// Calculate Edge Weights using "Loop Branch Heuristics". Predict backedges
// as taken, exiting edges as not-taken.
bool BranchProbabilityInfo::calcLoopBranchHeuristics(BasicBlock *BB) {
Loop *L = LI->getLoopFor(BB);
if (!L)
return false;
SmallPtrSet<BasicBlock *, 8> BackEdges;
SmallPtrSet<BasicBlock *, 8> ExitingEdges;
SmallPtrSet<BasicBlock *, 8> InEdges; // Edges from header to the loop.
for (succ_iterator I = succ_begin(BB), E = succ_end(BB); I != E; ++I) {
if (!L->contains(*I))
ExitingEdges.insert(*I);
else if (L->getHeader() == *I)
BackEdges.insert(*I);
else
InEdges.insert(*I);
}
if (uint32_t numBackEdges = BackEdges.size()) {
uint32_t backWeight = LBH_TAKEN_WEIGHT / numBackEdges;
if (backWeight < NORMAL_WEIGHT)
backWeight = NORMAL_WEIGHT;
for (SmallPtrSet<BasicBlock *, 8>::iterator EI = BackEdges.begin(),
EE = BackEdges.end(); EI != EE; ++EI) {
BasicBlock *Back = *EI;
setEdgeWeight(BB, Back, backWeight);
}
}
if (uint32_t numInEdges = InEdges.size()) {
uint32_t inWeight = LBH_TAKEN_WEIGHT / numInEdges;
if (inWeight < NORMAL_WEIGHT)
inWeight = NORMAL_WEIGHT;
for (SmallPtrSet<BasicBlock *, 8>::iterator EI = InEdges.begin(),
EE = InEdges.end(); EI != EE; ++EI) {
BasicBlock *Back = *EI;
setEdgeWeight(BB, Back, inWeight);
}
}
if (uint32_t numExitingEdges = ExitingEdges.size()) {
uint32_t exitWeight = LBH_NONTAKEN_WEIGHT / numExitingEdges;
if (exitWeight < MIN_WEIGHT)
exitWeight = MIN_WEIGHT;
for (SmallPtrSet<BasicBlock *, 8>::iterator EI = ExitingEdges.begin(),
EE = ExitingEdges.end(); EI != EE; ++EI) {
BasicBlock *Exiting = *EI;
setEdgeWeight(BB, Exiting, exitWeight);
}
}
return true;
}
TEST(SmallPtrSetTest, GrowthTest) {
int i;
int buf[8];
for(i=0; i<8; ++i) buf[i]=0;
SmallPtrSet<int *, 4> s;
typedef SmallPtrSet<int *, 4>::iterator iter;
s.insert(&buf[0]);
s.insert(&buf[1]);
s.insert(&buf[2]);
s.insert(&buf[3]);
EXPECT_EQ(4U, s.size());
i = 0;
for(iter I=s.begin(), E=s.end(); I!=E; ++I, ++i)
(**I)++;
EXPECT_EQ(4, i);
for(i=0; i<8; ++i)
EXPECT_EQ(i<4?1:0,buf[i]);
s.insert(&buf[4]);
s.insert(&buf[5]);
s.insert(&buf[6]);
s.insert(&buf[7]);
i = 0;
for(iter I=s.begin(), E=s.end(); I!=E; ++I, ++i)
(**I)++;
EXPECT_EQ(8, i);
s.erase(&buf[4]);
s.erase(&buf[5]);
s.erase(&buf[6]);
s.erase(&buf[7]);
EXPECT_EQ(4U, s.size());
i = 0;
for(iter I=s.begin(), E=s.end(); I!=E; ++I, ++i)
(**I)++;
EXPECT_EQ(4, i);
for(i=0; i<8; ++i)
EXPECT_EQ(i<4?3:1,buf[i]);
s.clear();
for(i=0; i<8; ++i) buf[i]=0;
for(i=0; i<128; ++i) s.insert(&buf[i%8]); // test repeated entires
EXPECT_EQ(8U, s.size());
for(iter I=s.begin(), E=s.end(); I!=E; ++I, ++i)
(**I)++;
for(i=0; i<8; ++i)
EXPECT_EQ(1,buf[i]);
}
void LTOCodeGenerator::applyScopeRestrictions() {
if (ScopeRestrictionsDone)
return;
Module *mergedModule = Linker.getModule();
// Start off with a verification pass.
PassManager passes;
passes.add(createVerifierPass());
// mark which symbols can not be internalized
Mangler Mangler(TargetMach);
std::vector<const char*> MustPreserveList;
SmallPtrSet<GlobalValue*, 8> AsmUsed;
std::vector<StringRef> Libcalls;
TargetLibraryInfo TLI(Triple(TargetMach->getTargetTriple()));
accumulateAndSortLibcalls(Libcalls, TLI, TargetMach->getTargetLowering());
for (Module::iterator f = mergedModule->begin(),
e = mergedModule->end(); f != e; ++f)
applyRestriction(*f, Libcalls, MustPreserveList, AsmUsed, Mangler);
for (Module::global_iterator v = mergedModule->global_begin(),
e = mergedModule->global_end(); v != e; ++v)
applyRestriction(*v, Libcalls, MustPreserveList, AsmUsed, Mangler);
for (Module::alias_iterator a = mergedModule->alias_begin(),
e = mergedModule->alias_end(); a != e; ++a)
applyRestriction(*a, Libcalls, MustPreserveList, AsmUsed, Mangler);
GlobalVariable *LLVMCompilerUsed =
mergedModule->getGlobalVariable("llvm.compiler.used");
findUsedValues(LLVMCompilerUsed, AsmUsed);
if (LLVMCompilerUsed)
LLVMCompilerUsed->eraseFromParent();
if (!AsmUsed.empty()) {
llvm::Type *i8PTy = llvm::Type::getInt8PtrTy(Context);
std::vector<Constant*> asmUsed2;
for (SmallPtrSet<GlobalValue*, 16>::const_iterator i = AsmUsed.begin(),
e = AsmUsed.end(); i !=e; ++i) {
GlobalValue *GV = *i;
Constant *c = ConstantExpr::getBitCast(GV, i8PTy);
asmUsed2.push_back(c);
}
llvm::ArrayType *ATy = llvm::ArrayType::get(i8PTy, asmUsed2.size());
LLVMCompilerUsed =
new llvm::GlobalVariable(*mergedModule, ATy, false,
llvm::GlobalValue::AppendingLinkage,
llvm::ConstantArray::get(ATy, asmUsed2),
"llvm.compiler.used");
LLVMCompilerUsed->setSection("llvm.metadata");
}
passes.add(createInternalizePass(MustPreserveList));
// apply scope restrictions
passes.run(*mergedModule);
ScopeRestrictionsDone = true;
}
/// forwardResume - Forward the 'resume' instruction to the caller's landing pad
/// block. When the landing pad block has only one predecessor, this is a simple
/// branch. When there is more than one predecessor, we need to split the
/// landing pad block after the landingpad instruction and jump to there.
void InvokeInliningInfo::forwardResume(ResumeInst *RI,
SmallPtrSet<LandingPadInst*, 16> &InlinedLPads) {
BasicBlock *Dest = getInnerResumeDest();
LandingPadInst *OuterLPad = getLandingPadInst();
BasicBlock *Src = RI->getParent();
BranchInst::Create(Dest, Src);
// Update the PHIs in the destination. They were inserted in an order which
// makes this work.
addIncomingPHIValuesForInto(Src, Dest);
InnerEHValuesPHI->addIncoming(RI->getOperand(0), Src);
RI->eraseFromParent();
// Append the clauses from the outer landing pad instruction into the inlined
// landing pad instructions.
for (SmallPtrSet<LandingPadInst*, 16>::iterator I = InlinedLPads.begin(),
E = InlinedLPads.end(); I != E; ++I) {
LandingPadInst *InlinedLPad = *I;
for (unsigned OuterIdx = 0, OuterNum = OuterLPad->getNumClauses();
OuterIdx != OuterNum; ++OuterIdx)
InlinedLPad->addClause(OuterLPad->getClause(OuterIdx));
}
}
/// ProcessLoop - Walk the loop structure in depth first order, ensuring that
/// all loops have preheaders.
///
bool LoopSimplify::ProcessLoop(Loop *L, LPPassManager &LPM) {
bool Changed = false;
ReprocessLoop:
// Check to see that no blocks (other than the header) in this loop have
// predecessors that are not in the loop. This is not valid for natural
// loops, but can occur if the blocks are unreachable. Since they are
// unreachable we can just shamelessly delete those CFG edges!
for (Loop::block_iterator BB = L->block_begin(), E = L->block_end();
BB != E; ++BB) {
if (*BB == L->getHeader()) continue;
SmallPtrSet<BasicBlock*, 4> BadPreds;
for (pred_iterator PI = pred_begin(*BB),
PE = pred_end(*BB); PI != PE; ++PI) {
BasicBlock *P = *PI;
if (!L->contains(P))
BadPreds.insert(P);
}
// Delete each unique out-of-loop (and thus dead) predecessor.
for (SmallPtrSet<BasicBlock*, 4>::iterator I = BadPreds.begin(),
E = BadPreds.end(); I != E; ++I) {
DEBUG(dbgs() << "LoopSimplify: Deleting edge from dead predecessor ";
WriteAsOperand(dbgs(), *I, false);
dbgs() << "\n");
// Inform each successor of each dead pred.
for (succ_iterator SI = succ_begin(*I), SE = succ_end(*I); SI != SE; ++SI)
(*SI)->removePredecessor(*I);
// Zap the dead pred's terminator and replace it with unreachable.
TerminatorInst *TI = (*I)->getTerminator();
TI->replaceAllUsesWith(UndefValue::get(TI->getType()));
(*I)->getTerminator()->eraseFromParent();
new UnreachableInst((*I)->getContext(), *I);
Changed = true;
}
}
// If there are exiting blocks with branches on undef, resolve the undef in
// the direction which will exit the loop. This will help simplify loop
// trip count computations.
SmallVector<BasicBlock*, 8> ExitingBlocks;
L->getExitingBlocks(ExitingBlocks);
for (SmallVectorImpl<BasicBlock *>::iterator I = ExitingBlocks.begin(),
E = ExitingBlocks.end(); I != E; ++I)
if (BranchInst *BI = dyn_cast<BranchInst>((*I)->getTerminator()))
if (BI->isConditional()) {
if (UndefValue *Cond = dyn_cast<UndefValue>(BI->getCondition())) {
DEBUG(dbgs() << "LoopSimplify: Resolving \"br i1 undef\" to exit in ";
WriteAsOperand(dbgs(), *I, false);
dbgs() << "\n");
BI->setCondition(ConstantInt::get(Cond->getType(),
!L->contains(BI->getSuccessor(0))));
Changed = true;
}
}
/// RemoveAccessedObjects - Check to see if the specified location may alias any
/// of the stack objects in the DeadStackObjects set. If so, they become live
/// because the location is being loaded.
void DSE::RemoveAccessedObjects(const AliasAnalysis::Location &LoadedLoc,
SmallPtrSet<Value*, 16> &DeadStackObjects) {
const Value *UnderlyingPointer = LoadedLoc.Ptr->getUnderlyingObject();
// A constant can't be in the dead pointer set.
if (isa<Constant>(UnderlyingPointer))
return;
// If the kill pointer can be easily reduced to an alloca, don't bother doing
// extraneous AA queries.
if (isa<AllocaInst>(UnderlyingPointer) || isa<Argument>(UnderlyingPointer)) {
DeadStackObjects.erase(const_cast<Value*>(UnderlyingPointer));
return;
}
SmallVector<Value*, 16> NowLive;
for (SmallPtrSet<Value*, 16>::iterator I = DeadStackObjects.begin(),
E = DeadStackObjects.end(); I != E; ++I) {
// See if the loaded location could alias the stack location.
AliasAnalysis::Location StackLoc(*I, getPointerSize(*I, *AA));
if (!AA->isNoAlias(StackLoc, LoadedLoc))
NowLive.push_back(*I);
}
for (SmallVector<Value*, 16>::iterator I = NowLive.begin(), E = NowLive.end();
I != E; ++I)
DeadStackObjects.erase(*I);
}
void LTOCodeGenerator::applyScopeRestrictions() {
if (_scopeRestrictionsDone) return;
Module *mergedModule = _linker.getModule();
// Start off with a verification pass.
PassManager passes;
passes.add(createVerifierPass());
// mark which symbols can not be internalized
MCContext Context(*_target->getMCAsmInfo(), *_target->getRegisterInfo(),NULL);
Mangler mangler(Context, *_target->getTargetData());
std::vector<const char*> mustPreserveList;
SmallPtrSet<GlobalValue*, 8> asmUsed;
for (Module::iterator f = mergedModule->begin(),
e = mergedModule->end(); f != e; ++f)
applyRestriction(*f, mustPreserveList, asmUsed, mangler);
for (Module::global_iterator v = mergedModule->global_begin(),
e = mergedModule->global_end(); v != e; ++v)
applyRestriction(*v, mustPreserveList, asmUsed, mangler);
for (Module::alias_iterator a = mergedModule->alias_begin(),
e = mergedModule->alias_end(); a != e; ++a)
applyRestriction(*a, mustPreserveList, asmUsed, mangler);
GlobalVariable *LLVMCompilerUsed =
mergedModule->getGlobalVariable("llvm.compiler.used");
findUsedValues(LLVMCompilerUsed, asmUsed);
if (LLVMCompilerUsed)
LLVMCompilerUsed->eraseFromParent();
llvm::Type *i8PTy = llvm::Type::getInt8PtrTy(_context);
std::vector<Constant*> asmUsed2;
for (SmallPtrSet<GlobalValue*, 16>::const_iterator i = asmUsed.begin(),
e = asmUsed.end(); i !=e; ++i) {
GlobalValue *GV = *i;
Constant *c = ConstantExpr::getBitCast(GV, i8PTy);
asmUsed2.push_back(c);
}
llvm::ArrayType *ATy = llvm::ArrayType::get(i8PTy, asmUsed2.size());
LLVMCompilerUsed =
new llvm::GlobalVariable(*mergedModule, ATy, false,
llvm::GlobalValue::AppendingLinkage,
llvm::ConstantArray::get(ATy, asmUsed2),
"llvm.compiler.used");
LLVMCompilerUsed->setSection("llvm.metadata");
// Add prerequisite passes needed by SAFECode
PassManagerBuilder().populateLTOPassManager(passes, /*Internalize=*/ false,
!DisableInline);
passes.add(createInternalizePass(mustPreserveList));
// apply scope restrictions
passes.run(*mergedModule);
_scopeRestrictionsDone = true;
}
bool CodeGenPrepare::runOnFunction(Function &F) {
bool EverMadeChange = false;
ModifiedDT = false;
TLInfo = &getAnalysis<TargetLibraryInfo>();
DT = getAnalysisIfAvailable<DominatorTree>();
PFI = getAnalysisIfAvailable<ProfileInfo>();
OptSize = F.hasFnAttr(Attribute::OptimizeForSize);
// First pass, eliminate blocks that contain only PHI nodes and an
// unconditional branch.
EverMadeChange |= EliminateMostlyEmptyBlocks(F);
// llvm.dbg.value is far away from the value then iSel may not be able
// handle it properly. iSel will drop llvm.dbg.value if it can not
// find a node corresponding to the value.
EverMadeChange |= PlaceDbgValues(F);
bool MadeChange = true;
while (MadeChange) {
MadeChange = false;
for (Function::iterator I = F.begin(), E = F.end(); I != E; ) {
BasicBlock *BB = I++;
MadeChange |= OptimizeBlock(*BB);
}
EverMadeChange |= MadeChange;
}
SunkAddrs.clear();
if (!DisableBranchOpts) {
MadeChange = false;
SmallPtrSet<BasicBlock*, 8> WorkList;
for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB) {
SmallVector<BasicBlock*, 2> Successors(succ_begin(BB), succ_end(BB));
MadeChange |= ConstantFoldTerminator(BB, true);
if (!MadeChange) continue;
for (SmallVectorImpl<BasicBlock*>::iterator
II = Successors.begin(), IE = Successors.end(); II != IE; ++II)
if (pred_begin(*II) == pred_end(*II))
WorkList.insert(*II);
}
if (!DisableDeleteDeadBlocks)
for (SmallPtrSet<BasicBlock*, 8>::iterator
I = WorkList.begin(), E = WorkList.end(); I != E; ++I)
DeleteDeadBlock(*I);
if (MadeChange)
ModifiedDT = true;
EverMadeChange |= MadeChange;
}
if (ModifiedDT && DT)
DT->DT->recalculate(F);
return EverMadeChange;
}
/// \brief Calculate edge weights for successors lead to unreachable.
///
/// Predict that a successor which leads necessarily to an
/// unreachable-terminated block as extremely unlikely.
bool BranchProbabilityInfo::calcUnreachableHeuristics(BasicBlock *BB) {
TerminatorInst *TI = BB->getTerminator();
if (TI->getNumSuccessors() == 0) {
if (isa<UnreachableInst>(TI))
PostDominatedByUnreachable.insert(BB);
return false;
}
SmallPtrSet<BasicBlock *, 4> UnreachableEdges;
SmallPtrSet<BasicBlock *, 4> ReachableEdges;
for (succ_iterator I = succ_begin(BB), E = succ_end(BB); I != E; ++I) {
if (PostDominatedByUnreachable.count(*I))
UnreachableEdges.insert(*I);
else
ReachableEdges.insert(*I);
}
// If all successors are in the set of blocks post-dominated by unreachable,
// this block is too.
if (UnreachableEdges.size() == TI->getNumSuccessors())
PostDominatedByUnreachable.insert(BB);
// Skip probabilities if this block has a single successor or if all were
// reachable.
if (TI->getNumSuccessors() == 1 || UnreachableEdges.empty())
return false;
uint32_t UnreachableWeight =
std::max(UR_TAKEN_WEIGHT / UnreachableEdges.size(), MIN_WEIGHT);
for (SmallPtrSet<BasicBlock *, 4>::iterator I = UnreachableEdges.begin(),
E = UnreachableEdges.end();
I != E; ++I)
setEdgeWeight(BB, *I, UnreachableWeight);
if (ReachableEdges.empty())
return true;
uint32_t ReachableWeight =
std::max(UR_NONTAKEN_WEIGHT / ReachableEdges.size(), NORMAL_WEIGHT);
for (SmallPtrSet<BasicBlock *, 4>::iterator I = ReachableEdges.begin(),
E = ReachableEdges.end();
I != E; ++I)
setEdgeWeight(BB, *I, ReachableWeight);
return true;
}
/// removeDeadFunctions - Remove dead functions that are not included in
/// DNR (Do Not Remove) list.
bool Inliner::removeDeadFunctions(CallGraph &CG,
SmallPtrSet<const Function *, 16> *DNR) {
SmallPtrSet<CallGraphNode*, 16> FunctionsToRemove;
// Scan for all of the functions, looking for ones that should now be removed
// from the program. Insert the dead ones in the FunctionsToRemove set.
for (CallGraph::iterator I = CG.begin(), E = CG.end(); I != E; ++I) {
CallGraphNode *CGN = I->second;
if (CGN->getFunction() == 0)
continue;
Function *F = CGN->getFunction();
// If the only remaining users of the function are dead constants, remove
// them.
F->removeDeadConstantUsers();
if (DNR && DNR->count(F))
continue;
if (!F->hasLinkOnceLinkage() && !F->hasLocalLinkage() &&
!F->hasAvailableExternallyLinkage())
continue;
if (!F->use_empty())
continue;
// Remove any call graph edges from the function to its callees.
CGN->removeAllCalledFunctions();
// Remove any edges from the external node to the function's call graph
// node. These edges might have been made irrelegant due to
// optimization of the program.
CG.getExternalCallingNode()->removeAnyCallEdgeTo(CGN);
// Removing the node for callee from the call graph and delete it.
FunctionsToRemove.insert(CGN);
}
// Now that we know which functions to delete, do so. We didn't want to do
// this inline, because that would invalidate our CallGraph::iterator
// objects. :(
//
// Note that it doesn't matter that we are iterating over a non-stable set
// here to do this, it doesn't matter which order the functions are deleted
// in.
bool Changed = false;
for (SmallPtrSet<CallGraphNode*, 16>::iterator I = FunctionsToRemove.begin(),
E = FunctionsToRemove.end(); I != E; ++I) {
resetCachedCostInfo((*I)->getFunction());
delete CG.removeFunctionFromModule(*I);
++NumDeleted;
Changed = true;
}
return Changed;
}
/// Test if there is any edge from V in the upper direction
bool ABCD::InequalityGraph::hasEdge(Value *V, bool upper) const {
SmallPtrSet<Edge *, 16> it = graph.lookup(V);
SmallPtrSet<Edge *, 16>::iterator begin = it.begin();
SmallPtrSet<Edge *, 16>::iterator end = it.end();
for (; begin != end; ++begin) {
if ((*begin)->isUpperBound() == upper) {
return true;
}
}
return false;
}
void LowerEmAsyncify::FindContextVariables(AsyncCallEntry & Entry) {
BasicBlock *AfterCallBlock = Entry.AfterCallBlock;
Function & F = *AfterCallBlock->getParent();
// Create a new entry block as if in the callback function
// theck check variables that no longer properly dominate their uses
BasicBlock *EntryBlock = BasicBlock::Create(TheModule->getContext(), "", &F, &F.getEntryBlock());
BranchInst::Create(AfterCallBlock, EntryBlock);
DominatorTreeWrapperPass DTW;
DTW.runOnFunction(F);
DominatorTree& DT = DTW.getDomTree();
// These blocks may be using some values defined at or before AsyncCallBlock
BasicBlockSet Ramifications = FindReachableBlocksFrom(AfterCallBlock);
SmallPtrSet<Value*, 256> ContextVariables;
Values Pending;
// Examine the instructions, find all variables that we need to store in the context
for (BasicBlockSet::iterator RI = Ramifications.begin(), RE = Ramifications.end(); RI != RE; ++RI) {
for (BasicBlock::iterator I = (*RI)->begin(), E = (*RI)->end(); I != E; ++I) {
for (unsigned i = 0, NumOperands = I->getNumOperands(); i < NumOperands; ++i) {
Value *O = I->getOperand(i);
if (Instruction *Inst = dyn_cast<Instruction>(O)) {
if (Inst == Entry.AsyncCallInst) continue; // for the original async call, we will load directly from async return value
if (ContextVariables.count(Inst) != 0) continue; // already examined
if (!DT.dominates(Inst, I->getOperandUse(i))) {
// `I` is using `Inst`, yet `Inst` does not dominate `I` if we arrive directly at AfterCallBlock
// so we need to save `Inst` in the context
ContextVariables.insert(Inst);
Pending.push_back(Inst);
}
} else if (Argument *Arg = dyn_cast<Argument>(O)) {
// count() should be as fast/slow as insert, so just insert here
ContextVariables.insert(Arg);
}
}
}
}
// restore F
EntryBlock->eraseFromParent();
Entry.ContextVariables.clear();
Entry.ContextVariables.reserve(ContextVariables.size());
for (SmallPtrSet<Value*, 256>::iterator I = ContextVariables.begin(), E = ContextVariables.end(); I != E; ++I) {
Entry.ContextVariables.push_back(*I);
}
}
bool ReduceCrashingInstructions::TestInsts(std::vector<const Instruction*>
&Insts) {
// Clone the program to try hacking it apart...
ValueToValueMapTy VMap;
Module *M = CloneModule(BD.getProgram(), VMap);
// Convert list to set for fast lookup...
SmallPtrSet<Instruction*, 64> Instructions;
for (unsigned i = 0, e = Insts.size(); i != e; ++i) {
assert(!isa<TerminatorInst>(Insts[i]));
Instructions.insert(cast<Instruction>(VMap[Insts[i]]));
}
outs() << "Checking for crash with only " << Instructions.size();
if (Instructions.size() == 1)
outs() << " instruction: ";
else
outs() << " instructions: ";
for (Module::iterator MI = M->begin(), ME = M->end(); MI != ME; ++MI)
for (Function::iterator FI = MI->begin(), FE = MI->end(); FI != FE; ++FI)
for (BasicBlock::iterator I = FI->begin(), E = FI->end(); I != E;) {
Instruction *Inst = I++;
if (!Instructions.count(Inst) && !isa<TerminatorInst>(Inst) &&
!isa<LandingPadInst>(Inst)) {
if (!Inst->getType()->isVoidTy())
Inst->replaceAllUsesWith(UndefValue::get(Inst->getType()));
Inst->eraseFromParent();
}
}
// Verify that this is still valid.
PassManager Passes;
Passes.add(createVerifierPass());
Passes.add(createDebugInfoVerifierPass());
Passes.run(*M);
// Try running on the hacked up program...
if (TestFn(BD, M)) {
BD.setNewProgram(M); // It crashed, keep the trimmed version...
// Make sure to use instruction pointers that point into the now-current
// module, and that they don't include any deleted blocks.
Insts.clear();
for (SmallPtrSet<Instruction*, 64>::const_iterator I = Instructions.begin(),
E = Instructions.end(); I != E; ++I)
Insts.push_back(*I);
return true;
}
delete M; // It didn't crash, try something else.
return false;
}
// Calculate Edge Weights using "Return Heuristics". Predict a successor which
// leads directly to Return Instruction will not be taken.
bool BranchProbabilityAnalysis::calcReturnHeuristics(BasicBlock *BB){
if (BB->getTerminator()->getNumSuccessors() == 1)
return false;
SmallPtrSet<BasicBlock *, 4> ReturningEdges;
SmallPtrSet<BasicBlock *, 4> StayEdges;
for (succ_iterator I = succ_begin(BB), E = succ_end(BB); I != E; ++I) {
BasicBlock *Succ = *I;
if (isReturningBlock(Succ))
ReturningEdges.insert(Succ);
else
StayEdges.insert(Succ);
}
if (uint32_t numStayEdges = StayEdges.size()) {
uint32_t stayWeight = RH_TAKEN_WEIGHT / numStayEdges;
if (stayWeight < NORMAL_WEIGHT)
stayWeight = NORMAL_WEIGHT;
for (SmallPtrSet<BasicBlock *, 4>::iterator I = StayEdges.begin(),
E = StayEdges.end(); I != E; ++I)
BP->setEdgeWeight(BB, *I, stayWeight);
}
if (uint32_t numRetEdges = ReturningEdges.size()) {
uint32_t retWeight = RH_NONTAKEN_WEIGHT / numRetEdges;
if (retWeight < MIN_WEIGHT)
retWeight = MIN_WEIGHT;
for (SmallPtrSet<BasicBlock *, 4>::iterator I = ReturningEdges.begin(),
E = ReturningEdges.end(); I != E; ++I) {
BP->setEdgeWeight(BB, *I, retWeight);
}
}
return ReturningEdges.size() > 0;
}
/// HandleInlinedInvoke - If we inlined an invoke site, we need to convert calls
/// in the body of the inlined function into invokes.
///
/// II is the invoke instruction being inlined. FirstNewBlock is the first
/// block of the inlined code (the last block is the end of the function),
/// and InlineCodeInfo is information about the code that got inlined.
static void HandleInlinedInvoke(InvokeInst *II, BasicBlock *FirstNewBlock,
ClonedCodeInfo &InlinedCodeInfo) {
BasicBlock *InvokeDest = II->getUnwindDest();
Function *Caller = FirstNewBlock->getParent();
// The inlined code is currently at the end of the function, scan from the
// start of the inlined code to its end, checking for stuff we need to
// rewrite.
InvokeInliningInfo Invoke(II);
// Get all of the inlined landing pad instructions.
SmallPtrSet<LandingPadInst*, 16> InlinedLPads;
for (Function::iterator I = FirstNewBlock, E = Caller->end(); I != E; ++I)
if (InvokeInst *II = dyn_cast<InvokeInst>(I->getTerminator()))
InlinedLPads.insert(II->getLandingPadInst());
// Append the clauses from the outer landing pad instruction into the inlined
// landing pad instructions.
LandingPadInst *OuterLPad = Invoke.getLandingPadInst();
for (SmallPtrSet<LandingPadInst*, 16>::iterator I = InlinedLPads.begin(),
E = InlinedLPads.end(); I != E; ++I) {
LandingPadInst *InlinedLPad = *I;
unsigned OuterNum = OuterLPad->getNumClauses();
InlinedLPad->reserveClauses(OuterNum);
for (unsigned OuterIdx = 0; OuterIdx != OuterNum; ++OuterIdx)
InlinedLPad->addClause(OuterLPad->getClause(OuterIdx));
if (OuterLPad->isCleanup())
InlinedLPad->setCleanup(true);
}
for (Function::iterator BB = FirstNewBlock, E = Caller->end(); BB != E; ++BB) {
if (InlinedCodeInfo.ContainsCalls)
HandleCallsInBlockInlinedThroughInvoke(BB, Invoke);
// Forward any resumes that are remaining here.
if (ResumeInst *RI = dyn_cast<ResumeInst>(BB->getTerminator()))
Invoke.forwardResume(RI, InlinedLPads);
}
// Now that everything is happy, we have one final detail. The PHI nodes in
// the exception destination block still have entries due to the original
// invoke instruction. Eliminate these entries (which might even delete the
// PHI node) now.
InvokeDest->removePredecessor(II->getParent());
}
/// Insert phi functions when necessary
///
void SSI::insertPhiFunctions(SmallPtrSet<Instruction*, 4> &value) {
DominanceFrontier *DF = &getAnalysis<DominanceFrontier>();
for (SmallPtrSet<Instruction*, 4>::iterator I = value.begin(),
E = value.end(); I != E; ++I) {
// Test if there were any sigmas for this variable
SmallPtrSet<BasicBlock *, 16> BB_visited;
// Insert phi functions if there is any sigma function
while (!defsites[*I].empty()) {
BasicBlock *BB = defsites[*I].back();
defsites[*I].pop_back();
DominanceFrontier::iterator DF_BB = DF->find(BB);
// The BB is unreachable. Skip it.
if (DF_BB == DF->end())
continue;
// Iterates through all the dominance frontier of BB
for (std::set<BasicBlock *>::iterator DF_BB_begin =
DF_BB->second.begin(), DF_BB_end = DF_BB->second.end();
DF_BB_begin != DF_BB_end; ++DF_BB_begin) {
BasicBlock *BB_dominated = *DF_BB_begin;
// Test if has not yet visited this node and if the
// original definition dominates this node
if (BB_visited.insert(BB_dominated) &&
DT_->properlyDominates(value_original[*I], BB_dominated) &&
dominateAny(BB_dominated, *I)) {
PHINode *PN = PHINode::Create(
(*I)->getType(), SSI_PHI, BB_dominated->begin());
phis.insert(std::make_pair(PN, *I));
created.insert(PN);
defsites[*I].push_back(BB_dominated);
++NumPhiInserted;
}
}
}
BB_visited.clear();
}
}
static void RemoveDeadConstant(Constant *C) {
assert(C->use_empty() && "Constant is not dead!");
SmallPtrSet<Constant*, 4> Operands;
for (unsigned i = 0, e = C->getNumOperands(); i != e; ++i)
if (OnlyUsedBy(C->getOperand(i), C))
Operands.insert(cast<Constant>(C->getOperand(i)));
if (GlobalVariable *GV = dyn_cast<GlobalVariable>(C)) {
if (!GV->hasLocalLinkage()) return; // Don't delete non static globals.
GV->eraseFromParent();
}
else if (!isa<Function>(C))
if (isa<CompositeType>(C->getType()))
C->destroyConstant();
// If the constant referenced anything, see if we can delete it as well.
for (SmallPtrSet<Constant*, 4>::iterator OI = Operands.begin(),
OE = Operands.end(); OI != OE; ++OI)
RemoveDeadConstant(*OI);
}
/// Insert sigma functions (a sigma function is a phi function with one
/// operator)
///
void SSI::insertSigmaFunctions(SmallPtrSet<Instruction*, 4> &value) {
for (SmallPtrSet<Instruction*, 4>::iterator I = value.begin(),
E = value.end(); I != E; ++I) {
for (Value::use_iterator begin = (*I)->use_begin(),
end = (*I)->use_end(); begin != end; ++begin) {
// Test if the Use of the Value is in a comparator
if (CmpInst *CI = dyn_cast<CmpInst>(begin)) {
// Iterates through all uses of CmpInst
for (Value::use_iterator begin_ci = CI->use_begin(),
end_ci = CI->use_end(); begin_ci != end_ci; ++begin_ci) {
// Test if any use of CmpInst is in a Terminator
if (TerminatorInst *TI = dyn_cast<TerminatorInst>(begin_ci)) {
insertSigma(TI, *I);
}
}
}
}
}
}
static bool isSafeToMove(Instruction *Inst, AliasAnalysis *AA,
SmallPtrSet<Instruction *, 8> &Stores) {
if (LoadInst *L = dyn_cast<LoadInst>(Inst)) {
if (L->isVolatile()) return false;
Value *Ptr = L->getPointerOperand();
unsigned Size = AA->getTypeStoreSize(L->getType());
for (SmallPtrSet<Instruction *, 8>::iterator I = Stores.begin(),
E = Stores.end(); I != E; ++I)
if (AA->getModRefInfo(*I, Ptr, Size) & AliasAnalysis::Mod)
return false;
}
if (Inst->mayWriteToMemory()) {
Stores.insert(Inst);
return false;
}
return Inst->isSafeToSpeculativelyExecute();
}
static bool isSafeToMove(Instruction *Inst, AliasAnalysis *AA,
SmallPtrSet<Instruction *, 8> &Stores) {
if (Inst->mayWriteToMemory()) {
Stores.insert(Inst);
return false;
}
if (LoadInst *L = dyn_cast<LoadInst>(Inst)) {
AliasAnalysis::Location Loc = AA->getLocation(L);
for (SmallPtrSet<Instruction *, 8>::iterator I = Stores.begin(),
E = Stores.end(); I != E; ++I)
if (AA->getModRefInfo(*I, Loc) & AliasAnalysis::Mod)
return false;
}
if (isa<TerminatorInst>(Inst) || isa<PHINode>(Inst))
return false;
return true;
}
/// Updates the distance value for a and b
void ABCD::updateMemDistance(Value *a, Value *b, Bound *bound, unsigned level,
meet_function meet) {
ABCD::ProveResult res = (meet == max) ? False : True;
SmallPtrSet<Edge *, 16> Edges = inequality_graph.getEdges(b);
SmallPtrSet<Edge *, 16>::iterator begin = Edges.begin(), end = Edges.end();
for (; begin != end ; ++begin) {
if (((res >= Reduced) && (meet == max)) ||
((res == False) && (meet == min))) {
break;
}
Edge *in = *begin;
if (in->isUpperBound() == bound->isUpperBound()) {
Value *succ = in->getVertex();
res = meet(res, prove(a, succ, new Bound(bound, in->getValue()),
level+1));
}
}
mem_result.updateBound(b, bound, res);
}
/// CleanupSelectors - Any remaining eh.selector intrinsic calls which still use
/// the "llvm.eh.catch.all.value" call need to convert to using its
/// initializer instead.
bool DwarfEHPrepare::CleanupSelectors(SmallPtrSet<IntrinsicInst*, 32> &Sels) {
if (!EHCatchAllValue) return false;
if (!SelectorIntrinsic) {
SelectorIntrinsic =
Intrinsic::getDeclaration(F->getParent(), Intrinsic::eh_selector);
if (!SelectorIntrinsic) return false;
}
bool Changed = false;
for (SmallPtrSet<IntrinsicInst*, 32>::iterator
I = Sels.begin(), E = Sels.end(); I != E; ++I) {
IntrinsicInst *Sel = *I;
// Index of the "llvm.eh.catch.all.value" variable.
unsigned OpIdx = Sel->getNumArgOperands() - 1;
GlobalVariable *GV = dyn_cast<GlobalVariable>(Sel->getArgOperand(OpIdx));
if (GV != EHCatchAllValue) continue;
Sel->setArgOperand(OpIdx, EHCatchAllValue->getInitializer());
Changed = true;
}
return Changed;
}
bool ClassTemplateToClass::hasUsedNameDecl(
ClassTemplatePartialSpecializationDecl *PartialD)
{
if (!PartialD->isCompleteDefinition())
return false;
SmallPtrSet<NamedDecl *, 8> Params;
TemplateParameterList *PartialTPList = PartialD->getTemplateParameters();
for (unsigned PI = 0; PI < PartialTPList->size(); ++PI) {
NamedDecl *ND = PartialTPList->getParam(PI);
if (dyn_cast<NonTypeTemplateParmDecl>(ND))
continue;
Params.insert(ND);
}
TemplateParameterTypeVisitor ParamVisitor(Context);
// Skip visiting parameters and arguments
for (CXXRecordDecl::base_class_iterator I = PartialD->bases_begin(),
E = PartialD->bases_end(); I != E; ++I) {
ParamVisitor.TraverseType(I->getType());
}
DeclContext *Ctx = dyn_cast<DeclContext>(PartialD);
for (DeclContext::decl_iterator DI = Ctx->decls_begin(),
DE = Ctx->decls_end(); DI != DE; ++DI) {
ParamVisitor.TraverseDecl(*DI);
}
for (SmallPtrSet<NamedDecl *, 8>::iterator I = Params.begin(),
E = Params.end(); I != E; ++I) {
if (ParamVisitor.isAUsedParameter(*I))
return true;
}
return false;
}
// Calculate Edge Weights using "Loop Branch Heuristics". Predict backedges
// as taken, exiting edges as not-taken.
bool BranchProbabilityAnalysis::calcLoopBranchHeuristics(BasicBlock *BB) {
uint32_t numSuccs = BB->getTerminator()->getNumSuccessors();
Loop *L = LI->getLoopFor(BB);
if (!L)
return false;
SmallPtrSet<BasicBlock *, 8> BackEdges;
SmallPtrSet<BasicBlock *, 8> ExitingEdges;
SmallPtrSet<BasicBlock *, 8> InEdges; // Edges from header to the loop.
bool isHeader = BB == L->getHeader();
for (succ_iterator I = succ_begin(BB), E = succ_end(BB); I != E; ++I) {
BasicBlock *Succ = *I;
Loop *SuccL = LI->getLoopFor(Succ);
if (SuccL != L)
ExitingEdges.insert(Succ);
else if (Succ == L->getHeader())
BackEdges.insert(Succ);
else if (isHeader)
InEdges.insert(Succ);
}
if (uint32_t numBackEdges = BackEdges.size()) {
uint32_t backWeight = LBH_TAKEN_WEIGHT / numBackEdges;
if (backWeight < NORMAL_WEIGHT)
backWeight = NORMAL_WEIGHT;
for (SmallPtrSet<BasicBlock *, 8>::iterator EI = BackEdges.begin(),
EE = BackEdges.end(); EI != EE; ++EI) {
BasicBlock *Back = *EI;
BP->setEdgeWeight(BB, Back, backWeight);
}
}
if (uint32_t numInEdges = InEdges.size()) {
uint32_t inWeight = LBH_TAKEN_WEIGHT / numInEdges;
if (inWeight < NORMAL_WEIGHT)
inWeight = NORMAL_WEIGHT;
for (SmallPtrSet<BasicBlock *, 8>::iterator EI = InEdges.begin(),
EE = InEdges.end(); EI != EE; ++EI) {
BasicBlock *Back = *EI;
BP->setEdgeWeight(BB, Back, inWeight);
}
}
uint32_t numExitingEdges = ExitingEdges.size();
if (uint32_t numNonExitingEdges = numSuccs - numExitingEdges) {
uint32_t exitWeight = LBH_NONTAKEN_WEIGHT / numNonExitingEdges;
if (exitWeight < MIN_WEIGHT)
exitWeight = MIN_WEIGHT;
for (SmallPtrSet<BasicBlock *, 8>::iterator EI = ExitingEdges.begin(),
EE = ExitingEdges.end(); EI != EE; ++EI) {
BasicBlock *Exiting = *EI;
BP->setEdgeWeight(BB, Exiting, exitWeight);
}
}
return true;
}
/// lowerAcrossUnwindEdges - Find all variables which are alive across an unwind
/// edge and spill them.
void SjLjEHPrepare::lowerAcrossUnwindEdges(Function &F,
ArrayRef<InvokeInst *> Invokes) {
// Finally, scan the code looking for instructions with bad live ranges.
for (Function::iterator BB = F.begin(), BBE = F.end(); BB != BBE; ++BB) {
for (BasicBlock::iterator II = BB->begin(), IIE = BB->end(); II != IIE;
++II) {
// Ignore obvious cases we don't have to handle. In particular, most
// instructions either have no uses or only have a single use inside the
// current block. Ignore them quickly.
Instruction *Inst = II;
if (Inst->use_empty())
continue;
if (Inst->hasOneUse() &&
cast<Instruction>(Inst->user_back())->getParent() == BB &&
!isa<PHINode>(Inst->user_back()))
continue;
// If this is an alloca in the entry block, it's not a real register
// value.
if (AllocaInst *AI = dyn_cast<AllocaInst>(Inst))
if (isa<ConstantInt>(AI->getArraySize()) && BB == F.begin())
continue;
// Avoid iterator invalidation by copying users to a temporary vector.
SmallVector<Instruction *, 16> Users;
for (User *U : Inst->users()) {
Instruction *UI = cast<Instruction>(U);
if (UI->getParent() != BB || isa<PHINode>(UI))
Users.push_back(UI);
}
// Find all of the blocks that this value is live in.
SmallPtrSet<BasicBlock *, 64> LiveBBs;
LiveBBs.insert(Inst->getParent());
while (!Users.empty()) {
Instruction *U = Users.back();
Users.pop_back();
if (!isa<PHINode>(U)) {
MarkBlocksLiveIn(U->getParent(), LiveBBs);
} else {
// Uses for a PHI node occur in their predecessor block.
PHINode *PN = cast<PHINode>(U);
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
if (PN->getIncomingValue(i) == Inst)
MarkBlocksLiveIn(PN->getIncomingBlock(i), LiveBBs);
}
}
// Now that we know all of the blocks that this thing is live in, see if
// it includes any of the unwind locations.
bool NeedsSpill = false;
for (unsigned i = 0, e = Invokes.size(); i != e; ++i) {
BasicBlock *UnwindBlock = Invokes[i]->getUnwindDest();
if (UnwindBlock != BB && LiveBBs.count(UnwindBlock)) {
DEBUG(dbgs() << "SJLJ Spill: " << *Inst << " around "
<< UnwindBlock->getName() << "\n");
NeedsSpill = true;
break;
}
}
// If we decided we need a spill, do it.
// FIXME: Spilling this way is overkill, as it forces all uses of
// the value to be reloaded from the stack slot, even those that aren't
// in the unwind blocks. We should be more selective.
if (NeedsSpill) {
DemoteRegToStack(*Inst, true);
++NumSpilled;
}
}
}
// Go through the landing pads and remove any PHIs there.
for (unsigned i = 0, e = Invokes.size(); i != e; ++i) {
BasicBlock *UnwindBlock = Invokes[i]->getUnwindDest();
LandingPadInst *LPI = UnwindBlock->getLandingPadInst();
// Place PHIs into a set to avoid invalidating the iterator.
SmallPtrSet<PHINode *, 8> PHIsToDemote;
for (BasicBlock::iterator PN = UnwindBlock->begin(); isa<PHINode>(PN); ++PN)
PHIsToDemote.insert(cast<PHINode>(PN));
if (PHIsToDemote.empty())
continue;
// Demote the PHIs to the stack.
for (SmallPtrSet<PHINode *, 8>::iterator I = PHIsToDemote.begin(),
E = PHIsToDemote.end();
I != E; ++I)
DemotePHIToStack(*I);
// Move the landingpad instruction back to the top of the landing pad block.
LPI->moveBefore(UnwindBlock->begin());
}
}
bool ReduceCrashingBlocks::TestBlocks(std::vector<const BasicBlock*> &BBs) {
// Clone the program to try hacking it apart...
ValueMap<const Value*, Value*> VMap;
Module *M = CloneModule(BD.getProgram(), VMap);
// Convert list to set for fast lookup...
SmallPtrSet<BasicBlock*, 8> Blocks;
for (unsigned i = 0, e = BBs.size(); i != e; ++i)
Blocks.insert(cast<BasicBlock>(VMap[BBs[i]]));
outs() << "Checking for crash with only these blocks:";
unsigned NumPrint = Blocks.size();
if (NumPrint > 10) NumPrint = 10;
for (unsigned i = 0, e = NumPrint; i != e; ++i)
outs() << " " << BBs[i]->getName();
if (NumPrint < Blocks.size())
outs() << "... <" << Blocks.size() << " total>";
outs() << ": ";
// Loop over and delete any hack up any blocks that are not listed...
for (Module::iterator I = M->begin(), E = M->end(); I != E; ++I)
for (Function::iterator BB = I->begin(), E = I->end(); BB != E; ++BB)
if (!Blocks.count(BB) && BB->getTerminator()->getNumSuccessors()) {
// Loop over all of the successors of this block, deleting any PHI nodes
// that might include it.
for (succ_iterator SI = succ_begin(BB), E = succ_end(BB); SI != E; ++SI)
(*SI)->removePredecessor(BB);
TerminatorInst *BBTerm = BB->getTerminator();
if (!BB->getTerminator()->getType()->isVoidTy())
BBTerm->replaceAllUsesWith(Constant::getNullValue(BBTerm->getType()));
// Replace the old terminator instruction.
BB->getInstList().pop_back();
new UnreachableInst(BB->getContext(), BB);
}
// The CFG Simplifier pass may delete one of the basic blocks we are
// interested in. If it does we need to take the block out of the list. Make
// a "persistent mapping" by turning basic blocks into <function, name> pairs.
// This won't work well if blocks are unnamed, but that is just the risk we
// have to take.
std::vector<std::pair<std::string, std::string> > BlockInfo;
for (SmallPtrSet<BasicBlock*, 8>::iterator I = Blocks.begin(),
E = Blocks.end(); I != E; ++I)
BlockInfo.push_back(std::make_pair((*I)->getParent()->getName(),
(*I)->getName()));
// Now run the CFG simplify pass on the function...
std::vector<std::string> Passes;
Passes.push_back("simplifycfg");
Passes.push_back("verify");
Module *New = BD.runPassesOn(M, Passes);
delete M;
if (!New) {
errs() << "simplifycfg failed!\n";
exit(1);
}
M = New;
// Try running on the hacked up program...
if (TestFn(BD, M)) {
BD.setNewProgram(M); // It crashed, keep the trimmed version...
// Make sure to use basic block pointers that point into the now-current
// module, and that they don't include any deleted blocks.
BBs.clear();
const ValueSymbolTable &GST = M->getValueSymbolTable();
for (unsigned i = 0, e = BlockInfo.size(); i != e; ++i) {
Function *F = cast<Function>(GST.lookup(BlockInfo[i].first));
ValueSymbolTable &ST = F->getValueSymbolTable();
Value* V = ST.lookup(BlockInfo[i].second);
if (V && V->getType() == Type::getLabelTy(V->getContext()))
BBs.push_back(cast<BasicBlock>(V));
}
return true;
}
delete M; // It didn't crash, try something else.
return false;
}
/// ProcessLoop - Walk the loop structure in depth first order, ensuring that
/// all loops have preheaders.
///
bool LoopSimplify::ProcessLoop(Loop *L, LPPassManager &LPM) {
bool Changed = false;
ReprocessLoop:
// Check to see that no blocks (other than the header) in this loop have
// predecessors that are not in the loop. This is not valid for natural
// loops, but can occur if the blocks are unreachable. Since they are
// unreachable we can just shamelessly delete those CFG edges!
for (Loop::block_iterator BB = L->block_begin(), E = L->block_end();
BB != E; ++BB) {
if (*BB == L->getHeader()) continue;
SmallPtrSet<BasicBlock*, 4> BadPreds;
for (pred_iterator PI = pred_begin(*BB),
PE = pred_end(*BB); PI != PE; ++PI) {
BasicBlock *P = *PI;
if (!L->contains(P))
BadPreds.insert(P);
}
// Delete each unique out-of-loop (and thus dead) predecessor.
for (SmallPtrSet<BasicBlock*, 4>::iterator I = BadPreds.begin(),
E = BadPreds.end(); I != E; ++I) {
DEBUG(dbgs() << "LoopSimplify: Deleting edge from dead predecessor "
<< (*I)->getName() << "\n");
// Inform each successor of each dead pred.
for (succ_iterator SI = succ_begin(*I), SE = succ_end(*I); SI != SE; ++SI)
(*SI)->removePredecessor(*I);
// Zap the dead pred's terminator and replace it with unreachable.
TerminatorInst *TI = (*I)->getTerminator();
TI->replaceAllUsesWith(UndefValue::get(TI->getType()));
(*I)->getTerminator()->eraseFromParent();
new UnreachableInst((*I)->getContext(), *I);
Changed = true;
}
}
// If there are exiting blocks with branches on undef, resolve the undef in
// the direction which will exit the loop. This will help simplify loop
// trip count computations.
SmallVector<BasicBlock*, 8> ExitingBlocks;
L->getExitingBlocks(ExitingBlocks);
for (SmallVectorImpl<BasicBlock *>::iterator I = ExitingBlocks.begin(),
E = ExitingBlocks.end(); I != E; ++I)
if (BranchInst *BI = dyn_cast<BranchInst>((*I)->getTerminator()))
if (BI->isConditional()) {
if (UndefValue *Cond = dyn_cast<UndefValue>(BI->getCondition())) {
DEBUG(dbgs() << "LoopSimplify: Resolving \"br i1 undef\" to exit in "
<< (*I)->getName() << "\n");
BI->setCondition(ConstantInt::get(Cond->getType(),
!L->contains(BI->getSuccessor(0))));
Changed = true;
}
}
// Does the loop already have a preheader? If so, don't insert one.
BasicBlock *Preheader = L->getLoopPreheader();
if (!Preheader) {
Preheader = InsertPreheaderForLoop(L);
if (Preheader) {
++NumInserted;
Changed = true;
}
}
// Next, check to make sure that all exit nodes of the loop only have
// predecessors that are inside of the loop. This check guarantees that the
// loop preheader/header will dominate the exit blocks. If the exit block has
// predecessors from outside of the loop, split the edge now.
SmallVector<BasicBlock*, 8> ExitBlocks;
L->getExitBlocks(ExitBlocks);
SmallSetVector<BasicBlock *, 8> ExitBlockSet(ExitBlocks.begin(),
ExitBlocks.end());
for (SmallSetVector<BasicBlock *, 8>::iterator I = ExitBlockSet.begin(),
E = ExitBlockSet.end(); I != E; ++I) {
BasicBlock *ExitBlock = *I;
for (pred_iterator PI = pred_begin(ExitBlock), PE = pred_end(ExitBlock);
PI != PE; ++PI)
// Must be exactly this loop: no subloops, parent loops, or non-loop preds
// allowed.
if (!L->contains(*PI)) {
if (RewriteLoopExitBlock(L, ExitBlock)) {
++NumInserted;
Changed = true;
}
break;
}
}
// If the header has more than two predecessors at this point (from the
// preheader and from multiple backedges), we must adjust the loop.
BasicBlock *LoopLatch = L->getLoopLatch();
if (!LoopLatch) {
// If this is really a nested loop, rip it out into a child loop. Don't do
// this for loops with a giant number of backedges, just factor them into a
// common backedge instead.
if (L->getNumBackEdges() < 8) {
if (SeparateNestedLoop(L, LPM)) {
++NumNested;
// This is a big restructuring change, reprocess the whole loop.
Changed = true;
// GCC doesn't tail recursion eliminate this.
goto ReprocessLoop;
}
}
// If we either couldn't, or didn't want to, identify nesting of the loops,
// insert a new block that all backedges target, then make it jump to the
// loop header.
LoopLatch = InsertUniqueBackedgeBlock(L, Preheader);
if (LoopLatch) {
++NumInserted;
Changed = true;
}
}
// Scan over the PHI nodes in the loop header. Since they now have only two
// incoming values (the loop is canonicalized), we may have simplified the PHI
// down to 'X = phi [X, Y]', which should be replaced with 'Y'.
PHINode *PN;
for (BasicBlock::iterator I = L->getHeader()->begin();
(PN = dyn_cast<PHINode>(I++)); )
if (Value *V = SimplifyInstruction(PN, 0, DT)) {
if (AA) AA->deleteValue(PN);
if (SE) SE->forgetValue(PN);
PN->replaceAllUsesWith(V);
PN->eraseFromParent();
}
// If this loop has multiple exits and the exits all go to the same
// block, attempt to merge the exits. This helps several passes, such
// as LoopRotation, which do not support loops with multiple exits.
// SimplifyCFG also does this (and this code uses the same utility
// function), however this code is loop-aware, where SimplifyCFG is
// not. That gives it the advantage of being able to hoist
// loop-invariant instructions out of the way to open up more
// opportunities, and the disadvantage of having the responsibility
// to preserve dominator information.
bool UniqueExit = true;
if (!ExitBlocks.empty())
for (unsigned i = 1, e = ExitBlocks.size(); i != e; ++i)
if (ExitBlocks[i] != ExitBlocks[0]) {
UniqueExit = false;
break;
}
if (UniqueExit) {
for (unsigned i = 0, e = ExitingBlocks.size(); i != e; ++i) {
BasicBlock *ExitingBlock = ExitingBlocks[i];
if (!ExitingBlock->getSinglePredecessor()) continue;
BranchInst *BI = dyn_cast<BranchInst>(ExitingBlock->getTerminator());
if (!BI || !BI->isConditional()) continue;
CmpInst *CI = dyn_cast<CmpInst>(BI->getCondition());
if (!CI || CI->getParent() != ExitingBlock) continue;
// Attempt to hoist out all instructions except for the
// comparison and the branch.
bool AllInvariant = true;
for (BasicBlock::iterator I = ExitingBlock->begin(); &*I != BI; ) {
Instruction *Inst = I++;
// Skip debug info intrinsics.
if (isa<DbgInfoIntrinsic>(Inst))
continue;
if (Inst == CI)
continue;
if (!L->makeLoopInvariant(Inst, Changed,
Preheader ? Preheader->getTerminator() : 0)) {
AllInvariant = false;
break;
}
}
if (!AllInvariant) continue;
// The block has now been cleared of all instructions except for
// a comparison and a conditional branch. SimplifyCFG may be able
// to fold it now.
if (!FoldBranchToCommonDest(BI)) continue;
// Success. The block is now dead, so remove it from the loop,
// update the dominator tree and delete it.
DEBUG(dbgs() << "LoopSimplify: Eliminating exiting block "
<< ExitingBlock->getName() << "\n");
// If any reachable control flow within this loop has changed, notify
// ScalarEvolution. Currently assume the parent loop doesn't change
// (spliting edges doesn't count). If blocks, CFG edges, or other values
// in the parent loop change, then we need call to forgetLoop() for the
// parent instead.
if (SE)
SE->forgetLoop(L);
assert(pred_begin(ExitingBlock) == pred_end(ExitingBlock));
Changed = true;
LI->removeBlock(ExitingBlock);
DomTreeNode *Node = DT->getNode(ExitingBlock);
const std::vector<DomTreeNodeBase<BasicBlock> *> &Children =
Node->getChildren();
while (!Children.empty()) {
DomTreeNode *Child = Children.front();
DT->changeImmediateDominator(Child, Node->getIDom());
}
DT->eraseNode(ExitingBlock);
BI->getSuccessor(0)->removePredecessor(ExitingBlock);
BI->getSuccessor(1)->removePredecessor(ExitingBlock);
ExitingBlock->eraseFromParent();
}
}
return Changed;
}
/// runOnLoop - Remove dead loops, by which we mean loops that do not impact the
/// observable behavior of the program other than finite running time. Note
/// we do ensure that this never remove a loop that might be infinite, as doing
/// so could change the halting/non-halting nature of a program.
/// NOTE: This entire process relies pretty heavily on LoopSimplify and LCSSA
/// in order to make various safety checks work.
bool LoopDeletion::runOnLoop(Loop *L, LPPassManager &LPM) {
// We can only remove the loop if there is a preheader that we can
// branch from after removing it.
BasicBlock *preheader = L->getLoopPreheader();
if (!preheader)
return false;
// If LoopSimplify form is not available, stay out of trouble.
if (!L->hasDedicatedExits())
return false;
// We can't remove loops that contain subloops. If the subloops were dead,
// they would already have been removed in earlier executions of this pass.
if (L->begin() != L->end())
return false;
SmallVector<BasicBlock*, 4> exitingBlocks;
L->getExitingBlocks(exitingBlocks);
SmallVector<BasicBlock*, 4> exitBlocks;
L->getUniqueExitBlocks(exitBlocks);
// We require that the loop only have a single exit block. Otherwise, we'd
// be in the situation of needing to be able to solve statically which exit
// block will be branched to, or trying to preserve the branching logic in
// a loop invariant manner.
if (exitBlocks.size() != 1)
return false;
// Finally, we have to check that the loop really is dead.
bool Changed = false;
if (!isLoopDead(L, exitingBlocks, exitBlocks, Changed, preheader))
return Changed;
// Don't remove loops for which we can't solve the trip count.
// They could be infinite, in which case we'd be changing program behavior.
ScalarEvolution &SE = getAnalysis<ScalarEvolution>();
const SCEV *S = SE.getMaxBackedgeTakenCount(L);
if (isa<SCEVCouldNotCompute>(S))
return Changed;
// Now that we know the removal is safe, remove the loop by changing the
// branch from the preheader to go to the single exit block.
BasicBlock *exitBlock = exitBlocks[0];
// Because we're deleting a large chunk of code at once, the sequence in which
// we remove things is very important to avoid invalidation issues. Don't
// mess with this unless you have good reason and know what you're doing.
// Tell ScalarEvolution that the loop is deleted. Do this before
// deleting the loop so that ScalarEvolution can look at the loop
// to determine what it needs to clean up.
SE.forgetLoop(L);
// Connect the preheader directly to the exit block.
TerminatorInst *TI = preheader->getTerminator();
TI->replaceUsesOfWith(L->getHeader(), exitBlock);
// Rewrite phis in the exit block to get their inputs from
// the preheader instead of the exiting block.
BasicBlock *exitingBlock = exitingBlocks[0];
BasicBlock::iterator BI = exitBlock->begin();
while (PHINode *P = dyn_cast<PHINode>(BI)) {
int j = P->getBasicBlockIndex(exitingBlock);
assert(j >= 0 && "Can't find exiting block in exit block's phi node!");
P->setIncomingBlock(j, preheader);
for (unsigned i = 1; i < exitingBlocks.size(); ++i)
P->removeIncomingValue(exitingBlocks[i]);
++BI;
}
// Update the dominator tree and remove the instructions and blocks that will
// be deleted from the reference counting scheme.
DominatorTree &DT = getAnalysis<DominatorTree>();
SmallVector<DomTreeNode*, 8> ChildNodes;
for (Loop::block_iterator LI = L->block_begin(), LE = L->block_end();
LI != LE; ++LI) {
// Move all of the block's children to be children of the preheader, which
// allows us to remove the domtree entry for the block.
ChildNodes.insert(ChildNodes.begin(), DT[*LI]->begin(), DT[*LI]->end());
for (SmallVectorImpl<DomTreeNode *>::iterator DI = ChildNodes.begin(),
DE = ChildNodes.end(); DI != DE; ++DI) {
DT.changeImmediateDominator(*DI, DT[preheader]);
}
ChildNodes.clear();
DT.eraseNode(*LI);
// Remove the block from the reference counting scheme, so that we can
// delete it freely later.
(*LI)->dropAllReferences();
}
// Erase the instructions and the blocks without having to worry
// about ordering because we already dropped the references.
// NOTE: This iteration is safe because erasing the block does not remove its
// entry from the loop's block list. We do that in the next section.
for (Loop::block_iterator LI = L->block_begin(), LE = L->block_end();
LI != LE; ++LI)
(*LI)->eraseFromParent();
// Finally, the blocks from loopinfo. This has to happen late because
// otherwise our loop iterators won't work.
LoopInfo &loopInfo = getAnalysis<LoopInfo>();
SmallPtrSet<BasicBlock*, 8> blocks;
blocks.insert(L->block_begin(), L->block_end());
for (SmallPtrSet<BasicBlock*,8>::iterator I = blocks.begin(),
E = blocks.end(); I != E; ++I)
loopInfo.removeBlock(*I);
// The last step is to inform the loop pass manager that we've
// eliminated this loop.
LPM.deleteLoopFromQueue(L);
Changed = true;
++NumDeleted;
return Changed;
}
/// At this point, we're committed to promoting the alloca using IDF's, and the
/// standard SSA construction algorithm. Determine which blocks need phi nodes
/// and see if we can optimize out some work by avoiding insertion of dead phi
/// nodes.
void PromoteMem2Reg::DetermineInsertionPoint(AllocaInst *AI, unsigned AllocaNum,
AllocaInfo &Info) {
// Unique the set of defining blocks for efficient lookup.
SmallPtrSet<BasicBlock *, 32> DefBlocks;
DefBlocks.insert(Info.DefiningBlocks.begin(), Info.DefiningBlocks.end());
// Determine which blocks the value is live in. These are blocks which lead
// to uses.
SmallPtrSet<BasicBlock *, 32> LiveInBlocks;
ComputeLiveInBlocks(AI, Info, DefBlocks, LiveInBlocks);
// Use a priority queue keyed on dominator tree level so that inserted nodes
// are handled from the bottom of the dominator tree upwards.
typedef std::priority_queue<DomTreeNodePair,
SmallVector<DomTreeNodePair, 32>,
DomTreeNodeCompare> IDFPriorityQueue;
IDFPriorityQueue PQ;
for (SmallPtrSet<BasicBlock *, 32>::const_iterator I = DefBlocks.begin(),
E = DefBlocks.end();
I != E; ++I) {
if (DomTreeNode *Node = DT.getNode(*I))
PQ.push(std::make_pair(Node, DomLevels[Node]));
}
SmallVector<std::pair<unsigned, BasicBlock *>, 32> DFBlocks;
SmallPtrSet<DomTreeNode *, 32> Visited;
SmallVector<DomTreeNode *, 32> Worklist;
while (!PQ.empty()) {
DomTreeNodePair RootPair = PQ.top();
PQ.pop();
DomTreeNode *Root = RootPair.first;
unsigned RootLevel = RootPair.second;
// Walk all dominator tree children of Root, inspecting their CFG edges with
// targets elsewhere on the dominator tree. Only targets whose level is at
// most Root's level are added to the iterated dominance frontier of the
// definition set.
Worklist.clear();
Worklist.push_back(Root);
while (!Worklist.empty()) {
DomTreeNode *Node = Worklist.pop_back_val();
BasicBlock *BB = Node->getBlock();
for (succ_iterator SI = succ_begin(BB), SE = succ_end(BB); SI != SE;
++SI) {
DomTreeNode *SuccNode = DT.getNode(*SI);
// Quickly skip all CFG edges that are also dominator tree edges instead
// of catching them below.
if (SuccNode->getIDom() == Node)
continue;
unsigned SuccLevel = DomLevels[SuccNode];
if (SuccLevel > RootLevel)
continue;
if (!Visited.insert(SuccNode))
continue;
BasicBlock *SuccBB = SuccNode->getBlock();
if (!LiveInBlocks.count(SuccBB))
continue;
DFBlocks.push_back(std::make_pair(BBNumbers[SuccBB], SuccBB));
if (!DefBlocks.count(SuccBB))
PQ.push(std::make_pair(SuccNode, SuccLevel));
}
for (DomTreeNode::iterator CI = Node->begin(), CE = Node->end(); CI != CE;
++CI) {
if (!Visited.count(*CI))
Worklist.push_back(*CI);
}
}
}
if (DFBlocks.size() > 1)
std::sort(DFBlocks.begin(), DFBlocks.end());
unsigned CurrentVersion = 0;
for (unsigned i = 0, e = DFBlocks.size(); i != e; ++i)
QueuePhiNode(DFBlocks[i].second, AllocaNum, CurrentVersion);
}
/// Unroll the given loop by Count. The loop must be in LCSSA form. Returns true
/// if unrolling was succesful, or false if the loop was unmodified. Unrolling
/// can only fail when the loop's latch block is not terminated by a conditional
/// branch instruction. However, if the trip count (and multiple) are not known,
/// loop unrolling will mostly produce more code that is no faster.
///
/// The LoopInfo Analysis that is passed will be kept consistent.
///
/// If a LoopPassManager is passed in, and the loop is fully removed, it will be
/// removed from the LoopPassManager as well. LPM can also be NULL.
bool llvm::UnrollLoop(Loop *L, unsigned Count, LoopInfo* LI, LPPassManager* LPM) {
assert(L->isLCSSAForm());
BasicBlock *Header = L->getHeader();
BasicBlock *LatchBlock = L->getLoopLatch();
BranchInst *BI = dyn_cast<BranchInst>(LatchBlock->getTerminator());
if (!BI || BI->isUnconditional()) {
// The loop-rotate pass can be helpful to avoid this in many cases.
DOUT << " Can't unroll; loop not terminated by a conditional branch.\n";
return false;
}
// Find trip count
unsigned TripCount = L->getSmallConstantTripCount();
// Find trip multiple if count is not available
unsigned TripMultiple = 1;
if (TripCount == 0)
TripMultiple = L->getSmallConstantTripMultiple();
if (TripCount != 0)
DOUT << " Trip Count = " << TripCount << "\n";
if (TripMultiple != 1)
DOUT << " Trip Multiple = " << TripMultiple << "\n";
// Effectively "DCE" unrolled iterations that are beyond the tripcount
// and will never be executed.
if (TripCount != 0 && Count > TripCount)
Count = TripCount;
assert(Count > 0);
assert(TripMultiple > 0);
assert(TripCount == 0 || TripCount % TripMultiple == 0);
// Are we eliminating the loop control altogether?
bool CompletelyUnroll = Count == TripCount;
// If we know the trip count, we know the multiple...
unsigned BreakoutTrip = 0;
if (TripCount != 0) {
BreakoutTrip = TripCount % Count;
TripMultiple = 0;
} else {
// Figure out what multiple to use.
BreakoutTrip = TripMultiple =
(unsigned)GreatestCommonDivisor64(Count, TripMultiple);
}
if (CompletelyUnroll) {
DEBUG(errs() << "COMPLETELY UNROLLING loop %" << Header->getName()
<< " with trip count " << TripCount << "!\n");
} else {
DEBUG(errs() << "UNROLLING loop %" << Header->getName()
<< " by " << Count);
if (TripMultiple == 0 || BreakoutTrip != TripMultiple) {
DOUT << " with a breakout at trip " << BreakoutTrip;
} else if (TripMultiple != 1) {
DOUT << " with " << TripMultiple << " trips per branch";
}
DOUT << "!\n";
}
std::vector<BasicBlock*> LoopBlocks = L->getBlocks();
bool ContinueOnTrue = L->contains(BI->getSuccessor(0));
BasicBlock *LoopExit = BI->getSuccessor(ContinueOnTrue);
// For the first iteration of the loop, we should use the precloned values for
// PHI nodes. Insert associations now.
typedef DenseMap<const Value*, Value*> ValueMapTy;
ValueMapTy LastValueMap;
std::vector<PHINode*> OrigPHINode;
for (BasicBlock::iterator I = Header->begin(); isa<PHINode>(I); ++I) {
PHINode *PN = cast<PHINode>(I);
OrigPHINode.push_back(PN);
if (Instruction *I =
dyn_cast<Instruction>(PN->getIncomingValueForBlock(LatchBlock)))
if (L->contains(I->getParent()))
LastValueMap[I] = I;
}
std::vector<BasicBlock*> Headers;
std::vector<BasicBlock*> Latches;
Headers.push_back(Header);
Latches.push_back(LatchBlock);
for (unsigned It = 1; It != Count; ++It) {
char SuffixBuffer[100];
sprintf(SuffixBuffer, ".%d", It);
std::vector<BasicBlock*> NewBlocks;
for (std::vector<BasicBlock*>::iterator BB = LoopBlocks.begin(),
E = LoopBlocks.end(); BB != E; ++BB) {
ValueMapTy ValueMap;
BasicBlock *New = CloneBasicBlock(*BB, ValueMap, SuffixBuffer);
Header->getParent()->getBasicBlockList().push_back(New);
// Loop over all of the PHI nodes in the block, changing them to use the
// incoming values from the previous block.
if (*BB == Header)
for (unsigned i = 0, e = OrigPHINode.size(); i != e; ++i) {
PHINode *NewPHI = cast<PHINode>(ValueMap[OrigPHINode[i]]);
Value *InVal = NewPHI->getIncomingValueForBlock(LatchBlock);
if (Instruction *InValI = dyn_cast<Instruction>(InVal))
if (It > 1 && L->contains(InValI->getParent()))
InVal = LastValueMap[InValI];
ValueMap[OrigPHINode[i]] = InVal;
New->getInstList().erase(NewPHI);
}
// Update our running map of newest clones
LastValueMap[*BB] = New;
for (ValueMapTy::iterator VI = ValueMap.begin(), VE = ValueMap.end();
VI != VE; ++VI)
LastValueMap[VI->first] = VI->second;
L->addBasicBlockToLoop(New, LI->getBase());
// Add phi entries for newly created values to all exit blocks except
// the successor of the latch block. The successor of the exit block will
// be updated specially after unrolling all the way.
if (*BB != LatchBlock)
for (Value::use_iterator UI = (*BB)->use_begin(), UE = (*BB)->use_end();
UI != UE;) {
Instruction *UseInst = cast<Instruction>(*UI);
++UI;
if (isa<PHINode>(UseInst) && !L->contains(UseInst->getParent())) {
PHINode *phi = cast<PHINode>(UseInst);
Value *Incoming = phi->getIncomingValueForBlock(*BB);
phi->addIncoming(Incoming, New);
}
}
// Keep track of new headers and latches as we create them, so that
// we can insert the proper branches later.
if (*BB == Header)
Headers.push_back(New);
if (*BB == LatchBlock) {
Latches.push_back(New);
// Also, clear out the new latch's back edge so that it doesn't look
// like a new loop, so that it's amenable to being merged with adjacent
// blocks later on.
TerminatorInst *Term = New->getTerminator();
assert(L->contains(Term->getSuccessor(!ContinueOnTrue)));
assert(Term->getSuccessor(ContinueOnTrue) == LoopExit);
Term->setSuccessor(!ContinueOnTrue, NULL);
}
NewBlocks.push_back(New);
}
// Remap all instructions in the most recent iteration
for (unsigned i = 0; i < NewBlocks.size(); ++i)
for (BasicBlock::iterator I = NewBlocks[i]->begin(),
E = NewBlocks[i]->end(); I != E; ++I)
RemapInstruction(I, LastValueMap);
}
// The latch block exits the loop. If there are any PHI nodes in the
// successor blocks, update them to use the appropriate values computed as the
// last iteration of the loop.
if (Count != 1) {
SmallPtrSet<PHINode*, 8> Users;
for (Value::use_iterator UI = LatchBlock->use_begin(),
UE = LatchBlock->use_end(); UI != UE; ++UI)
if (PHINode *phi = dyn_cast<PHINode>(*UI))
Users.insert(phi);
BasicBlock *LastIterationBB = cast<BasicBlock>(LastValueMap[LatchBlock]);
for (SmallPtrSet<PHINode*,8>::iterator SI = Users.begin(), SE = Users.end();
SI != SE; ++SI) {
PHINode *PN = *SI;
Value *InVal = PN->removeIncomingValue(LatchBlock, false);
// If this value was defined in the loop, take the value defined by the
// last iteration of the loop.
if (Instruction *InValI = dyn_cast<Instruction>(InVal)) {
if (L->contains(InValI->getParent()))
InVal = LastValueMap[InVal];
}
PN->addIncoming(InVal, LastIterationBB);
}
}
// Now, if we're doing complete unrolling, loop over the PHI nodes in the
// original block, setting them to their incoming values.
if (CompletelyUnroll) {
BasicBlock *Preheader = L->getLoopPreheader();
for (unsigned i = 0, e = OrigPHINode.size(); i != e; ++i) {
PHINode *PN = OrigPHINode[i];
PN->replaceAllUsesWith(PN->getIncomingValueForBlock(Preheader));
Header->getInstList().erase(PN);
}
}
// Now that all the basic blocks for the unrolled iterations are in place,
// set up the branches to connect them.
for (unsigned i = 0, e = Latches.size(); i != e; ++i) {
// The original branch was replicated in each unrolled iteration.
BranchInst *Term = cast<BranchInst>(Latches[i]->getTerminator());
// The branch destination.
unsigned j = (i + 1) % e;
BasicBlock *Dest = Headers[j];
bool NeedConditional = true;
// For a complete unroll, make the last iteration end with a branch
// to the exit block.
if (CompletelyUnroll && j == 0) {
Dest = LoopExit;
NeedConditional = false;
}
// If we know the trip count or a multiple of it, we can safely use an
// unconditional branch for some iterations.
if (j != BreakoutTrip && (TripMultiple == 0 || j % TripMultiple != 0)) {
NeedConditional = false;
}
if (NeedConditional) {
// Update the conditional branch's successor for the following
// iteration.
Term->setSuccessor(!ContinueOnTrue, Dest);
} else {
Term->setUnconditionalDest(Dest);
// Merge adjacent basic blocks, if possible.
if (BasicBlock *Fold = FoldBlockIntoPredecessor(Dest, LI)) {
std::replace(Latches.begin(), Latches.end(), Dest, Fold);
std::replace(Headers.begin(), Headers.end(), Dest, Fold);
}
}
}
// At this point, the code is well formed. We now do a quick sweep over the
// inserted code, doing constant propagation and dead code elimination as we
// go.
const std::vector<BasicBlock*> &NewLoopBlocks = L->getBlocks();
for (std::vector<BasicBlock*>::const_iterator BB = NewLoopBlocks.begin(),
BBE = NewLoopBlocks.end(); BB != BBE; ++BB)
for (BasicBlock::iterator I = (*BB)->begin(), E = (*BB)->end(); I != E; ) {
Instruction *Inst = I++;
if (isInstructionTriviallyDead(Inst))
(*BB)->getInstList().erase(Inst);
else if (Constant *C = ConstantFoldInstruction(Inst,
Header->getContext())) {
Inst->replaceAllUsesWith(C);
(*BB)->getInstList().erase(Inst);
}
}
NumCompletelyUnrolled += CompletelyUnroll;
++NumUnrolled;
// Remove the loop from the LoopPassManager if it's completely removed.
if (CompletelyUnroll && LPM != NULL)
LPM->deleteLoopFromQueue(L);
// If we didn't completely unroll the loop, it should still be in LCSSA form.
if (!CompletelyUnroll)
assert(L->isLCSSAForm());
return true;
}
