#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallPtrSet.h"
+#include "llvm/ADT/SmallSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/ConstantFolding.h"
#include "llvm/Analysis/MemoryBuiltins.h"
+#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/IR/CallSite.h"
#include "llvm/IR/CallingConv.h"
#include "llvm/IR/Constants.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/raw_ostream.h"
-#include "llvm/Target/TargetLibraryInfo.h"
#include "llvm/Transforms/Utils/CtorUtils.h"
#include "llvm/Transforms/Utils/GlobalStatus.h"
#include "llvm/Transforms/Utils/ModuleUtils.h"
namespace {
struct GlobalOpt : public ModulePass {
void getAnalysisUsage(AnalysisUsage &AU) const override {
- AU.addRequired<TargetLibraryInfo>();
+ AU.addRequired<TargetLibraryInfoWrapperPass>();
}
static char ID; // Pass identification, replacement for typeid
GlobalOpt() : ModulePass(ID) {
const GlobalStatus &GS);
bool OptimizeEmptyGlobalCXXDtors(Function *CXAAtExitFn);
- const DataLayout *DL;
TargetLibraryInfo *TLI;
+ SmallSet<const Comdat *, 8> NotDiscardableComdats;
};
}
char GlobalOpt::ID = 0;
INITIALIZE_PASS_BEGIN(GlobalOpt, "globalopt",
"Global Variable Optimizer", false, false)
-INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfo)
+INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfoWrapperPass)
INITIALIZE_PASS_END(GlobalOpt, "globalopt",
"Global Variable Optimizer", false, false)
/// quick scan over the use list to clean up the easy and obvious cruft. This
/// returns true if it made a change.
static bool CleanupConstantGlobalUsers(Value *V, Constant *Init,
- const DataLayout *DL,
+ const DataLayout &DL,
TargetLibraryInfo *TLI) {
bool Changed = false;
// Note that we need to use a weak value handle for the worklist items. When
// and will invalidate our notion of what Init is.
Constant *SubInit = nullptr;
if (!isa<ConstantExpr>(GEP->getOperand(0))) {
- ConstantExpr *CE =
- dyn_cast_or_null<ConstantExpr>(ConstantFoldInstruction(GEP, DL, TLI));
+ ConstantExpr *CE = dyn_cast_or_null<ConstantExpr>(
+ ConstantFoldInstruction(GEP, DL, TLI));
if (Init && CE && CE->getOpcode() == Instruction::GetElementPtr)
SubInit = ConstantFoldLoadThroughGEPConstantExpr(Init, CE);
In, GV->getName()+"."+Twine(i),
GV->getThreadLocalMode(),
GV->getType()->getAddressSpace());
- Globals.insert(GV, NGV);
+ Globals.insert(GV->getIterator(), NGV);
NewGlobals.push_back(NGV);
// Calculate the known alignment of the field. If the original aggregate
In, GV->getName()+"."+Twine(i),
GV->getThreadLocalMode(),
GV->getType()->getAddressSpace());
- Globals.insert(GV, NGV);
+ Globals.insert(GV->getIterator(), NGV);
NewGlobals.push_back(NGV);
// Calculate the known alignment of the field. If the original aggregate
if (NewGlobals.empty())
return nullptr;
- DEBUG(dbgs() << "PERFORMING GLOBAL SRA ON: " << *GV);
+ DEBUG(dbgs() << "PERFORMING GLOBAL SRA ON: " << *GV << "\n");
Constant *NullInt =Constant::getNullValue(Type::getInt32Ty(GV->getContext()));
if (Val >= NewGlobals.size()) Val = 0; // Out of bound array access.
Value *NewPtr = NewGlobals[Val];
+ Type *NewTy = NewGlobals[Val]->getValueType();
// Form a shorter GEP if needed.
if (GEP->getNumOperands() > 3) {
Idxs.push_back(NullInt);
for (unsigned i = 3, e = CE->getNumOperands(); i != e; ++i)
Idxs.push_back(CE->getOperand(i));
- NewPtr = ConstantExpr::getGetElementPtr(cast<Constant>(NewPtr), Idxs);
+ NewPtr =
+ ConstantExpr::getGetElementPtr(NewTy, cast<Constant>(NewPtr), Idxs);
} else {
GetElementPtrInst *GEPI = cast<GetElementPtrInst>(GEP);
SmallVector<Value*, 8> Idxs;
Idxs.push_back(NullInt);
for (unsigned i = 3, e = GEPI->getNumOperands(); i != e; ++i)
Idxs.push_back(GEPI->getOperand(i));
- NewPtr = GetElementPtrInst::Create(NewPtr, Idxs,
- GEPI->getName()+"."+Twine(Val),GEPI);
+ NewPtr = GetElementPtrInst::Create(
+ NewTy, NewPtr, Idxs, GEPI->getName() + "." + Twine(Val), GEPI);
}
}
GEP->replaceAllUsesWith(NewPtr);
/// value will trap if the value is dynamically null. PHIs keeps track of any
/// phi nodes we've seen to avoid reprocessing them.
static bool AllUsesOfValueWillTrapIfNull(const Value *V,
- SmallPtrSet<const PHINode*, 8> &PHIs) {
+ SmallPtrSetImpl<const PHINode*> &PHIs) {
for (const User *U : V->users())
if (isa<LoadInst>(U)) {
// Will trap.
} else if (const PHINode *PN = dyn_cast<PHINode>(U)) {
// If we've already seen this phi node, ignore it, it has already been
// checked.
- if (PHIs.insert(PN) && !AllUsesOfValueWillTrapIfNull(PN, PHIs))
+ if (PHIs.insert(PN).second && !AllUsesOfValueWillTrapIfNull(PN, PHIs))
return false;
} else if (isa<ICmpInst>(U) &&
isa<ConstantPointerNull>(U->getOperand(1))) {
else
break;
if (Idxs.size() == GEPI->getNumOperands()-1)
- Changed |= OptimizeAwayTrappingUsesOfValue(GEPI,
- ConstantExpr::getGetElementPtr(NewV, Idxs));
+ Changed |= OptimizeAwayTrappingUsesOfValue(
+ GEPI, ConstantExpr::getGetElementPtr(nullptr, NewV, Idxs));
if (GEPI->use_empty()) {
Changed = true;
GEPI->eraseFromParent();
/// if the loaded value is dynamically null, then we know that they cannot be
/// reachable with a null optimize away the load.
static bool OptimizeAwayTrappingUsesOfLoads(GlobalVariable *GV, Constant *LV,
- const DataLayout *DL,
+ const DataLayout &DL,
TargetLibraryInfo *TLI) {
bool Changed = false;
}
if (Changed) {
- DEBUG(dbgs() << "OPTIMIZED LOADS FROM STORED ONCE POINTER: " << *GV);
+ DEBUG(dbgs() << "OPTIMIZED LOADS FROM STORED ONCE POINTER: " << *GV << "\n");
++NumGlobUses;
}
/// ConstantPropUsersOf - Walk the use list of V, constant folding all of the
/// instructions that are foldable.
-static void ConstantPropUsersOf(Value *V, const DataLayout *DL,
+static void ConstantPropUsersOf(Value *V, const DataLayout &DL,
TargetLibraryInfo *TLI) {
for (Value::user_iterator UI = V->user_begin(), E = V->user_end(); UI != E; )
if (Instruction *I = dyn_cast<Instruction>(*UI++))
/// the specified malloc. Because it is always the result of the specified
/// malloc, there is no reason to actually DO the malloc. Instead, turn the
/// malloc into a global, and any loads of GV as uses of the new global.
-static GlobalVariable *OptimizeGlobalAddressOfMalloc(GlobalVariable *GV,
- CallInst *CI,
- Type *AllocTy,
- ConstantInt *NElements,
- const DataLayout *DL,
- TargetLibraryInfo *TLI) {
+static GlobalVariable *
+OptimizeGlobalAddressOfMalloc(GlobalVariable *GV, CallInst *CI, Type *AllocTy,
+ ConstantInt *NElements, const DataLayout &DL,
+ TargetLibraryInfo *TLI) {
DEBUG(errs() << "PROMOTING GLOBAL: " << *GV << " CALL = " << *CI << '\n');
Type *GlobalType;
cast<StoreInst>(InitBool->user_back())->eraseFromParent();
delete InitBool;
} else
- GV->getParent()->getGlobalList().insert(GV, InitBool);
+ GV->getParent()->getGlobalList().insert(GV->getIterator(), InitBool);
// Now the GV is dead, nuke it and the malloc..
GV->eraseFromParent();
/// it is to the specified global.
static bool ValueIsOnlyUsedLocallyOrStoredToOneGlobal(const Instruction *V,
const GlobalVariable *GV,
- SmallPtrSet<const PHINode*, 8> &PHIs) {
+ SmallPtrSetImpl<const PHINode*> &PHIs) {
for (const User *U : V->users()) {
const Instruction *Inst = cast<Instruction>(U);
if (const PHINode *PN = dyn_cast<PHINode>(Inst)) {
// PHIs are ok if all uses are ok. Don't infinitely recurse through PHI
// cycles.
- if (PHIs.insert(PN))
+ if (PHIs.insert(PN).second)
if (!ValueIsOnlyUsedLocallyOrStoredToOneGlobal(PN, GV, PHIs))
return false;
continue;
/// of a load) are simple enough to perform heap SRA on. This permits GEP's
/// that index through the array and struct field, icmps of null, and PHIs.
static bool LoadUsesSimpleEnoughForHeapSRA(const Value *V,
- SmallPtrSet<const PHINode*, 32> &LoadUsingPHIs,
- SmallPtrSet<const PHINode*, 32> &LoadUsingPHIsPerLoad) {
+ SmallPtrSetImpl<const PHINode*> &LoadUsingPHIs,
+ SmallPtrSetImpl<const PHINode*> &LoadUsingPHIsPerLoad) {
// We permit two users of the load: setcc comparing against the null
// pointer, and a getelementptr of a specific form.
for (const User *U : V->users()) {
}
if (const PHINode *PN = dyn_cast<PHINode>(UI)) {
- if (!LoadUsingPHIsPerLoad.insert(PN))
+ if (!LoadUsingPHIsPerLoad.insert(PN).second)
// This means some phi nodes are dependent on each other.
// Avoid infinite looping!
return false;
- if (!LoadUsingPHIs.insert(PN))
+ if (!LoadUsingPHIs.insert(PN).second)
// If we have already analyzed this PHI, then it is safe.
continue;
// that all inputs the to the PHI nodes are in the same equivalence sets.
// Check to verify that all operands of the PHIs are either PHIS that can be
// transformed, loads from GV, or MI itself.
- for (SmallPtrSet<const PHINode*, 32>::const_iterator I = LoadUsingPHIs.begin()
- , E = LoadUsingPHIs.end(); I != E; ++I) {
- const PHINode *PN = *I;
+ for (const PHINode *PN : LoadUsingPHIs) {
for (unsigned op = 0, e = PN->getNumIncomingValues(); op != e; ++op) {
Value *InVal = PN->getIncomingValue(op);
InsertedScalarizedValues,
PHIsToRewrite),
LI->getName()+".f"+Twine(FieldNo), LI);
- } else if (PHINode *PN = dyn_cast<PHINode>(V)) {
+ } else {
+ PHINode *PN = cast<PHINode>(V);
// PN's type is pointer to struct. Make a new PHI of pointer to struct
// field.
PN->getName()+".f"+Twine(FieldNo), PN);
Result = NewPN;
PHIsToRewrite.push_back(std::make_pair(PN, FieldNo));
- } else {
- llvm_unreachable("Unknown usable value");
}
return FieldVals[FieldNo] = Result;
GEPIdx.push_back(GEPI->getOperand(1));
GEPIdx.append(GEPI->op_begin()+3, GEPI->op_end());
- Value *NGEPI = GetElementPtrInst::Create(NewPtr, GEPIdx,
+ Value *NGEPI = GetElementPtrInst::Create(GEPI->getResultElementType(), NewPtr, GEPIdx,
GEPI->getName(), GEPI);
GEPI->replaceAllUsesWith(NGEPI);
GEPI->eraseFromParent();
/// PerformHeapAllocSRoA - CI is an allocation of an array of structures. Break
/// it up into multiple allocations of arrays of the fields.
static GlobalVariable *PerformHeapAllocSRoA(GlobalVariable *GV, CallInst *CI,
- Value *NElems, const DataLayout *DL,
+ Value *NElems, const DataLayout &DL,
const TargetLibraryInfo *TLI) {
DEBUG(dbgs() << "SROA HEAP ALLOC: " << *GV << " MALLOC = " << *CI << '\n');
Type *MAT = getMallocAllocatedType(CI, TLI);
GV->getThreadLocalMode());
FieldGlobals.push_back(NGV);
- unsigned TypeSize = DL->getTypeAllocSize(FieldTy);
+ unsigned TypeSize = DL.getTypeAllocSize(FieldTy);
if (StructType *ST = dyn_cast<StructType>(FieldTy))
- TypeSize = DL->getStructLayout(ST)->getSizeInBytes();
- Type *IntPtrTy = DL->getIntPtrType(CI->getType());
+ TypeSize = DL.getStructLayout(ST)->getSizeInBytes();
+ Type *IntPtrTy = DL.getIntPtrType(CI->getType());
Value *NMI = CallInst::CreateMalloc(CI, IntPtrTy, FieldTy,
ConstantInt::get(IntPtrTy, TypeSize),
NElems, nullptr,
// Split the basic block at the old malloc.
BasicBlock *OrigBB = CI->getParent();
- BasicBlock *ContBB = OrigBB->splitBasicBlock(CI, "malloc_cont");
+ BasicBlock *ContBB =
+ OrigBB->splitBasicBlock(CI->getIterator(), "malloc_cont");
// Create the block to check the first condition. Put all these blocks at the
// end of the function as they are unlikely to be executed.
/// TryToOptimizeStoreOfMallocToGlobal - This function is called when we see a
/// pointer global variable with a single value stored it that is a malloc or
/// cast of malloc.
-static bool TryToOptimizeStoreOfMallocToGlobal(GlobalVariable *GV,
- CallInst *CI,
+static bool TryToOptimizeStoreOfMallocToGlobal(GlobalVariable *GV, CallInst *CI,
Type *AllocTy,
AtomicOrdering Ordering,
Module::global_iterator &GVI,
- const DataLayout *DL,
+ const DataLayout &DL,
TargetLibraryInfo *TLI) {
- if (!DL)
- return false;
-
// If this is a malloc of an abstract type, don't touch it.
if (!AllocTy->isSized())
return false;
// Restrict this transformation to only working on small allocations
// (2048 bytes currently), as we don't want to introduce a 16M global or
// something.
- if (NElements->getZExtValue() * DL->getTypeAllocSize(AllocTy) < 2048) {
- GVI = OptimizeGlobalAddressOfMalloc(GV, CI, AllocTy, NElements, DL, TLI);
+ if (NElements->getZExtValue() * DL.getTypeAllocSize(AllocTy) < 2048) {
+ GVI = OptimizeGlobalAddressOfMalloc(GV, CI, AllocTy, NElements, DL, TLI)
+ ->getIterator();
return true;
}
// If this is a fixed size array, transform the Malloc to be an alloc of
// structs. malloc [100 x struct],1 -> malloc struct, 100
if (ArrayType *AT = dyn_cast<ArrayType>(getMallocAllocatedType(CI, TLI))) {
- Type *IntPtrTy = DL->getIntPtrType(CI->getType());
- unsigned TypeSize = DL->getStructLayout(AllocSTy)->getSizeInBytes();
+ Type *IntPtrTy = DL.getIntPtrType(CI->getType());
+ unsigned TypeSize = DL.getStructLayout(AllocSTy)->getSizeInBytes();
Value *AllocSize = ConstantInt::get(IntPtrTy, TypeSize);
Value *NumElements = ConstantInt::get(IntPtrTy, AT->getNumElements());
Instruction *Malloc = CallInst::CreateMalloc(CI, IntPtrTy, AllocSTy,
}
GVI = PerformHeapAllocSRoA(GV, CI, getMallocArraySize(CI, DL, TLI, true),
- DL, TLI);
+ DL, TLI)
+ ->getIterator();
return true;
}
static bool OptimizeOnceStoredGlobal(GlobalVariable *GV, Value *StoredOnceVal,
AtomicOrdering Ordering,
Module::global_iterator &GVI,
- const DataLayout *DL,
+ const DataLayout &DL,
TargetLibraryInfo *TLI) {
// Ignore no-op GEPs and bitcasts.
StoredOnceVal = StoredOnceVal->stripPointerCasts();
if (!isa<LoadInst>(U) && !isa<StoreInst>(U))
return false;
- DEBUG(dbgs() << " *** SHRINKING TO BOOL: " << *GV);
+ DEBUG(dbgs() << " *** SHRINKING TO BOOL: " << *GV << "\n");
// Create the new global, initializing it to false.
GlobalVariable *NewGV = new GlobalVariable(Type::getInt1Ty(GV->getContext()),
GV->getName()+".b",
GV->getThreadLocalMode(),
GV->getType()->getAddressSpace());
- GV->getParent()->getGlobalList().insert(GV, NewGV);
+ GV->getParent()->getGlobalList().insert(GV->getIterator(), NewGV);
Constant *InitVal = GV->getInitializer();
assert(InitVal->getType() != Type::getInt1Ty(GV->getContext()) &&
/// possible. If we make a change, return true.
bool GlobalOpt::ProcessGlobal(GlobalVariable *GV,
Module::global_iterator &GVI) {
- if (!GV->isDiscardableIfUnused())
- return false;
-
// Do more involved optimizations if the global is internal.
GV->removeDeadConstantUsers();
if (GV->use_empty()) {
- DEBUG(dbgs() << "GLOBAL DEAD: " << *GV);
+ DEBUG(dbgs() << "GLOBAL DEAD: " << *GV << "\n");
GV->eraseFromParent();
++NumDeleted;
return true;
bool GlobalOpt::ProcessInternalGlobal(GlobalVariable *GV,
Module::global_iterator &GVI,
const GlobalStatus &GS) {
+ auto &DL = GV->getParent()->getDataLayout();
// If this is a first class global and has only one accessing function
// and this function is main (which we know is not recursive), we replace
// the global with a local alloca in this function.
++NumMarked;
return true;
} else if (!GV->getInitializer()->getType()->isSingleValueType()) {
- if (DataLayoutPass *DLP = getAnalysisIfAvailable<DataLayoutPass>()) {
- const DataLayout &DL = DLP->getDataLayout();
- if (GlobalVariable *FirstNewGV = SRAGlobal(GV, DL)) {
- GVI = FirstNewGV; // Don't skip the newly produced globals!
- return true;
- }
+ const DataLayout &DL = GV->getParent()->getDataLayout();
+ if (GlobalVariable *FirstNewGV = SRAGlobal(GV, DL)) {
+ GVI = FirstNewGV->getIterator(); // Don't skip the newly produced globals!
+ return true;
}
- } else if (GS.StoredType == GlobalStatus::StoredOnce) {
+ } else if (GS.StoredType == GlobalStatus::StoredOnce && GS.StoredOnceValue) {
// If the initial value for the global was an undef value, and if only
// one other value was stored into it, we can just change the
// initializer to be the stored value, then delete all stores to the
GV->eraseFromParent();
++NumDeleted;
} else {
- GVI = GV;
+ GVI = GV->getIterator();
}
++NumSubstitute;
return true;
bool Changed = false;
// Optimize functions.
for (Module::iterator FI = M.begin(), E = M.end(); FI != E; ) {
- Function *F = FI++;
+ Function *F = &*FI++;
// Functions without names cannot be referenced outside this module.
- if (!F->hasName() && !F->isDeclaration())
+ if (!F->hasName() && !F->isDeclaration() && !F->hasLocalLinkage())
F->setLinkage(GlobalValue::InternalLinkage);
+
+ const Comdat *C = F->getComdat();
+ bool inComdat = C && NotDiscardableComdats.count(C);
F->removeDeadConstantUsers();
- if (F->isDefTriviallyDead()) {
+ if ((!inComdat || F->hasLocalLinkage()) && F->isDefTriviallyDead()) {
F->eraseFromParent();
Changed = true;
++NumFnDeleted;
bool GlobalOpt::OptimizeGlobalVars(Module &M) {
bool Changed = false;
+
for (Module::global_iterator GVI = M.global_begin(), E = M.global_end();
GVI != E; ) {
- GlobalVariable *GV = GVI++;
+ GlobalVariable *GV = &*GVI++;
// Global variables without names cannot be referenced outside this module.
- if (!GV->hasName() && !GV->isDeclaration())
+ if (!GV->hasName() && !GV->isDeclaration() && !GV->hasLocalLinkage())
GV->setLinkage(GlobalValue::InternalLinkage);
// Simplify the initializer.
if (GV->hasInitializer())
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(GV->getInitializer())) {
+ auto &DL = M.getDataLayout();
Constant *New = ConstantFoldConstantExpression(CE, DL, TLI);
if (New && New != CE)
GV->setInitializer(New);
}
- Changed |= ProcessGlobal(GV, GVI);
+ if (GV->isDiscardableIfUnused()) {
+ if (const Comdat *C = GV->getComdat())
+ if (NotDiscardableComdats.count(C) && !GV->hasLocalLinkage())
+ continue;
+ Changed |= ProcessGlobal(GV, GVI);
+ }
}
return Changed;
}
static inline bool
isSimpleEnoughValueToCommit(Constant *C,
- SmallPtrSet<Constant*, 8> &SimpleConstants,
- const DataLayout *DL);
-
+ SmallPtrSetImpl<Constant *> &SimpleConstants,
+ const DataLayout &DL);
/// isSimpleEnoughValueToCommit - Return true if the specified constant can be
/// handled by the code generator. We don't want to generate something like:
/// This function should be called if C was not found (but just got inserted)
/// in SimpleConstants to avoid having to rescan the same constants all the
/// time.
-static bool isSimpleEnoughValueToCommitHelper(Constant *C,
- SmallPtrSet<Constant*, 8> &SimpleConstants,
- const DataLayout *DL) {
- // Simple integer, undef, constant aggregate zero, global addresses, etc are
- // all supported.
- if (C->getNumOperands() == 0 || isa<BlockAddress>(C) ||
- isa<GlobalValue>(C))
+static bool
+isSimpleEnoughValueToCommitHelper(Constant *C,
+ SmallPtrSetImpl<Constant *> &SimpleConstants,
+ const DataLayout &DL) {
+ // Simple global addresses are supported, do not allow dllimport or
+ // thread-local globals.
+ if (auto *GV = dyn_cast<GlobalValue>(C))
+ return !GV->hasDLLImportStorageClass() && !GV->isThreadLocal();
+
+ // Simple integer, undef, constant aggregate zero, etc are all supported.
+ if (C->getNumOperands() == 0 || isa<BlockAddress>(C))
return true;
// Aggregate values are safe if all their elements are.
if (isa<ConstantArray>(C) || isa<ConstantStruct>(C) ||
isa<ConstantVector>(C)) {
- for (unsigned i = 0, e = C->getNumOperands(); i != e; ++i) {
- Constant *Op = cast<Constant>(C->getOperand(i));
- if (!isSimpleEnoughValueToCommit(Op, SimpleConstants, DL))
+ for (Value *Op : C->operands())
+ if (!isSimpleEnoughValueToCommit(cast<Constant>(Op), SimpleConstants, DL))
return false;
- }
return true;
}
case Instruction::PtrToInt:
// int <=> ptr is fine if the int type is the same size as the
// pointer type.
- if (!DL || DL->getTypeSizeInBits(CE->getType()) !=
- DL->getTypeSizeInBits(CE->getOperand(0)->getType()))
+ if (DL.getTypeSizeInBits(CE->getType()) !=
+ DL.getTypeSizeInBits(CE->getOperand(0)->getType()))
return false;
return isSimpleEnoughValueToCommit(CE->getOperand(0), SimpleConstants, DL);
static inline bool
isSimpleEnoughValueToCommit(Constant *C,
- SmallPtrSet<Constant*, 8> &SimpleConstants,
- const DataLayout *DL) {
+ SmallPtrSetImpl<Constant *> &SimpleConstants,
+ const DataLayout &DL) {
// If we already checked this constant, we win.
- if (!SimpleConstants.insert(C)) return true;
+ if (!SimpleConstants.insert(C).second)
+ return true;
// Check the constant.
return isSimpleEnoughValueToCommitHelper(C, SimpleConstants, DL);
}
return false;
if (GlobalVariable *GV = dyn_cast<GlobalVariable>(C))
- // Do not allow weak/*_odr/linkonce/dllimport/dllexport linkage or
- // external globals.
+ // Do not allow weak/*_odr/linkonce linkage or external globals.
return GV->hasUniqueInitializer();
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
/// Once an evaluation call fails, the evaluation object should not be reused.
class Evaluator {
public:
- Evaluator(const DataLayout *DL, const TargetLibraryInfo *TLI)
- : DL(DL), TLI(TLI) {
+ Evaluator(const DataLayout &DL, const TargetLibraryInfo *TLI)
+ : DL(DL), TLI(TLI) {
ValueStack.emplace_back();
}
return MutatedMemory;
}
- const SmallPtrSet<GlobalVariable*, 8> &getInvariants() const {
+ const SmallPtrSetImpl<GlobalVariable*> &getInvariants() const {
return Invariants;
}
/// simple enough to live in a static initializer of a global.
SmallPtrSet<Constant*, 8> SimpleConstants;
- const DataLayout *DL;
+ const DataLayout &DL;
const TargetLibraryInfo *TLI;
};
Constant *IdxZero = ConstantInt::get(IdxTy, 0, false);
Constant * const IdxList[] = {IdxZero, IdxZero};
- Ptr = ConstantExpr::getGetElementPtr(Ptr, IdxList);
+ Ptr = ConstantExpr::getGetElementPtr(nullptr, Ptr, IdxList);
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Ptr))
Ptr = ConstantFoldConstantExpression(CE, DL, TLI);
getVal(SI->getOperand(2)));
DEBUG(dbgs() << "Found a Select! Simplifying: " << *InstResult
<< "\n");
+ } else if (auto *EVI = dyn_cast<ExtractValueInst>(CurInst)) {
+ InstResult = ConstantExpr::getExtractValue(
+ getVal(EVI->getAggregateOperand()), EVI->getIndices());
+ DEBUG(dbgs() << "Found an ExtractValueInst! Simplifying: " << *InstResult
+ << "\n");
+ } else if (auto *IVI = dyn_cast<InsertValueInst>(CurInst)) {
+ InstResult = ConstantExpr::getInsertValue(
+ getVal(IVI->getAggregateOperand()),
+ getVal(IVI->getInsertedValueOperand()), IVI->getIndices());
+ DEBUG(dbgs() << "Found an InsertValueInst! Simplifying: " << *InstResult
+ << "\n");
} else if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(CurInst)) {
Constant *P = getVal(GEP->getOperand(0));
SmallVector<Constant*, 8> GEPOps;
i != e; ++i)
GEPOps.push_back(getVal(*i));
InstResult =
- ConstantExpr::getGetElementPtr(P, GEPOps,
- cast<GEPOperator>(GEP)->isInBounds());
+ ConstantExpr::getGetElementPtr(GEP->getSourceElementType(), P, GEPOps,
+ cast<GEPOperator>(GEP)->isInBounds());
DEBUG(dbgs() << "Found a GEP! Simplifying: " << *InstResult
<< "\n");
} else if (LoadInst *LI = dyn_cast<LoadInst>(CurInst)) {
InstResult = AllocaTmps.back().get();
DEBUG(dbgs() << "Found an alloca. Result: " << *InstResult << "\n");
} else if (isa<CallInst>(CurInst) || isa<InvokeInst>(CurInst)) {
- CallSite CS(CurInst);
+ CallSite CS(&*CurInst);
// Debug info can safely be ignored here.
if (isa<DbgInfoIntrinsic>(CS.getInstruction())) {
Value *Ptr = PtrArg->stripPointerCasts();
if (GlobalVariable *GV = dyn_cast<GlobalVariable>(Ptr)) {
Type *ElemTy = cast<PointerType>(GV->getType())->getElementType();
- if (DL && !Size->isAllOnesValue() &&
+ if (!Size->isAllOnesValue() &&
Size->getValue().getLimitedValue() >=
- DL->getTypeStoreSize(ElemTy)) {
+ DL.getTypeStoreSize(ElemTy)) {
Invariants.insert(GV);
DEBUG(dbgs() << "Found a global var that is an invariant: " << *GV
<< "\n");
// Continue even if we do nothing.
++CurInst;
continue;
+ } else if (II->getIntrinsicID() == Intrinsic::assume) {
+ DEBUG(dbgs() << "Skipping assume intrinsic.\n");
+ ++CurInst;
+ continue;
}
DEBUG(dbgs() << "Unknown intrinsic. Can not evaluate.\n");
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(InstResult))
InstResult = ConstantFoldConstantExpression(CE, DL, TLI);
- setVal(CurInst, InstResult);
+ setVal(&*CurInst, InstResult);
}
// If we just processed an invoke, we finished evaluating the block.
unsigned ArgNo = 0;
for (Function::arg_iterator AI = F->arg_begin(), E = F->arg_end(); AI != E;
++AI, ++ArgNo)
- setVal(AI, ActualArgs[ArgNo]);
+ setVal(&*AI, ActualArgs[ArgNo]);
// ExecutedBlocks - We only handle non-looping, non-recursive code. As such,
// we can only evaluate any one basic block at most once. This set keeps
SmallPtrSet<BasicBlock*, 32> ExecutedBlocks;
// CurBB - The current basic block we're evaluating.
- BasicBlock *CurBB = F->begin();
+ BasicBlock *CurBB = &F->front();
BasicBlock::iterator CurInst = CurBB->begin();
// Okay, we succeeded in evaluating this control flow. See if we have
// executed the new block before. If so, we have a looping function,
// which we cannot evaluate in reasonable time.
- if (!ExecutedBlocks.insert(NextBB))
+ if (!ExecutedBlocks.insert(NextBB).second)
return false; // looped!
// Okay, we have never been in this block before. Check to see if there
/// EvaluateStaticConstructor - Evaluate static constructors in the function, if
/// we can. Return true if we can, false otherwise.
-static bool EvaluateStaticConstructor(Function *F, const DataLayout *DL,
+static bool EvaluateStaticConstructor(Function *F, const DataLayout &DL,
const TargetLibraryInfo *TLI) {
// Call the function.
Evaluator Eval(DL, TLI);
Eval.getMutatedMemory().begin(), E = Eval.getMutatedMemory().end();
I != E; ++I)
CommitValueTo(I->second, I->first);
- for (SmallPtrSet<GlobalVariable*, 8>::const_iterator I =
- Eval.getInvariants().begin(), E = Eval.getInvariants().end();
- I != E; ++I)
- (*I)->setConstant(true);
+ for (GlobalVariable *GV : Eval.getInvariants())
+ GV->setConstant(true);
}
return EvalSuccess;
}
static int compareNames(Constant *const *A, Constant *const *B) {
- return (*A)->getName().compare((*B)->getName());
+ return (*A)->stripPointerCasts()->getName().compare(
+ (*B)->stripPointerCasts()->getName());
}
static void setUsedInitializer(GlobalVariable &V,
- SmallPtrSet<GlobalValue *, 8> Init) {
+ const SmallPtrSet<GlobalValue *, 8> &Init) {
if (Init.empty()) {
V.eraseFromParent();
return;
PointerType *Int8PtrTy = Type::getInt8PtrTy(V.getContext(), 0);
SmallVector<llvm::Constant *, 8> UsedArray;
- for (SmallPtrSet<GlobalValue *, 8>::iterator I = Init.begin(), E = Init.end();
- I != E; ++I) {
+ for (GlobalValue *GV : Init) {
Constant *Cast
- = ConstantExpr::getPointerBitCastOrAddrSpaceCast(*I, Int8PtrTy);
+ = ConstantExpr::getPointerBitCastOrAddrSpaceCast(GV, Int8PtrTy);
UsedArray.push_back(Cast);
}
// Sort to get deterministic order.
CompilerUsedV = collectUsedGlobalVariables(M, CompilerUsed, true);
}
typedef SmallPtrSet<GlobalValue *, 8>::iterator iterator;
+ typedef iterator_range<iterator> used_iterator_range;
iterator usedBegin() { return Used.begin(); }
iterator usedEnd() { return Used.end(); }
+ used_iterator_range used() {
+ return used_iterator_range(usedBegin(), usedEnd());
+ }
iterator compilerUsedBegin() { return CompilerUsed.begin(); }
iterator compilerUsedEnd() { return CompilerUsed.end(); }
+ used_iterator_range compilerUsed() {
+ return used_iterator_range(compilerUsedBegin(), compilerUsedEnd());
+ }
bool usedCount(GlobalValue *GV) const { return Used.count(GV); }
bool compilerUsedCount(GlobalValue *GV) const {
return CompilerUsed.count(GV);
}
bool usedErase(GlobalValue *GV) { return Used.erase(GV); }
bool compilerUsedErase(GlobalValue *GV) { return CompilerUsed.erase(GV); }
- bool usedInsert(GlobalValue *GV) { return Used.insert(GV); }
- bool compilerUsedInsert(GlobalValue *GV) { return CompilerUsed.insert(GV); }
+ bool usedInsert(GlobalValue *GV) { return Used.insert(GV).second; }
+ bool compilerUsedInsert(GlobalValue *GV) {
+ return CompilerUsed.insert(GV).second;
+ }
void syncVariablesAndSets() {
if (UsedV)
return U.usedCount(&GA) || U.compilerUsedCount(&GA);
}
-static bool hasUsesToReplace(GlobalAlias &GA, LLVMUsed &U, bool &RenameTarget) {
+static bool hasUsesToReplace(GlobalAlias &GA, const LLVMUsed &U,
+ bool &RenameTarget) {
RenameTarget = false;
bool Ret = false;
if (hasUseOtherThanLLVMUsed(GA, U))
bool Changed = false;
LLVMUsed Used(M);
- for (SmallPtrSet<GlobalValue *, 8>::iterator I = Used.usedBegin(),
- E = Used.usedEnd();
- I != E; ++I)
- Used.compilerUsedErase(*I);
+ for (GlobalValue *GV : Used.used())
+ Used.compilerUsedErase(GV);
for (Module::alias_iterator I = M.alias_begin(), E = M.alias_end();
I != E;) {
Module::alias_iterator J = I++;
// Aliases without names cannot be referenced outside this module.
- if (!J->hasName() && !J->isDeclaration())
+ if (!J->hasName() && !J->isDeclaration() && !J->hasLocalLinkage())
J->setLinkage(GlobalValue::InternalLinkage);
// If the aliasee may change at link time, nothing can be done - bail out.
if (J->mayBeOverridden())
continue;
Constant *Aliasee = J->getAliasee();
- GlobalValue *Target = cast<GlobalValue>(Aliasee->stripPointerCasts());
+ GlobalValue *Target = dyn_cast<GlobalValue>(Aliasee->stripPointerCasts());
+ // We can't trivially replace the alias with the aliasee if the aliasee is
+ // non-trivial in some way.
+ // TODO: Try to handle non-zero GEPs of local aliasees.
+ if (!Target)
+ continue;
Target->removeDeadConstantUsers();
// Make all users of the alias use the aliasee instead.
if (RenameTarget) {
// Give the aliasee the name, linkage and other attributes of the alias.
- Target->takeName(J);
+ Target->takeName(&*J);
Target->setLinkage(J->getLinkage());
Target->setVisibility(J->getVisibility());
Target->setDLLStorageClass(J->getDLLStorageClass());
- if (Used.usedErase(J))
+ if (Used.usedErase(&*J))
Used.usedInsert(Target);
- if (Used.compilerUsedErase(J))
+ if (Used.compilerUsedErase(&*J))
Used.compilerUsedInsert(Target);
} else if (mayHaveOtherReferences(*J, Used))
continue;
SmallPtrSet<const Function *, 8> NewCalledFunctions(CalledFunctions);
// Don't treat recursive functions as empty.
- if (!NewCalledFunctions.insert(CalledFn))
+ if (!NewCalledFunctions.insert(CalledFn).second)
return false;
if (!cxxDtorIsEmpty(*CalledFn, NewCalledFunctions))
bool GlobalOpt::runOnModule(Module &M) {
bool Changed = false;
- DataLayoutPass *DLP = getAnalysisIfAvailable<DataLayoutPass>();
- DL = DLP ? &DLP->getDataLayout() : nullptr;
- TLI = &getAnalysis<TargetLibraryInfo>();
+ auto &DL = M.getDataLayout();
+ TLI = &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
bool LocalChange = true;
while (LocalChange) {
LocalChange = false;
+ NotDiscardableComdats.clear();
+ for (const GlobalVariable &GV : M.globals())
+ if (const Comdat *C = GV.getComdat())
+ if (!GV.isDiscardableIfUnused() || !GV.use_empty())
+ NotDiscardableComdats.insert(C);
+ for (Function &F : M)
+ if (const Comdat *C = F.getComdat())
+ if (!F.isDefTriviallyDead())
+ NotDiscardableComdats.insert(C);
+ for (GlobalAlias &GA : M.aliases())
+ if (const Comdat *C = GA.getComdat())
+ if (!GA.isDiscardableIfUnused() || !GA.use_empty())
+ NotDiscardableComdats.insert(C);
+
// Delete functions that are trivially dead, ccc -> fastcc
LocalChange |= OptimizeFunctions(M);
return Changed;
}
+
projects / oota-llvm.git / blobdiff