancyHint = ST.getWavesPerEU(F).second;
if (OccupancyHint == 0)
OccupancyHint = 7;
// Clamp to max value.
OccupancyHint = std::min(OccupancyHint, ST.getMaxWavesPerEU());
// Check the hint but ignore it if it's obviously wrong from the existing LDS
// usage.
MaxOccupancy = std::min(OccupancyHint, MaxOccupancy);
// Round up to the next tier of usage.
unsigned MaxSizeWithWaveCount
= ST.getMaxLocalMemSizeWithWaveCount(MaxOccupancy, F);
// Program is possibly broken by using more local mem than available.
if (CurrentLocalMemUsage > MaxSizeWithWaveCount)
return false;
LocalMemLimit = MaxSizeWithWaveCount;
DEBUG(
dbgs() << F.getName() << " uses " << CurrentLocalMemUsage << " bytes of LDS\n"
<< " Rounding size to " << MaxSizeWithWaveCount
<< " with a maximum occupancy of " << MaxOccupancy << '\n'
<< " and " << (LocalMemLimit - CurrentLocalMemUsage)
<< " available for promotion\n"
);
return true;
}
/// InstCombineStoreToCast - Fold store V, (cast P) -> store (cast V), P
/// when possible. This makes it generally easy to do alias analysis and/or
/// SROA/mem2reg of the memory object.
static Instruction *InstCombineStoreToCast(InstCombiner &IC, StoreInst &SI) {
User *CI = cast<User>(SI.getOperand(1));
Value *CastOp = CI->getOperand(0);
Type *DestPTy = CI->getType()->getPointerElementType();
PointerType *SrcTy = dyn_cast<PointerType>(CastOp->getType());
if (!SrcTy) return nullptr;
Type *SrcPTy = SrcTy->getElementType();
if (!DestPTy->isIntegerTy() && !DestPTy->isPointerTy())
return nullptr;
/// NewGEPIndices - If SrcPTy is an aggregate type, we can emit a "noop gep"
/// to its first element. This allows us to handle things like:
/// store i32 xxx, (bitcast {foo*, float}* %P to i32*)
/// on 32-bit hosts.
SmallVector<Value*, 4> NewGEPIndices;
// If the source is an array, the code below will not succeed. Check to
// see if a trivial 'gep P, 0, 0' will help matters. Only do this for
// constants.
if (SrcPTy->isArrayTy() || SrcPTy->isStructTy()) {
// Index through pointer.
Constant *Zero = Constant::getNullValue(Type::getInt32Ty(SI.getContext()));
NewGEPIndices.push_back(Zero);
while (1) {
if (StructType *STy = dyn_cast<StructType>(SrcPTy)) {
if (!STy->getNumElements()) /* Struct can be empty {} */
break;
NewGEPIndices.push_back(Zero);
SrcPTy = STy->getElementType(0);
} else if (ArrayType *ATy = dyn_cast<ArrayType>(SrcPTy)) {
NewGEPIndices.push_back(Zero);
SrcPTy = ATy->getElementType();
} else {
break;
}
}
SrcTy = PointerType::get(SrcPTy, SrcTy->getAddressSpace());
}
if (!SrcPTy->isIntegerTy() && !SrcPTy->isPointerTy())
return nullptr;
// If the pointers point into different address spaces don't do the
// transformation.
if (SrcTy->getAddressSpace() != CI->getType()->getPointerAddressSpace())
return nullptr;
// If the pointers point to values of different sizes don't do the
// transformation.
if (!IC.getDataLayout() ||
IC.getDataLayout()->getTypeSizeInBits(SrcPTy) !=
IC.getDataLayout()->getTypeSizeInBits(DestPTy))
return nullptr;
// If the pointers point to pointers to different address spaces don't do the
// transformation. It is not safe to introduce an addrspacecast instruction in
// this case since, depending on the target, addrspacecast may not be a no-op
// cast.
if (SrcPTy->isPointerTy() && DestPTy->isPointerTy() &&
SrcPTy->getPointerAddressSpace() != DestPTy->getPointerAddressSpace())
return nullptr;
// Okay, we are casting from one integer or pointer type to another of
// the same size. Instead of casting the pointer before
// the store, cast the value to be stored.
Value *NewCast;
Instruction::CastOps opcode = Instruction::BitCast;
Type* CastSrcTy = DestPTy;
Type* CastDstTy = SrcPTy;
if (CastDstTy->isPointerTy()) {
if (CastSrcTy->isIntegerTy())
opcode = Instruction::IntToPtr;
} else if (CastDstTy->isIntegerTy()) {
if (CastSrcTy->isPointerTy())
opcode = Instruction::PtrToInt;
}
// SIOp0 is a pointer to aggregate and this is a store to the first field,
// emit a GEP to index into its first field.
if (!NewGEPIndices.empty())
CastOp = IC.Builder->CreateInBoundsGEP(CastOp, NewGEPIndices);
Value *SIOp0 = SI.getOperand(0);
NewCast = IC.Builder->CreateCast(opcode, SIOp0, CastDstTy,
SIOp0->getName()+".c");
SI.setOperand(0, NewCast);
SI.setOperand(1, CastOp);
return &SI;
}
/// Constants comparison:
/// 1. Check whether type of L constant could be losslessly bitcasted to R
/// type.
/// 2. Compare constant contents.
/// For more details see declaration comments.
int FunctionComparator::cmpConstants(const Constant *L,
const Constant *R) const {
Type *TyL = L->getType();
Type *TyR = R->getType();
// Check whether types are bitcastable. This part is just re-factored
// Type::canLosslesslyBitCastTo method, but instead of returning true/false,
// we also pack into result which type is "less" for us.
int TypesRes = cmpTypes(TyL, TyR);
if (TypesRes != 0) {
// Types are different, but check whether we can bitcast them.
if (!TyL->isFirstClassType()) {
if (TyR->isFirstClassType())
return -1;
// Neither TyL nor TyR are values of first class type. Return the result
// of comparing the types
return TypesRes;
}
if (!TyR->isFirstClassType()) {
if (TyL->isFirstClassType())
return 1;
return TypesRes;
}
// Vector -> Vector conversions are always lossless if the two vector types
// have the same size, otherwise not.
unsigned TyLWidth = 0;
unsigned TyRWidth = 0;
if (auto *VecTyL = dyn_cast<VectorType>(TyL))
TyLWidth = VecTyL->getBitWidth();
if (auto *VecTyR = dyn_cast<VectorType>(TyR))
TyRWidth = VecTyR->getBitWidth();
if (TyLWidth != TyRWidth)
return cmpNumbers(TyLWidth, TyRWidth);
// Zero bit-width means neither TyL nor TyR are vectors.
if (!TyLWidth) {
PointerType *PTyL = dyn_cast<PointerType>(TyL);
PointerType *PTyR = dyn_cast<PointerType>(TyR);
if (PTyL && PTyR) {
unsigned AddrSpaceL = PTyL->getAddressSpace();
unsigned AddrSpaceR = PTyR->getAddressSpace();
if (int Res = cmpNumbers(AddrSpaceL, AddrSpaceR))
return Res;
}
if (PTyL)
return 1;
if (PTyR)
return -1;
// TyL and TyR aren't vectors, nor pointers. We don't know how to
// bitcast them.
return TypesRes;
}
}
// OK, types are bitcastable, now check constant contents.
if (L->isNullValue() && R->isNullValue())
return TypesRes;
if (L->isNullValue() && !R->isNullValue())
return 1;
if (!L->isNullValue() && R->isNullValue())
return -1;
auto GlobalValueL = const_cast<GlobalValue*>(dyn_cast<GlobalValue>(L));
auto GlobalValueR = const_cast<GlobalValue*>(dyn_cast<GlobalValue>(R));
if (GlobalValueL && GlobalValueR) {
return cmpGlobalValues(GlobalValueL, GlobalValueR);
}
if (int Res = cmpNumbers(L->getValueID(), R->getValueID()))
return Res;
if (const auto *SeqL = dyn_cast<ConstantDataSequential>(L)) {
const auto *SeqR = cast<ConstantDataSequential>(R);
// This handles ConstantDataArray and ConstantDataVector. Note that we
// compare the two raw data arrays, which might differ depending on the host
// endianness. This isn't a problem though, because the endiness of a module
// will affect the order of the constants, but this order is the same
// for a given input module and host platform.
return cmpMem(SeqL->getRawDataValues(), SeqR->getRawDataValues());
}
switch (L->getValueID()) {
case Value::UndefValueVal:
case Value::ConstantTokenNoneVal:
return TypesRes;
case Value::ConstantIntVal: {
const APInt &LInt = cast<ConstantInt>(L)->getValue();
const APInt &RInt = cast<ConstantInt>(R)->getValue();
return cmpAPInts(LInt, RInt);
}
case Value::ConstantFPVal: {
const APFloat &LAPF = cast<ConstantFP>(L)->getValueAPF();
const APFloat &RAPF = cast<ConstantFP>(R)->getValueAPF();
return cmpAPFloats(LAPF, RAPF);
}
case Value::ConstantArrayVal: {
const ConstantArray *LA = cast<ConstantArray>(L);
const ConstantArray *RA = cast<ConstantArray>(R);
uint64_t NumElementsL = cast<ArrayType>(TyL)->getNumElements();
uint64_t NumElementsR = cast<ArrayType>(TyR)->getNumElements();
if (int Res = cmpNumbers(NumElementsL, NumElementsR))
return Res;
for (uint64_t i = 0; i < NumElementsL; ++i) {
if (int Res = cmpConstants(cast<Constant>(LA->getOperand(i)),
cast<Constant>(RA->getOperand(i))))
return Res;
}
return 0;
}
case Value::ConstantStructVal: {
const ConstantStruct *LS = cast<ConstantStruct>(L);
const ConstantStruct *RS = cast<ConstantStruct>(R);
unsigned NumElementsL = cast<StructType>(TyL)->getNumElements();
unsigned NumElementsR = cast<StructType>(TyR)->getNumElements();
if (int Res = cmpNumbers(NumElementsL, NumElementsR))
return Res;
for (unsigned i = 0; i != NumElementsL; ++i) {
if (int Res = cmpConstants(cast<Constant>(LS->getOperand(i)),
cast<Constant>(RS->getOperand(i))))
return Res;
}
return 0;
}
case Value::ConstantVectorVal: {
const ConstantVector *LV = cast<ConstantVector>(L);
const ConstantVector *RV = cast<ConstantVector>(R);
unsigned NumElementsL = cast<VectorType>(TyL)->getNumElements();
unsigned NumElementsR = cast<VectorType>(TyR)->getNumElements();
if (int Res = cmpNumbers(NumElementsL, NumElementsR))
return Res;
for (uint64_t i = 0; i < NumElementsL; ++i) {
if (int Res = cmpConstants(cast<Constant>(LV->getOperand(i)),
cast<Constant>(RV->getOperand(i))))
return Res;
}
return 0;
}
case Value::ConstantExprVal: {
const ConstantExpr *LE = cast<ConstantExpr>(L);
const ConstantExpr *RE = cast<ConstantExpr>(R);
unsigned NumOperandsL = LE->getNumOperands();
unsigned NumOperandsR = RE->getNumOperands();
if (int Res = cmpNumbers(NumOperandsL, NumOperandsR))
return Res;
for (unsigned i = 0; i < NumOperandsL; ++i) {
if (int Res = cmpConstants(cast<Constant>(LE->getOperand(i)),
cast<Constant>(RE->getOperand(i))))
return Res;
}
return 0;
}
case Value::BlockAddressVal: {
const BlockAddress *LBA = cast<BlockAddress>(L);
const BlockAddress *RBA = cast<BlockAddress>(R);
if (int Res = cmpValues(LBA->getFunction(), RBA->getFunction()))
return Res;
if (LBA->getFunction() == RBA->getFunction()) {
// They are BBs in the same function. Order by which comes first in the
// BB order of the function. This order is deterministic.
Function* F = LBA->getFunction();
BasicBlock *LBB = LBA->getBasicBlock();
BasicBlock *RBB = RBA->getBasicBlock();
if (LBB == RBB)
return 0;
for(BasicBlock &BB : F->getBasicBlockList()) {
if (&BB == LBB) {
assert(&BB != RBB);
return -1;
}
if (&BB == RBB)
return 1;
}
llvm_unreachable("Basic Block Address does not point to a basic block in "
"its function.");
return -1;
} else {
// cmpValues said the functions are the same. So because they aren't
// literally the same pointer, they must respectively be the left and
// right functions.
assert(LBA->getFunction() == FnL && RBA->getFunction() == FnR);
// cmpValues will tell us if these are equivalent BasicBlocks, in the
// context of their respective functions.
return cmpValues(LBA->getBasicBlock(), RBA->getBasicBlock());
}
}
default: // Unknown constant, abort.
DEBUG(dbgs() << "Looking at valueID " << L->getValueID() << "\n");
llvm_unreachable("Constant ValueID not recognized.");
return -1;
}
}
/// cmpType - compares two types,
/// defines total ordering among the types set.
/// See method declaration comments for more details.
int FunctionComparator::cmpTypes(Type *TyL, Type *TyR) const {
PointerType *PTyL = dyn_cast<PointerType>(TyL);
PointerType *PTyR = dyn_cast<PointerType>(TyR);
const DataLayout &DL = FnL->getParent()->getDataLayout();
if (PTyL && PTyL->getAddressSpace() == 0)
TyL = DL.getIntPtrType(TyL);
if (PTyR && PTyR->getAddressSpace() == 0)
TyR = DL.getIntPtrType(TyR);
if (TyL == TyR)
return 0;
if (int Res = cmpNumbers(TyL->getTypeID(), TyR->getTypeID()))
return Res;
switch (TyL->getTypeID()) {
default:
llvm_unreachable("Unknown type!");
// Fall through in Release mode.
LLVM_FALLTHROUGH;
case Type::IntegerTyID:
return cmpNumbers(cast<IntegerType>(TyL)->getBitWidth(),
cast<IntegerType>(TyR)->getBitWidth());
// TyL == TyR would have returned true earlier, because types are uniqued.
case Type::VoidTyID:
case Type::FloatTyID:
case Type::DoubleTyID:
case Type::X86_FP80TyID:
case Type::FP128TyID:
case Type::PPC_FP128TyID:
case Type::LabelTyID:
case Type::MetadataTyID:
case Type::TokenTyID:
return 0;
case Type::PointerTyID: {
assert(PTyL && PTyR && "Both types must be pointers here.");
return cmpNumbers(PTyL->getAddressSpace(), PTyR->getAddressSpace());
}
case Type::StructTyID: {
StructType *STyL = cast<StructType>(TyL);
StructType *STyR = cast<StructType>(TyR);
if (STyL->getNumElements() != STyR->getNumElements())
return cmpNumbers(STyL->getNumElements(), STyR->getNumElements());
if (STyL->isPacked() != STyR->isPacked())
return cmpNumbers(STyL->isPacked(), STyR->isPacked());
for (unsigned i = 0, e = STyL->getNumElements(); i != e; ++i) {
if (int Res = cmpTypes(STyL->getElementType(i), STyR->getElementType(i)))
return Res;
}
return 0;
}
case Type::FunctionTyID: {
FunctionType *FTyL = cast<FunctionType>(TyL);
FunctionType *FTyR = cast<FunctionType>(TyR);
if (FTyL->getNumParams() != FTyR->getNumParams())
return cmpNumbers(FTyL->getNumParams(), FTyR->getNumParams());
if (FTyL->isVarArg() != FTyR->isVarArg())
return cmpNumbers(FTyL->isVarArg(), FTyR->isVarArg());
if (int Res = cmpTypes(FTyL->getReturnType(), FTyR->getReturnType()))
return Res;
for (unsigned i = 0, e = FTyL->getNumParams(); i != e; ++i) {
if (int Res = cmpTypes(FTyL->getParamType(i), FTyR->getParamType(i)))
return Res;
}
return 0;
}
case Type::ArrayTyID:
case Type::VectorTyID: {
auto *STyL = cast<SequentialType>(TyL);
auto *STyR = cast<SequentialType>(TyR);
if (STyL->getNumElements() != STyR->getNumElements())
return cmpNumbers(STyL->getNumElements(), STyR->getNumElements());
return cmpTypes(STyL->getElementType(), STyR->getElementType());
}
}
}
GlobalAlias *GlobalAlias::create(LinkageTypes Link, const Twine &Name,
GlobalObject *Aliasee) {
PointerType *PTy = Aliasee->getType();
return create(PTy->getElementType(), PTy->getAddressSpace(), Link, Name,
Aliasee);
}
/// cmpType - compares two types,
/// defines total ordering among the types set.
/// See method declaration comments for more details.
int FunctionComparator::cmpType(Type *TyL, Type *TyR) const {
PointerType *PTyL = dyn_cast<PointerType>(TyL);
PointerType *PTyR = dyn_cast<PointerType>(TyR);
if (DL) {
if (PTyL && PTyL->getAddressSpace() == 0) TyL = DL->getIntPtrType(TyL);
if (PTyR && PTyR->getAddressSpace() == 0) TyR = DL->getIntPtrType(TyR);
}
if (TyL == TyR)
return 0;
if (int Res = cmpNumbers(TyL->getTypeID(), TyR->getTypeID()))
return Res;
switch (TyL->getTypeID()) {
default:
llvm_unreachable("Unknown type!");
// Fall through in Release mode.
case Type::IntegerTyID:
case Type::VectorTyID:
// TyL == TyR would have returned true earlier.
return cmpNumbers((uint64_t)TyL, (uint64_t)TyR);
case Type::VoidTyID:
case Type::FloatTyID:
case Type::DoubleTyID:
case Type::X86_FP80TyID:
case Type::FP128TyID:
case Type::PPC_FP128TyID:
case Type::LabelTyID:
case Type::MetadataTyID:
return 0;
case Type::PointerTyID: {
assert(PTyL && PTyR && "Both types must be pointers here.");
return cmpNumbers(PTyL->getAddressSpace(), PTyR->getAddressSpace());
}
case Type::StructTyID: {
StructType *STyL = cast<StructType>(TyL);
StructType *STyR = cast<StructType>(TyR);
if (STyL->getNumElements() != STyR->getNumElements())
return cmpNumbers(STyL->getNumElements(), STyR->getNumElements());
if (STyL->isPacked() != STyR->isPacked())
return cmpNumbers(STyL->isPacked(), STyR->isPacked());
for (unsigned i = 0, e = STyL->getNumElements(); i != e; ++i) {
if (int Res = cmpType(STyL->getElementType(i),
STyR->getElementType(i)))
return Res;
}
return 0;
}
case Type::FunctionTyID: {
FunctionType *FTyL = cast<FunctionType>(TyL);
FunctionType *FTyR = cast<FunctionType>(TyR);
if (FTyL->getNumParams() != FTyR->getNumParams())
return cmpNumbers(FTyL->getNumParams(), FTyR->getNumParams());
if (FTyL->isVarArg() != FTyR->isVarArg())
return cmpNumbers(FTyL->isVarArg(), FTyR->isVarArg());
if (int Res = cmpType(FTyL->getReturnType(), FTyR->getReturnType()))
return Res;
for (unsigned i = 0, e = FTyL->getNumParams(); i != e; ++i) {
if (int Res = cmpType(FTyL->getParamType(i), FTyR->getParamType(i)))
return Res;
}
return 0;
}
case Type::ArrayTyID: {
ArrayType *ATyL = cast<ArrayType>(TyL);
ArrayType *ATyR = cast<ArrayType>(TyR);
if (ATyL->getNumElements() != ATyR->getNumElements())
return cmpNumbers(ATyL->getNumElements(), ATyR->getNumElements());
return cmpType(ATyL->getElementType(), ATyR->getElementType());
}
}
}
unsigned GlobalValue::getAddressSpace() const {
PointerType *PtrTy = getType();
return PtrTy->getAddressSpace();
}