void PromoteMem2Reg::run() {
Function &F = *DT.getRoot()->getParent();
if (AST) PointerAllocaValues.resize(Allocas.size());
AllocaDbgDeclares.resize(Allocas.size());
AllocaInfo Info;
LargeBlockInfo LBI;
for (unsigned AllocaNum = 0; AllocaNum != Allocas.size(); ++AllocaNum) {
AllocaInst *AI = Allocas[AllocaNum];
assert(isAllocaPromotable(AI) &&
"Cannot promote non-promotable alloca!");
assert(AI->getParent()->getParent() == &F &&
"All allocas should be in the same function, which is same as DF!");
removeLifetimeIntrinsicUsers(AI);
if (AI->use_empty()) {
// If there are no uses of the alloca, just delete it now.
if (AST) AST->deleteValue(AI);
AI->eraseFromParent();
// Remove the alloca from the Allocas list, since it has been processed
RemoveFromAllocasList(AllocaNum);
++NumDeadAlloca;
continue;
}
// Calculate the set of read and write-locations for each alloca. This is
// analogous to finding the 'uses' and 'definitions' of each variable.
Info.AnalyzeAlloca(AI);
// If there is only a single store to this value, replace any loads of
// it that are directly dominated by the definition with the value stored.
if (Info.DefiningBlocks.size() == 1) {
RewriteSingleStoreAlloca(AI, Info, LBI);
// Finally, after the scan, check to see if the store is all that is left.
if (Info.UsingBlocks.empty()) {
// Record debuginfo for the store and remove the declaration's
// debuginfo.
if (DbgDeclareInst *DDI = Info.DbgDeclare) {
if (!DIB)
DIB = new DIBuilder(*DDI->getParent()->getParent()->getParent());
ConvertDebugDeclareToDebugValue(DDI, Info.OnlyStore, *DIB);
DDI->eraseFromParent();
}
// Remove the (now dead) store and alloca.
Info.OnlyStore->eraseFromParent();
LBI.deleteValue(Info.OnlyStore);
if (AST) AST->deleteValue(AI);
AI->eraseFromParent();
LBI.deleteValue(AI);
// The alloca has been processed, move on.
RemoveFromAllocasList(AllocaNum);
++NumSingleStore;
continue;
}
}
// If the alloca is only read and written in one basic block, just perform a
// linear sweep over the block to eliminate it.
if (Info.OnlyUsedInOneBlock) {
PromoteSingleBlockAlloca(AI, Info, LBI);
// Finally, after the scan, check to see if the stores are all that is
// left.
if (Info.UsingBlocks.empty()) {
// Remove the (now dead) stores and alloca.
while (!AI->use_empty()) {
StoreInst *SI = cast<StoreInst>(AI->use_back());
// Record debuginfo for the store before removing it.
if (DbgDeclareInst *DDI = Info.DbgDeclare) {
if (!DIB)
DIB = new DIBuilder(*SI->getParent()->getParent()->getParent());
ConvertDebugDeclareToDebugValue(DDI, SI, *DIB);
}
SI->eraseFromParent();
LBI.deleteValue(SI);
}
if (AST) AST->deleteValue(AI);
AI->eraseFromParent();
LBI.deleteValue(AI);
// The alloca has been processed, move on.
RemoveFromAllocasList(AllocaNum);
// The alloca's debuginfo can be removed as well.
if (DbgDeclareInst *DDI = Info.DbgDeclare)
DDI->eraseFromParent();
++NumLocalPromoted;
continue;
}
}
// If we haven't computed dominator tree levels, do so now.
if (DomLevels.empty()) {
SmallVector<DomTreeNode*, 32> Worklist;
DomTreeNode *Root = DT.getRootNode();
DomLevels[Root] = 0;
Worklist.push_back(Root);
while (!Worklist.empty()) {
DomTreeNode *Node = Worklist.pop_back_val();
unsigned ChildLevel = DomLevels[Node] + 1;
for (DomTreeNode::iterator CI = Node->begin(), CE = Node->end();
CI != CE; ++CI) {
DomLevels[*CI] = ChildLevel;
Worklist.push_back(*CI);
}
}
}
// If we haven't computed a numbering for the BB's in the function, do so
// now.
if (BBNumbers.empty()) {
unsigned ID = 0;
for (Function::iterator I = F.begin(), E = F.end(); I != E; ++I)
BBNumbers[I] = ID++;
}
// If we have an AST to keep updated, remember some pointer value that is
// stored into the alloca.
if (AST)
PointerAllocaValues[AllocaNum] = Info.AllocaPointerVal;
// Remember the dbg.declare intrinsic describing this alloca, if any.
if (Info.DbgDeclare) AllocaDbgDeclares[AllocaNum] = Info.DbgDeclare;
// Keep the reverse mapping of the 'Allocas' array for the rename pass.
AllocaLookup[Allocas[AllocaNum]] = AllocaNum;
// 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.
DetermineInsertionPoint(AI, AllocaNum, Info);
}
if (Allocas.empty())
return; // All of the allocas must have been trivial!
LBI.clear();
// Set the incoming values for the basic block to be null values for all of
// the alloca's. We do this in case there is a load of a value that has not
// been stored yet. In this case, it will get this null value.
//
RenamePassData::ValVector Values(Allocas.size());
for (unsigned i = 0, e = Allocas.size(); i != e; ++i)
Values[i] = UndefValue::get(Allocas[i]->getAllocatedType());
// Walks all basic blocks in the function performing the SSA rename algorithm
// and inserting the phi nodes we marked as necessary
//
std::vector<RenamePassData> RenamePassWorkList;
RenamePassWorkList.push_back(RenamePassData(F.begin(), 0, Values));
do {
RenamePassData RPD;
RPD.swap(RenamePassWorkList.back());
RenamePassWorkList.pop_back();
// RenamePass may add new worklist entries.
RenamePass(RPD.BB, RPD.Pred, RPD.Values, RenamePassWorkList);
} while (!RenamePassWorkList.empty());
// The renamer uses the Visited set to avoid infinite loops. Clear it now.
Visited.clear();
// Remove the allocas themselves from the function.
for (unsigned i = 0, e = Allocas.size(); i != e; ++i) {
Instruction *A = Allocas[i];
// If there are any uses of the alloca instructions left, they must be in
// unreachable basic blocks that were not processed by walking the dominator
// tree. Just delete the users now.
if (!A->use_empty())
A->replaceAllUsesWith(UndefValue::get(A->getType()));
if (AST) AST->deleteValue(A);
A->eraseFromParent();
}
// Remove alloca's dbg.declare instrinsics from the function.
for (unsigned i = 0, e = AllocaDbgDeclares.size(); i != e; ++i)
if (DbgDeclareInst *DDI = AllocaDbgDeclares[i])
DDI->eraseFromParent();
// Loop over all of the PHI nodes and see if there are any that we can get
// rid of because they merge all of the same incoming values. This can
// happen due to undef values coming into the PHI nodes. This process is
// iterative, because eliminating one PHI node can cause others to be removed.
bool EliminatedAPHI = true;
while (EliminatedAPHI) {
EliminatedAPHI = false;
// Iterating over NewPhiNodes is deterministic, so it is safe to try to
// simplify and RAUW them as we go. If it was not, we could add uses to
// the values we replace with in a non deterministic order, thus creating
// non deterministic def->use chains.
for (DenseMap<std::pair<unsigned, unsigned>, PHINode*>::iterator I =
NewPhiNodes.begin(), E = NewPhiNodes.end(); I != E;) {
PHINode *PN = I->second;
// If this PHI node merges one value and/or undefs, get the value.
if (Value *V = SimplifyInstruction(PN, 0, 0, &DT)) {
if (AST && PN->getType()->isPointerTy())
AST->deleteValue(PN);
PN->replaceAllUsesWith(V);
PN->eraseFromParent();
NewPhiNodes.erase(I++);
EliminatedAPHI = true;
continue;
}
++I;
}
}
// At this point, the renamer has added entries to PHI nodes for all reachable
// code. Unfortunately, there may be unreachable blocks which the renamer
// hasn't traversed. If this is the case, the PHI nodes may not
// have incoming values for all predecessors. Loop over all PHI nodes we have
// created, inserting undef values if they are missing any incoming values.
//
for (DenseMap<std::pair<unsigned, unsigned>, PHINode*>::iterator I =
NewPhiNodes.begin(), E = NewPhiNodes.end(); I != E; ++I) {
// We want to do this once per basic block. As such, only process a block
// when we find the PHI that is the first entry in the block.
PHINode *SomePHI = I->second;
BasicBlock *BB = SomePHI->getParent();
if (&BB->front() != SomePHI)
continue;
// Only do work here if there the PHI nodes are missing incoming values. We
// know that all PHI nodes that were inserted in a block will have the same
// number of incoming values, so we can just check any of them.
if (SomePHI->getNumIncomingValues() == getNumPreds(BB))
continue;
// Get the preds for BB.
SmallVector<BasicBlock*, 16> Preds(pred_begin(BB), pred_end(BB));
// Ok, now we know that all of the PHI nodes are missing entries for some
// basic blocks. Start by sorting the incoming predecessors for efficient
// access.
std::sort(Preds.begin(), Preds.end());
// Now we loop through all BB's which have entries in SomePHI and remove
// them from the Preds list.
for (unsigned i = 0, e = SomePHI->getNumIncomingValues(); i != e; ++i) {
// Do a log(n) search of the Preds list for the entry we want.
SmallVector<BasicBlock*, 16>::iterator EntIt =
std::lower_bound(Preds.begin(), Preds.end(),
SomePHI->getIncomingBlock(i));
assert(EntIt != Preds.end() && *EntIt == SomePHI->getIncomingBlock(i)&&
"PHI node has entry for a block which is not a predecessor!");
// Remove the entry
Preds.erase(EntIt);
}
// At this point, the blocks left in the preds list must have dummy
// entries inserted into every PHI nodes for the block. Update all the phi
// nodes in this block that we are inserting (there could be phis before
// mem2reg runs).
unsigned NumBadPreds = SomePHI->getNumIncomingValues();
BasicBlock::iterator BBI = BB->begin();
while ((SomePHI = dyn_cast<PHINode>(BBI++)) &&
SomePHI->getNumIncomingValues() == NumBadPreds) {
Value *UndefVal = UndefValue::get(SomePHI->getType());
for (unsigned pred = 0, e = Preds.size(); pred != e; ++pred)
SomePHI->addIncoming(UndefVal, Preds[pred]);
}
}
NewPhiNodes.clear();
}
/// 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->use_back())->getParent() == BB &&
!isa<PHINode>(Inst->use_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 (Value::use_iterator
UI = Inst->use_begin(), E = Inst->use_end(); UI != E; ++UI) {
Instruction *User = cast<Instruction>(*UI);
if (User->getParent() != BB || isa<PHINode>(User))
Users.push_back(User);
}
// 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());
}
}
/// EliminateMostlyEmptyBlock - Eliminate a basic block that have only phi's and
/// an unconditional branch in it.
void CodeGenPrepare::EliminateMostlyEmptyBlock(BasicBlock *BB) {
BranchInst *BI = cast<BranchInst>(BB->getTerminator());
BasicBlock *DestBB = BI->getSuccessor(0);
DEBUG(dbgs() << "MERGING MOSTLY EMPTY BLOCKS - BEFORE:\n" << *BB << *DestBB);
// If the destination block has a single pred, then this is a trivial edge,
// just collapse it.
if (BasicBlock *SinglePred = DestBB->getSinglePredecessor()) {
if (SinglePred != DestBB) {
// Remember if SinglePred was the entry block of the function. If so, we
// will need to move BB back to the entry position.
bool isEntry = SinglePred == &SinglePred->getParent()->getEntryBlock();
MergeBasicBlockIntoOnlyPred(DestBB, this);
if (isEntry && BB != &BB->getParent()->getEntryBlock())
BB->moveBefore(&BB->getParent()->getEntryBlock());
DEBUG(dbgs() << "AFTER:\n" << *DestBB << "\n\n\n");
return;
}
}
// Otherwise, we have multiple predecessors of BB. Update the PHIs in DestBB
// to handle the new incoming edges it is about to have.
PHINode *PN;
for (BasicBlock::iterator BBI = DestBB->begin();
(PN = dyn_cast<PHINode>(BBI)); ++BBI) {
// Remove the incoming value for BB, and remember it.
Value *InVal = PN->removeIncomingValue(BB, false);
// Two options: either the InVal is a phi node defined in BB or it is some
// value that dominates BB.
PHINode *InValPhi = dyn_cast<PHINode>(InVal);
if (InValPhi && InValPhi->getParent() == BB) {
// Add all of the input values of the input PHI as inputs of this phi.
for (unsigned i = 0, e = InValPhi->getNumIncomingValues(); i != e; ++i)
PN->addIncoming(InValPhi->getIncomingValue(i),
InValPhi->getIncomingBlock(i));
} else {
// Otherwise, add one instance of the dominating value for each edge that
// we will be adding.
if (PHINode *BBPN = dyn_cast<PHINode>(BB->begin())) {
for (unsigned i = 0, e = BBPN->getNumIncomingValues(); i != e; ++i)
PN->addIncoming(InVal, BBPN->getIncomingBlock(i));
} else {
for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI)
PN->addIncoming(InVal, *PI);
}
}
}
// The PHIs are now updated, change everything that refers to BB to use
// DestBB and remove BB.
BB->replaceAllUsesWith(DestBB);
if (DT) {
BasicBlock *BBIDom = DT->getNode(BB)->getIDom()->getBlock();
BasicBlock *DestBBIDom = DT->getNode(DestBB)->getIDom()->getBlock();
BasicBlock *NewIDom = DT->findNearestCommonDominator(BBIDom, DestBBIDom);
DT->changeImmediateDominator(DestBB, NewIDom);
DT->eraseNode(BB);
}
if (PFI) {
PFI->replaceAllUses(BB, DestBB);
PFI->removeEdge(ProfileInfo::getEdge(BB, DestBB));
}
BB->eraseFromParent();
++NumBlocksElim;
DEBUG(dbgs() << "AFTER:\n" << *DestBB << "\n\n\n");
}
/// CloneAndPruneFunctionInto - This works exactly like CloneFunctionInto,
/// except that it does some simple constant prop and DCE on the fly. The
/// effect of this is to copy significantly less code in cases where (for
/// example) a function call with constant arguments is inlined, and those
/// constant arguments cause a significant amount of code in the callee to be
/// dead. Since this doesn't produce an exact copy of the input, it can't be
/// used for things like CloneFunction or CloneModule.
void llvm::CloneAndPruneFunctionInto(Function *NewFunc, const Function *OldFunc,
ValueToValueMapTy &VMap,
SmallVectorImpl<ReturnInst*> &Returns,
const char *NameSuffix,
ClonedCodeInfo *CodeInfo,
const TargetData *TD,
Instruction *TheCall) {
assert(NameSuffix && "NameSuffix cannot be null!");
#ifndef NDEBUG
for (Function::const_arg_iterator II = OldFunc->arg_begin(),
E = OldFunc->arg_end(); II != E; ++II)
assert(VMap.count(II) && "No mapping from source argument specified!");
#endif
PruningFunctionCloner PFC(NewFunc, OldFunc, VMap, Returns,
NameSuffix, CodeInfo, TD);
// Clone the entry block, and anything recursively reachable from it.
std::vector<const BasicBlock*> CloneWorklist;
CloneWorklist.push_back(&OldFunc->getEntryBlock());
while (!CloneWorklist.empty()) {
const BasicBlock *BB = CloneWorklist.back();
CloneWorklist.pop_back();
PFC.CloneBlock(BB, CloneWorklist);
}
// Loop over all of the basic blocks in the old function. If the block was
// reachable, we have cloned it and the old block is now in the value map:
// insert it into the new function in the right order. If not, ignore it.
//
// Defer PHI resolution until rest of function is resolved.
SmallVector<const PHINode*, 16> PHIToResolve;
for (Function::const_iterator BI = OldFunc->begin(), BE = OldFunc->end();
BI != BE; ++BI) {
BasicBlock *NewBB = cast_or_null<BasicBlock>(VMap[BI]);
if (NewBB == 0) continue; // Dead block.
// Add the new block to the new function.
NewFunc->getBasicBlockList().push_back(NewBB);
// Loop over all of the instructions in the block, fixing up operand
// references as we go. This uses VMap to do all the hard work.
//
BasicBlock::iterator I = NewBB->begin();
unsigned DbgKind = OldFunc->getContext().getMDKindID("dbg");
MDNode *TheCallMD = NULL;
if (TheCall && TheCall->hasMetadata())
TheCallMD = TheCall->getMetadata(DbgKind);
// Handle PHI nodes specially, as we have to remove references to dead
// blocks.
if (PHINode *PN = dyn_cast<PHINode>(I)) {
// Skip over all PHI nodes, remembering them for later.
BasicBlock::const_iterator OldI = BI->begin();
for (; (PN = dyn_cast<PHINode>(I)); ++I, ++OldI) {
if (I->hasMetadata()) {
if (TheCallMD) {
if (MDNode *IMD = I->getMetadata(DbgKind)) {
MDNode *NewMD = UpdateInlinedAtInfo(IMD, TheCallMD);
I->setMetadata(DbgKind, NewMD);
}
} else {
// The cloned instruction has dbg info but the call instruction
// does not have dbg info. Remove dbg info from cloned instruction.
I->setMetadata(DbgKind, 0);
}
}
PHIToResolve.push_back(cast<PHINode>(OldI));
}
}
// FIXME:
// FIXME:
// FIXME: Unclone all this metadata stuff.
// FIXME:
// FIXME:
// Otherwise, remap the rest of the instructions normally.
for (; I != NewBB->end(); ++I) {
if (I->hasMetadata()) {
if (TheCallMD) {
if (MDNode *IMD = I->getMetadata(DbgKind)) {
MDNode *NewMD = UpdateInlinedAtInfo(IMD, TheCallMD);
I->setMetadata(DbgKind, NewMD);
}
} else {
// The cloned instruction has dbg info but the call instruction
// does not have dbg info. Remove dbg info from cloned instruction.
I->setMetadata(DbgKind, 0);
}
}
RemapInstruction(I, VMap);
}
}
// Defer PHI resolution until rest of function is resolved, PHI resolution
// requires the CFG to be up-to-date.
for (unsigned phino = 0, e = PHIToResolve.size(); phino != e; ) {
const PHINode *OPN = PHIToResolve[phino];
unsigned NumPreds = OPN->getNumIncomingValues();
const BasicBlock *OldBB = OPN->getParent();
BasicBlock *NewBB = cast<BasicBlock>(VMap[OldBB]);
// Map operands for blocks that are live and remove operands for blocks
// that are dead.
for (; phino != PHIToResolve.size() &&
PHIToResolve[phino]->getParent() == OldBB; ++phino) {
OPN = PHIToResolve[phino];
PHINode *PN = cast<PHINode>(VMap[OPN]);
for (unsigned pred = 0, e = NumPreds; pred != e; ++pred) {
if (BasicBlock *MappedBlock =
cast_or_null<BasicBlock>(VMap[PN->getIncomingBlock(pred)])) {
Value *InVal = MapValue(PN->getIncomingValue(pred),
VMap);
assert(InVal && "Unknown input value?");
PN->setIncomingValue(pred, InVal);
PN->setIncomingBlock(pred, MappedBlock);
} else {
PN->removeIncomingValue(pred, false);
--pred, --e; // Revisit the next entry.
}
}
}
// The loop above has removed PHI entries for those blocks that are dead
// and has updated others. However, if a block is live (i.e. copied over)
// but its terminator has been changed to not go to this block, then our
// phi nodes will have invalid entries. Update the PHI nodes in this
// case.
PHINode *PN = cast<PHINode>(NewBB->begin());
NumPreds = std::distance(pred_begin(NewBB), pred_end(NewBB));
if (NumPreds != PN->getNumIncomingValues()) {
assert(NumPreds < PN->getNumIncomingValues());
// Count how many times each predecessor comes to this block.
std::map<BasicBlock*, unsigned> PredCount;
for (pred_iterator PI = pred_begin(NewBB), E = pred_end(NewBB);
PI != E; ++PI)
--PredCount[*PI];
// Figure out how many entries to remove from each PHI.
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
++PredCount[PN->getIncomingBlock(i)];
// At this point, the excess predecessor entries are positive in the
// map. Loop over all of the PHIs and remove excess predecessor
// entries.
BasicBlock::iterator I = NewBB->begin();
for (; (PN = dyn_cast<PHINode>(I)); ++I) {
for (std::map<BasicBlock*, unsigned>::iterator PCI =PredCount.begin(),
E = PredCount.end(); PCI != E; ++PCI) {
BasicBlock *Pred = PCI->first;
for (unsigned NumToRemove = PCI->second; NumToRemove; --NumToRemove)
PN->removeIncomingValue(Pred, false);
}
}
}
// If the loops above have made these phi nodes have 0 or 1 operand,
// replace them with undef or the input value. We must do this for
// correctness, because 0-operand phis are not valid.
PN = cast<PHINode>(NewBB->begin());
if (PN->getNumIncomingValues() == 0) {
BasicBlock::iterator I = NewBB->begin();
BasicBlock::const_iterator OldI = OldBB->begin();
while ((PN = dyn_cast<PHINode>(I++))) {
Value *NV = UndefValue::get(PN->getType());
PN->replaceAllUsesWith(NV);
assert(VMap[OldI] == PN && "VMap mismatch");
VMap[OldI] = NV;
PN->eraseFromParent();
++OldI;
}
}
// NOTE: We cannot eliminate single entry phi nodes here, because of
// VMap. Single entry phi nodes can have multiple VMap entries
// pointing at them. Thus, deleting one would require scanning the VMap
// to update any entries in it that would require that. This would be
// really slow.
}
// Now that the inlined function body has been fully constructed, go through
// and zap unconditional fall-through branches. This happen all the time when
// specializing code: code specialization turns conditional branches into
// uncond branches, and this code folds them.
Function::iterator I = cast<BasicBlock>(VMap[&OldFunc->getEntryBlock()]);
while (I != NewFunc->end()) {
BranchInst *BI = dyn_cast<BranchInst>(I->getTerminator());
if (!BI || BI->isConditional()) { ++I; continue; }
// Note that we can't eliminate uncond branches if the destination has
// single-entry PHI nodes. Eliminating the single-entry phi nodes would
// require scanning the VMap to update any entries that point to the phi
// node.
BasicBlock *Dest = BI->getSuccessor(0);
if (!Dest->getSinglePredecessor() || isa<PHINode>(Dest->begin())) {
++I; continue;
}
// We know all single-entry PHI nodes in the inlined function have been
// removed, so we just need to splice the blocks.
BI->eraseFromParent();
// Move all the instructions in the succ to the pred.
I->getInstList().splice(I->end(), Dest->getInstList());
// Make all PHI nodes that referred to Dest now refer to I as their source.
Dest->replaceAllUsesWith(I);
// Remove the dest block.
Dest->eraseFromParent();
// Do not increment I, iteratively merge all things this block branches to.
}
}
/// splitLiveRangesAcrossInvokes - Each value that is live across an unwind edge
/// we spill into a stack location, guaranteeing that there is nothing live
/// across the unwind edge. This process also splits all critical edges
/// coming out of invoke's.
void SjLjEHPass::
splitLiveRangesLiveAcrossInvokes(SmallVector<InvokeInst*,16> &Invokes) {
// First step, split all critical edges from invoke instructions.
for (unsigned i = 0, e = Invokes.size(); i != e; ++i) {
InvokeInst *II = Invokes[i];
SplitCriticalEdge(II, 0, this);
SplitCriticalEdge(II, 1, this);
assert(!isa<PHINode>(II->getNormalDest()) &&
!isa<PHINode>(II->getUnwindDest()) &&
"critical edge splitting left single entry phi nodes?");
}
Function *F = Invokes.back()->getParent()->getParent();
// To avoid having to handle incoming arguments specially, we lower each arg
// to a copy instruction in the entry block. This ensures that the argument
// value itself cannot be live across the entry block.
BasicBlock::iterator AfterAllocaInsertPt = F->begin()->begin();
while (isa<AllocaInst>(AfterAllocaInsertPt) &&
isa<ConstantInt>(cast<AllocaInst>(AfterAllocaInsertPt)->getArraySize()))
++AfterAllocaInsertPt;
for (Function::arg_iterator AI = F->arg_begin(), E = F->arg_end();
AI != E; ++AI) {
// This is always a no-op cast because we're casting AI to AI->getType() so
// src and destination types are identical. BitCast is the only possibility.
CastInst *NC = new BitCastInst(
AI, AI->getType(), AI->getName()+".tmp", AfterAllocaInsertPt);
AI->replaceAllUsesWith(NC);
// Normally its is forbidden to replace a CastInst's operand because it
// could cause the opcode to reflect an illegal conversion. However, we're
// replacing it here with the same value it was constructed with to simply
// make NC its user.
NC->setOperand(0, AI);
}
// Finally, scan the code looking for instructions with bad live ranges.
for (Function::iterator BB = F->begin(), E = F->end(); BB != E; ++BB)
for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++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->use_back())->getParent() == BB &&
!isa<PHINode>(Inst->use_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 (Value::use_iterator UI = Inst->use_begin(), E = Inst->use_end();
UI != E; ++UI) {
Instruction *User = cast<Instruction>(*UI);
if (User->getParent() != BB || isa<PHINode>(User))
Users.push_back(User);
}
// Find all of the blocks that this value is live in.
std::set<BasicBlock*> 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)) {
NeedsSpill = true;
}
}
// If we decided we need a spill, do it.
if (NeedsSpill) {
++NumSpilled;
DemoteRegToStack(*Inst, true);
}
}
}
// This function indicates the current limitations in the transform as a result
// of which we do not proceed.
bool LoopInterchangeLegality::currentLimitations() {
BasicBlock *InnerLoopPreHeader = InnerLoop->getLoopPreheader();
BasicBlock *InnerLoopHeader = InnerLoop->getHeader();
BasicBlock *OuterLoopHeader = OuterLoop->getHeader();
BasicBlock *InnerLoopLatch = InnerLoop->getLoopLatch();
BasicBlock *OuterLoopLatch = OuterLoop->getLoopLatch();
PHINode *InnerInductionVar;
PHINode *OuterInductionVar;
// We currently handle only 1 induction variable inside the loop. We also do
// not handle reductions as of now.
if (getPHICount(InnerLoopHeader) > 1)
return true;
if (getPHICount(OuterLoopHeader) > 1)
return true;
InnerInductionVar = getInductionVariable(InnerLoop, SE);
OuterInductionVar = getInductionVariable(OuterLoop, SE);
if (!OuterInductionVar || !InnerInductionVar) {
DEBUG(dbgs() << "Induction variable not found\n");
return true;
}
// TODO: Triangular loops are not handled for now.
if (!isLoopStructureUnderstood(InnerInductionVar)) {
DEBUG(dbgs() << "Loop structure not understood by pass\n");
return true;
}
// TODO: Loops with LCSSA PHI's are currently not handled.
if (isa<PHINode>(OuterLoopLatch->begin())) {
DEBUG(dbgs() << "Found and LCSSA PHI in outer loop latch\n");
return true;
}
if (InnerLoopLatch != InnerLoopHeader &&
isa<PHINode>(InnerLoopLatch->begin())) {
DEBUG(dbgs() << "Found and LCSSA PHI in inner loop latch\n");
return true;
}
// TODO: Current limitation: Since we split the inner loop latch at the point
// were induction variable is incremented (induction.next); We cannot have
// more than 1 user of induction.next since it would result in broken code
// after split.
// e.g.
// for(i=0;i<N;i++) {
// for(j = 0;j<M;j++) {
// A[j+1][i+2] = A[j][i]+k;
// }
// }
bool FoundInduction = false;
Instruction *InnerIndexVarInc = nullptr;
if (InnerInductionVar->getIncomingBlock(0) == InnerLoopPreHeader)
InnerIndexVarInc =
dyn_cast<Instruction>(InnerInductionVar->getIncomingValue(1));
else
InnerIndexVarInc =
dyn_cast<Instruction>(InnerInductionVar->getIncomingValue(0));
if (!InnerIndexVarInc)
return true;
// Since we split the inner loop latch on this induction variable. Make sure
// we do not have any instruction between the induction variable and branch
// instruction.
for (auto I = InnerLoopLatch->rbegin(), E = InnerLoopLatch->rend();
I != E && !FoundInduction; ++I) {
if (isa<BranchInst>(*I) || isa<CmpInst>(*I) || isa<TruncInst>(*I))
continue;
const Instruction &Ins = *I;
// We found an instruction. If this is not induction variable then it is not
// safe to split this loop latch.
if (!Ins.isIdenticalTo(InnerIndexVarInc))
return true;
else
FoundInduction = true;
}
// The loop latch ended and we didnt find the induction variable return as
// current limitation.
if (!FoundInduction)
return true;
return false;
}
/// SplitCriticalEdge - If this edge is a critical edge, insert a new node to
/// split the critical edge. This will update DominatorTree information if it
/// is available, thus calling this pass will not invalidate either of them.
/// This returns the new block if the edge was split, null otherwise.
///
/// If MergeIdenticalEdges is true (not the default), *all* edges from TI to the
/// specified successor will be merged into the same critical edge block.
/// This is most commonly interesting with switch instructions, which may
/// have many edges to any one destination. This ensures that all edges to that
/// dest go to one block instead of each going to a different block, but isn't
/// the standard definition of a "critical edge".
///
/// It is invalid to call this function on a critical edge that starts at an
/// IndirectBrInst. Splitting these edges will almost always create an invalid
/// program because the address of the new block won't be the one that is jumped
/// to.
///
BasicBlock *llvm::SplitCriticalEdge(TerminatorInst *TI, unsigned SuccNum,
Pass *P, bool MergeIdenticalEdges,
bool DontDeleteUselessPhis,
bool SplitLandingPads) {
if (!isCriticalEdge(TI, SuccNum, MergeIdenticalEdges)) return nullptr;
assert(!isa<IndirectBrInst>(TI) &&
"Cannot split critical edge from IndirectBrInst");
BasicBlock *TIBB = TI->getParent();
BasicBlock *DestBB = TI->getSuccessor(SuccNum);
// Splitting the critical edge to a landing pad block is non-trivial. Don't do
// it in this generic function.
if (DestBB->isLandingPad()) return nullptr;
// Create a new basic block, linking it into the CFG.
BasicBlock *NewBB = BasicBlock::Create(TI->getContext(),
TIBB->getName() + "." + DestBB->getName() + "_crit_edge");
// Create our unconditional branch.
BranchInst *NewBI = BranchInst::Create(DestBB, NewBB);
NewBI->setDebugLoc(TI->getDebugLoc());
// Branch to the new block, breaking the edge.
TI->setSuccessor(SuccNum, NewBB);
// Insert the block into the function... right after the block TI lives in.
Function &F = *TIBB->getParent();
Function::iterator FBBI = TIBB;
F.getBasicBlockList().insert(++FBBI, NewBB);
// If there are any PHI nodes in DestBB, we need to update them so that they
// merge incoming values from NewBB instead of from TIBB.
{
unsigned BBIdx = 0;
for (BasicBlock::iterator I = DestBB->begin(); isa<PHINode>(I); ++I) {
// We no longer enter through TIBB, now we come in through NewBB.
// Revector exactly one entry in the PHI node that used to come from
// TIBB to come from NewBB.
PHINode *PN = cast<PHINode>(I);
// Reuse the previous value of BBIdx if it lines up. In cases where we
// have multiple phi nodes with *lots* of predecessors, this is a speed
// win because we don't have to scan the PHI looking for TIBB. This
// happens because the BB list of PHI nodes are usually in the same
// order.
if (PN->getIncomingBlock(BBIdx) != TIBB)
BBIdx = PN->getBasicBlockIndex(TIBB);
PN->setIncomingBlock(BBIdx, NewBB);
}
}
// If there are any other edges from TIBB to DestBB, update those to go
// through the split block, making those edges non-critical as well (and
// reducing the number of phi entries in the DestBB if relevant).
if (MergeIdenticalEdges) {
for (unsigned i = SuccNum+1, e = TI->getNumSuccessors(); i != e; ++i) {
if (TI->getSuccessor(i) != DestBB) continue;
// Remove an entry for TIBB from DestBB phi nodes.
DestBB->removePredecessor(TIBB, DontDeleteUselessPhis);
// We found another edge to DestBB, go to NewBB instead.
TI->setSuccessor(i, NewBB);
}
}
// If we don't have a pass object, we can't update anything...
if (!P) return NewBB;
DominatorTreeWrapperPass *DTWP =
P->getAnalysisIfAvailable<DominatorTreeWrapperPass>();
DominatorTree *DT = DTWP ? &DTWP->getDomTree() : nullptr;
LoopInfo *LI = P->getAnalysisIfAvailable<LoopInfo>();
// If we have nothing to update, just return.
if (!DT && !LI)
return NewBB;
// Now update analysis information. Since the only predecessor of NewBB is
// the TIBB, TIBB clearly dominates NewBB. TIBB usually doesn't dominate
// anything, as there are other successors of DestBB. However, if all other
// predecessors of DestBB are already dominated by DestBB (e.g. DestBB is a
// loop header) then NewBB dominates DestBB.
SmallVector<BasicBlock*, 8> OtherPreds;
// If there is a PHI in the block, loop over predecessors with it, which is
// faster than iterating pred_begin/end.
if (PHINode *PN = dyn_cast<PHINode>(DestBB->begin())) {
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
if (PN->getIncomingBlock(i) != NewBB)
OtherPreds.push_back(PN->getIncomingBlock(i));
} else {
for (pred_iterator I = pred_begin(DestBB), E = pred_end(DestBB);
I != E; ++I) {
BasicBlock *P = *I;
if (P != NewBB)
OtherPreds.push_back(P);
}
}
bool NewBBDominatesDestBB = true;
// Should we update DominatorTree information?
if (DT) {
DomTreeNode *TINode = DT->getNode(TIBB);
// The new block is not the immediate dominator for any other nodes, but
// TINode is the immediate dominator for the new node.
//
if (TINode) { // Don't break unreachable code!
DomTreeNode *NewBBNode = DT->addNewBlock(NewBB, TIBB);
DomTreeNode *DestBBNode = nullptr;
// If NewBBDominatesDestBB hasn't been computed yet, do so with DT.
if (!OtherPreds.empty()) {
DestBBNode = DT->getNode(DestBB);
while (!OtherPreds.empty() && NewBBDominatesDestBB) {
if (DomTreeNode *OPNode = DT->getNode(OtherPreds.back()))
NewBBDominatesDestBB = DT->dominates(DestBBNode, OPNode);
OtherPreds.pop_back();
}
OtherPreds.clear();
}
// If NewBBDominatesDestBB, then NewBB dominates DestBB, otherwise it
// doesn't dominate anything.
if (NewBBDominatesDestBB) {
if (!DestBBNode) DestBBNode = DT->getNode(DestBB);
DT->changeImmediateDominator(DestBBNode, NewBBNode);
}
}
}
// Update LoopInfo if it is around.
if (LI) {
if (Loop *TIL = LI->getLoopFor(TIBB)) {
// If one or the other blocks were not in a loop, the new block is not
// either, and thus LI doesn't need to be updated.
if (Loop *DestLoop = LI->getLoopFor(DestBB)) {
if (TIL == DestLoop) {
// Both in the same loop, the NewBB joins loop.
DestLoop->addBasicBlockToLoop(NewBB, LI->getBase());
} else if (TIL->contains(DestLoop)) {
// Edge from an outer loop to an inner loop. Add to the outer loop.
TIL->addBasicBlockToLoop(NewBB, LI->getBase());
} else if (DestLoop->contains(TIL)) {
// Edge from an inner loop to an outer loop. Add to the outer loop.
DestLoop->addBasicBlockToLoop(NewBB, LI->getBase());
} else {
// Edge from two loops with no containment relation. Because these
// are natural loops, we know that the destination block must be the
// header of its loop (adding a branch into a loop elsewhere would
// create an irreducible loop).
assert(DestLoop->getHeader() == DestBB &&
"Should not create irreducible loops!");
if (Loop *P = DestLoop->getParentLoop())
P->addBasicBlockToLoop(NewBB, LI->getBase());
}
}
// If TIBB is in a loop and DestBB is outside of that loop, we may need
// to update LoopSimplify form and LCSSA form.
if (!TIL->contains(DestBB) &&
P->mustPreserveAnalysisID(LoopSimplifyID)) {
assert(!TIL->contains(NewBB) &&
"Split point for loop exit is contained in loop!");
// Update LCSSA form in the newly created exit block.
if (P->mustPreserveAnalysisID(LCSSAID))
createPHIsForSplitLoopExit(TIBB, NewBB, DestBB);
// The only that we can break LoopSimplify form by splitting a critical
// edge is if after the split there exists some edge from TIL to DestBB
// *and* the only edge into DestBB from outside of TIL is that of
// NewBB. If the first isn't true, then LoopSimplify still holds, NewBB
// is the new exit block and it has no non-loop predecessors. If the
// second isn't true, then DestBB was not in LoopSimplify form prior to
// the split as it had a non-loop predecessor. In both of these cases,
// the predecessor must be directly in TIL, not in a subloop, or again
// LoopSimplify doesn't hold.
SmallVector<BasicBlock *, 4> LoopPreds;
for (pred_iterator I = pred_begin(DestBB), E = pred_end(DestBB); I != E;
++I) {
BasicBlock *P = *I;
if (P == NewBB)
continue; // The new block is known.
if (LI->getLoopFor(P) != TIL) {
// No need to re-simplify, it wasn't to start with.
LoopPreds.clear();
break;
}
LoopPreds.push_back(P);
}
if (!LoopPreds.empty()) {
assert(!DestBB->isLandingPad() &&
"We don't split edges to landing pads!");
BasicBlock *NewExitBB =
SplitBlockPredecessors(DestBB, LoopPreds, "split", P);
if (P->mustPreserveAnalysisID(LCSSAID))
createPHIsForSplitLoopExit(LoopPreds, NewExitBB, DestBB);
}
}
// LCSSA form was updated above for the case where LoopSimplify is
// available, which means that all predecessors of loop exit blocks
// are within the loop. Without LoopSimplify form, it would be
// necessary to insert a new phi.
assert((!P->mustPreserveAnalysisID(LCSSAID) ||
P->mustPreserveAnalysisID(LoopSimplifyID)) &&
"SplitCriticalEdge doesn't know how to update LCCSA form "
"without LoopSimplify!");
}
}
return NewBB;
}
/// severSplitPHINodes - If a PHI node has multiple inputs from outside of the
/// region, we need to split the entry block of the region so that the PHI node
/// is easier to deal with.
void CodeExtractor::severSplitPHINodes(BasicBlock *&Header) {
unsigned NumPredsFromRegion = 0;
unsigned NumPredsOutsideRegion = 0;
if (Header != &Header->getParent()->getEntryBlock()) {
PHINode *PN = dyn_cast<PHINode>(Header->begin());
if (!PN) return; // No PHI nodes.
// If the header node contains any PHI nodes, check to see if there is more
// than one entry from outside the region. If so, we need to sever the
// header block into two.
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
if (Blocks.count(PN->getIncomingBlock(i)))
++NumPredsFromRegion;
else
++NumPredsOutsideRegion;
// If there is one (or fewer) predecessor from outside the region, we don't
// need to do anything special.
if (NumPredsOutsideRegion <= 1) return;
}
// Otherwise, we need to split the header block into two pieces: one
// containing PHI nodes merging values from outside of the region, and a
// second that contains all of the code for the block and merges back any
// incoming values from inside of the region.
BasicBlock *NewBB = llvm::SplitBlock(Header, Header->getFirstNonPHI(), DT);
// We only want to code extract the second block now, and it becomes the new
// header of the region.
BasicBlock *OldPred = Header;
Blocks.remove(OldPred);
Blocks.insert(NewBB);
Header = NewBB;
// Okay, now we need to adjust the PHI nodes and any branches from within the
// region to go to the new header block instead of the old header block.
if (NumPredsFromRegion) {
PHINode *PN = cast<PHINode>(OldPred->begin());
// Loop over all of the predecessors of OldPred that are in the region,
// changing them to branch to NewBB instead.
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
if (Blocks.count(PN->getIncomingBlock(i))) {
TerminatorInst *TI = PN->getIncomingBlock(i)->getTerminator();
TI->replaceUsesOfWith(OldPred, NewBB);
}
// Okay, everything within the region is now branching to the right block, we
// just have to update the PHI nodes now, inserting PHI nodes into NewBB.
BasicBlock::iterator AfterPHIs;
for (AfterPHIs = OldPred->begin(); isa<PHINode>(AfterPHIs); ++AfterPHIs) {
PHINode *PN = cast<PHINode>(AfterPHIs);
// Create a new PHI node in the new region, which has an incoming value
// from OldPred of PN.
PHINode *NewPN = PHINode::Create(PN->getType(), 1 + NumPredsFromRegion,
PN->getName() + ".ce", &NewBB->front());
PN->replaceAllUsesWith(NewPN);
NewPN->addIncoming(PN, OldPred);
// Loop over all of the incoming value in PN, moving them to NewPN if they
// are from the extracted region.
for (unsigned i = 0; i != PN->getNumIncomingValues(); ++i) {
if (Blocks.count(PN->getIncomingBlock(i))) {
NewPN->addIncoming(PN->getIncomingValue(i), PN->getIncomingBlock(i));
PN->removeIncomingValue(i);
--i;
}
}
}
}
}
BasicBlock *getIncomingBlock() { return PHI->getIncomingBlock(idx); }
bool isLinkListLoop(Loop * pLoop, map<PHINode *, set<Value *> > & LinkListHeaderMapping )
{
BasicBlock * pHeader = pLoop->getHeader();
vector<PHINode *> vecToDo;
for(BasicBlock::iterator II = pHeader->begin(); II != pHeader->end(); II ++ )
{
if(isa<PHINode>(II))
{
if(PointerType * pPointerType = dyn_cast<PointerType>(II->getType()))
{
if(isa<StructType>(pPointerType->getElementType()))
{
vecToDo.push_back(cast<PHINode>(II));
}
}
}
else
{
break;
}
}
if(vecToDo.size() == 0)
{
return false;
}
vector<PHINode *>::iterator itVecPhiBegin = vecToDo.begin();
vector<PHINode *>::iterator itVecPhiEnd = vecToDo.end();
for(; itVecPhiBegin != itVecPhiEnd; itVecPhiBegin ++ )
{
PHINode * pPHI = *itVecPhiBegin;
bool bFlag = true;
for(unsigned i = 0; i < pPHI->getNumIncomingValues(); i ++ )
{
if(pLoop->contains(pPHI->getIncomingBlock(i)))
{
if(Instruction * pInst = dyn_cast<Instruction>(pPHI->getIncomingValue(i)))
{
if(!isReachableThroughLinkListDereference(pPHI, pInst, pLoop))
{
bFlag = false;
break;
}
}
else
{
bFlag = false;
break;
}
}
}
if(bFlag)
{
for(unsigned i = 0; i < pPHI->getNumIncomingValues(); i ++ )
{
if(!pLoop->contains(pPHI->getIncomingBlock(i)))
{
LinkListHeaderMapping[pPHI].insert(pPHI->getIncomingValue(i));
}
}
}
}
if(LinkListHeaderMapping.size() > 0)
{
return true;
}
else
{
return false;
}
}
bool llvm::UnrollRuntimeLoopRemainder(Loop *L, unsigned Count,
bool AllowExpensiveTripCount,
bool UseEpilogRemainder,
bool UnrollRemainder, LoopInfo *LI,
ScalarEvolution *SE, DominatorTree *DT,
AssumptionCache *AC, bool PreserveLCSSA,
Loop **ResultLoop) {
LLVM_DEBUG(dbgs() << "Trying runtime unrolling on Loop: \n");
LLVM_DEBUG(L->dump());
LLVM_DEBUG(UseEpilogRemainder ? dbgs() << "Using epilog remainder.\n"
: dbgs() << "Using prolog remainder.\n");
// Make sure the loop is in canonical form.
if (!L->isLoopSimplifyForm()) {
LLVM_DEBUG(dbgs() << "Not in simplify form!\n");
return false;
}
// Guaranteed by LoopSimplifyForm.
BasicBlock *Latch = L->getLoopLatch();
BasicBlock *Header = L->getHeader();
BranchInst *LatchBR = cast<BranchInst>(Latch->getTerminator());
if (!LatchBR || LatchBR->isUnconditional()) {
// The loop-rotate pass can be helpful to avoid this in many cases.
LLVM_DEBUG(
dbgs()
<< "Loop latch not terminated by a conditional branch.\n");
return false;
}
unsigned ExitIndex = LatchBR->getSuccessor(0) == Header ? 1 : 0;
BasicBlock *LatchExit = LatchBR->getSuccessor(ExitIndex);
if (L->contains(LatchExit)) {
// Cloning the loop basic blocks (`CloneLoopBlocks`) requires that one of the
// targets of the Latch be an exit block out of the loop.
LLVM_DEBUG(
dbgs()
<< "One of the loop latch successors must be the exit block.\n");
return false;
}
// These are exit blocks other than the target of the latch exiting block.
SmallVector<BasicBlock *, 4> OtherExits;
bool isMultiExitUnrollingEnabled =
canSafelyUnrollMultiExitLoop(L, OtherExits, LatchExit, PreserveLCSSA,
UseEpilogRemainder) &&
canProfitablyUnrollMultiExitLoop(L, OtherExits, LatchExit, PreserveLCSSA,
UseEpilogRemainder);
// Support only single exit and exiting block unless multi-exit loop unrolling is enabled.
if (!isMultiExitUnrollingEnabled &&
(!L->getExitingBlock() || OtherExits.size())) {
LLVM_DEBUG(
dbgs()
<< "Multiple exit/exiting blocks in loop and multi-exit unrolling not "
"enabled!\n");
return false;
}
// Use Scalar Evolution to compute the trip count. This allows more loops to
// be unrolled than relying on induction var simplification.
if (!SE)
return false;
// Only unroll loops with a computable trip count, and the trip count needs
// to be an int value (allowing a pointer type is a TODO item).
// We calculate the backedge count by using getExitCount on the Latch block,
// which is proven to be the only exiting block in this loop. This is same as
// calculating getBackedgeTakenCount on the loop (which computes SCEV for all
// exiting blocks).
const SCEV *BECountSC = SE->getExitCount(L, Latch);
if (isa<SCEVCouldNotCompute>(BECountSC) ||
!BECountSC->getType()->isIntegerTy()) {
LLVM_DEBUG(dbgs() << "Could not compute exit block SCEV\n");
return false;
}
unsigned BEWidth = cast<IntegerType>(BECountSC->getType())->getBitWidth();
// Add 1 since the backedge count doesn't include the first loop iteration.
const SCEV *TripCountSC =
SE->getAddExpr(BECountSC, SE->getConstant(BECountSC->getType(), 1));
if (isa<SCEVCouldNotCompute>(TripCountSC)) {
LLVM_DEBUG(dbgs() << "Could not compute trip count SCEV.\n");
return false;
}
BasicBlock *PreHeader = L->getLoopPreheader();
BranchInst *PreHeaderBR = cast<BranchInst>(PreHeader->getTerminator());
const DataLayout &DL = Header->getModule()->getDataLayout();
SCEVExpander Expander(*SE, DL, "loop-unroll");
if (!AllowExpensiveTripCount &&
Expander.isHighCostExpansion(TripCountSC, L, PreHeaderBR)) {
LLVM_DEBUG(dbgs() << "High cost for expanding trip count scev!\n");
return false;
}
// This constraint lets us deal with an overflowing trip count easily; see the
// comment on ModVal below.
if (Log2_32(Count) > BEWidth) {
LLVM_DEBUG(
dbgs()
<< "Count failed constraint on overflow trip count calculation.\n");
return false;
}
// Loop structure is the following:
//
// PreHeader
// Header
// ...
// Latch
// LatchExit
BasicBlock *NewPreHeader;
BasicBlock *NewExit = nullptr;
BasicBlock *PrologExit = nullptr;
BasicBlock *EpilogPreHeader = nullptr;
BasicBlock *PrologPreHeader = nullptr;
if (UseEpilogRemainder) {
// If epilog remainder
// Split PreHeader to insert a branch around loop for unrolling.
NewPreHeader = SplitBlock(PreHeader, PreHeader->getTerminator(), DT, LI);
NewPreHeader->setName(PreHeader->getName() + ".new");
// Split LatchExit to create phi nodes from branch above.
SmallVector<BasicBlock*, 4> Preds(predecessors(LatchExit));
NewExit = SplitBlockPredecessors(LatchExit, Preds, ".unr-lcssa", DT, LI,
nullptr, PreserveLCSSA);
// NewExit gets its DebugLoc from LatchExit, which is not part of the
// original Loop.
// Fix this by setting Loop's DebugLoc to NewExit.
auto *NewExitTerminator = NewExit->getTerminator();
NewExitTerminator->setDebugLoc(Header->getTerminator()->getDebugLoc());
// Split NewExit to insert epilog remainder loop.
EpilogPreHeader = SplitBlock(NewExit, NewExitTerminator, DT, LI);
EpilogPreHeader->setName(Header->getName() + ".epil.preheader");
} else {
// If prolog remainder
// Split the original preheader twice to insert prolog remainder loop
PrologPreHeader = SplitEdge(PreHeader, Header, DT, LI);
PrologPreHeader->setName(Header->getName() + ".prol.preheader");
PrologExit = SplitBlock(PrologPreHeader, PrologPreHeader->getTerminator(),
DT, LI);
PrologExit->setName(Header->getName() + ".prol.loopexit");
// Split PrologExit to get NewPreHeader.
NewPreHeader = SplitBlock(PrologExit, PrologExit->getTerminator(), DT, LI);
NewPreHeader->setName(PreHeader->getName() + ".new");
}
// Loop structure should be the following:
// Epilog Prolog
//
// PreHeader PreHeader
// *NewPreHeader *PrologPreHeader
// Header *PrologExit
// ... *NewPreHeader
// Latch Header
// *NewExit ...
// *EpilogPreHeader Latch
// LatchExit LatchExit
// Calculate conditions for branch around loop for unrolling
// in epilog case and around prolog remainder loop in prolog case.
// Compute the number of extra iterations required, which is:
// extra iterations = run-time trip count % loop unroll factor
PreHeaderBR = cast<BranchInst>(PreHeader->getTerminator());
Value *TripCount = Expander.expandCodeFor(TripCountSC, TripCountSC->getType(),
PreHeaderBR);
Value *BECount = Expander.expandCodeFor(BECountSC, BECountSC->getType(),
PreHeaderBR);
IRBuilder<> B(PreHeaderBR);
Value *ModVal;
// Calculate ModVal = (BECount + 1) % Count.
// Note that TripCount is BECount + 1.
if (isPowerOf2_32(Count)) {
// When Count is power of 2 we don't BECount for epilog case, however we'll
// need it for a branch around unrolling loop for prolog case.
ModVal = B.CreateAnd(TripCount, Count - 1, "xtraiter");
// 1. There are no iterations to be run in the prolog/epilog loop.
// OR
// 2. The addition computing TripCount overflowed.
//
// If (2) is true, we know that TripCount really is (1 << BEWidth) and so
// the number of iterations that remain to be run in the original loop is a
// multiple Count == (1 << Log2(Count)) because Log2(Count) <= BEWidth (we
// explicitly check this above).
} else {
// As (BECount + 1) can potentially unsigned overflow we count
// (BECount % Count) + 1 which is overflow safe as BECount % Count < Count.
Value *ModValTmp = B.CreateURem(BECount,
ConstantInt::get(BECount->getType(),
Count));
Value *ModValAdd = B.CreateAdd(ModValTmp,
ConstantInt::get(ModValTmp->getType(), 1));
// At that point (BECount % Count) + 1 could be equal to Count.
// To handle this case we need to take mod by Count one more time.
ModVal = B.CreateURem(ModValAdd,
ConstantInt::get(BECount->getType(), Count),
"xtraiter");
}
Value *BranchVal =
UseEpilogRemainder ? B.CreateICmpULT(BECount,
ConstantInt::get(BECount->getType(),
Count - 1)) :
B.CreateIsNotNull(ModVal, "lcmp.mod");
BasicBlock *RemainderLoop = UseEpilogRemainder ? NewExit : PrologPreHeader;
BasicBlock *UnrollingLoop = UseEpilogRemainder ? NewPreHeader : PrologExit;
// Branch to either remainder (extra iterations) loop or unrolling loop.
B.CreateCondBr(BranchVal, RemainderLoop, UnrollingLoop);
PreHeaderBR->eraseFromParent();
if (DT) {
if (UseEpilogRemainder)
DT->changeImmediateDominator(NewExit, PreHeader);
else
DT->changeImmediateDominator(PrologExit, PreHeader);
}
Function *F = Header->getParent();
// Get an ordered list of blocks in the loop to help with the ordering of the
// cloned blocks in the prolog/epilog code
LoopBlocksDFS LoopBlocks(L);
LoopBlocks.perform(LI);
//
// For each extra loop iteration, create a copy of the loop's basic blocks
// and generate a condition that branches to the copy depending on the
// number of 'left over' iterations.
//
std::vector<BasicBlock *> NewBlocks;
ValueToValueMapTy VMap;
// For unroll factor 2 remainder loop will have 1 iterations.
// Do not create 1 iteration loop.
bool CreateRemainderLoop = (Count != 2);
// Clone all the basic blocks in the loop. If Count is 2, we don't clone
// the loop, otherwise we create a cloned loop to execute the extra
// iterations. This function adds the appropriate CFG connections.
BasicBlock *InsertBot = UseEpilogRemainder ? LatchExit : PrologExit;
BasicBlock *InsertTop = UseEpilogRemainder ? EpilogPreHeader : PrologPreHeader;
Loop *remainderLoop = CloneLoopBlocks(
L, ModVal, CreateRemainderLoop, UseEpilogRemainder, UnrollRemainder,
InsertTop, InsertBot,
NewPreHeader, NewBlocks, LoopBlocks, VMap, DT, LI);
// Insert the cloned blocks into the function.
F->getBasicBlockList().splice(InsertBot->getIterator(),
F->getBasicBlockList(),
NewBlocks[0]->getIterator(),
F->end());
// Now the loop blocks are cloned and the other exiting blocks from the
// remainder are connected to the original Loop's exit blocks. The remaining
// work is to update the phi nodes in the original loop, and take in the
// values from the cloned region.
for (auto *BB : OtherExits) {
for (auto &II : *BB) {
// Given we preserve LCSSA form, we know that the values used outside the
// loop will be used through these phi nodes at the exit blocks that are
// transformed below.
if (!isa<PHINode>(II))
break;
PHINode *Phi = cast<PHINode>(&II);
unsigned oldNumOperands = Phi->getNumIncomingValues();
// Add the incoming values from the remainder code to the end of the phi
// node.
for (unsigned i =0; i < oldNumOperands; i++){
Value *newVal = VMap.lookup(Phi->getIncomingValue(i));
// newVal can be a constant or derived from values outside the loop, and
// hence need not have a VMap value. Also, since lookup already generated
// a default "null" VMap entry for this value, we need to populate that
// VMap entry correctly, with the mapped entry being itself.
if (!newVal) {
newVal = Phi->getIncomingValue(i);
VMap[Phi->getIncomingValue(i)] = Phi->getIncomingValue(i);
}
Phi->addIncoming(newVal,
cast<BasicBlock>(VMap[Phi->getIncomingBlock(i)]));
}
}
#if defined(EXPENSIVE_CHECKS) && !defined(NDEBUG)
for (BasicBlock *SuccBB : successors(BB)) {
assert(!(any_of(OtherExits,
[SuccBB](BasicBlock *EB) { return EB == SuccBB; }) ||
SuccBB == LatchExit) &&
"Breaks the definition of dedicated exits!");
}
#endif
}
// Update the immediate dominator of the exit blocks and blocks that are
// reachable from the exit blocks. This is needed because we now have paths
// from both the original loop and the remainder code reaching the exit
// blocks. While the IDom of these exit blocks were from the original loop,
// now the IDom is the preheader (which decides whether the original loop or
// remainder code should run).
if (DT && !L->getExitingBlock()) {
SmallVector<BasicBlock *, 16> ChildrenToUpdate;
// NB! We have to examine the dom children of all loop blocks, not just
// those which are the IDom of the exit blocks. This is because blocks
// reachable from the exit blocks can have their IDom as the nearest common
// dominator of the exit blocks.
for (auto *BB : L->blocks()) {
auto *DomNodeBB = DT->getNode(BB);
for (auto *DomChild : DomNodeBB->getChildren()) {
auto *DomChildBB = DomChild->getBlock();
if (!L->contains(LI->getLoopFor(DomChildBB)))
ChildrenToUpdate.push_back(DomChildBB);
}
}
for (auto *BB : ChildrenToUpdate)
DT->changeImmediateDominator(BB, PreHeader);
}
// Loop structure should be the following:
// Epilog Prolog
//
// PreHeader PreHeader
// NewPreHeader PrologPreHeader
// Header PrologHeader
// ... ...
// Latch PrologLatch
// NewExit PrologExit
// EpilogPreHeader NewPreHeader
// EpilogHeader Header
// ... ...
// EpilogLatch Latch
// LatchExit LatchExit
// Rewrite the cloned instruction operands to use the values created when the
// clone is created.
for (BasicBlock *BB : NewBlocks) {
for (Instruction &I : *BB) {
RemapInstruction(&I, VMap,
RF_NoModuleLevelChanges | RF_IgnoreMissingLocals);
}
}
if (UseEpilogRemainder) {
// Connect the epilog code to the original loop and update the
// PHI functions.
ConnectEpilog(L, ModVal, NewExit, LatchExit, PreHeader,
EpilogPreHeader, NewPreHeader, VMap, DT, LI,
PreserveLCSSA);
// Update counter in loop for unrolling.
// I should be multiply of Count.
IRBuilder<> B2(NewPreHeader->getTerminator());
Value *TestVal = B2.CreateSub(TripCount, ModVal, "unroll_iter");
BranchInst *LatchBR = cast<BranchInst>(Latch->getTerminator());
B2.SetInsertPoint(LatchBR);
PHINode *NewIdx = PHINode::Create(TestVal->getType(), 2, "niter",
Header->getFirstNonPHI());
Value *IdxSub =
B2.CreateSub(NewIdx, ConstantInt::get(NewIdx->getType(), 1),
NewIdx->getName() + ".nsub");
Value *IdxCmp;
if (LatchBR->getSuccessor(0) == Header)
IdxCmp = B2.CreateIsNotNull(IdxSub, NewIdx->getName() + ".ncmp");
else
IdxCmp = B2.CreateIsNull(IdxSub, NewIdx->getName() + ".ncmp");
NewIdx->addIncoming(TestVal, NewPreHeader);
NewIdx->addIncoming(IdxSub, Latch);
LatchBR->setCondition(IdxCmp);
} else {
// Connect the prolog code to the original loop and update the
// PHI functions.
ConnectProlog(L, BECount, Count, PrologExit, LatchExit, PreHeader,
NewPreHeader, VMap, DT, LI, PreserveLCSSA);
}
// If this loop is nested, then the loop unroller changes the code in the any
// of its parent loops, so the Scalar Evolution pass needs to be run again.
SE->forgetTopmostLoop(L);
// Verify that the Dom Tree is correct.
#if defined(EXPENSIVE_CHECKS) && !defined(NDEBUG)
if (DT)
assert(DT->verify(DominatorTree::VerificationLevel::Full));
#endif
// Canonicalize to LoopSimplifyForm both original and remainder loops. We
// cannot rely on the LoopUnrollPass to do this because it only does
// canonicalization for parent/subloops and not the sibling loops.
if (OtherExits.size() > 0) {
// Generate dedicated exit blocks for the original loop, to preserve
// LoopSimplifyForm.
formDedicatedExitBlocks(L, DT, LI, nullptr, PreserveLCSSA);
// Generate dedicated exit blocks for the remainder loop if one exists, to
// preserve LoopSimplifyForm.
if (remainderLoop)
formDedicatedExitBlocks(remainderLoop, DT, LI, nullptr, PreserveLCSSA);
}
auto UnrollResult = LoopUnrollResult::Unmodified;
if (remainderLoop && UnrollRemainder) {
LLVM_DEBUG(dbgs() << "Unrolling remainder loop\n");
UnrollResult =
UnrollLoop(remainderLoop, /*Count*/ Count - 1, /*TripCount*/ Count - 1,
/*Force*/ false, /*AllowRuntime*/ false,
/*AllowExpensiveTripCount*/ false, /*PreserveCondBr*/ true,
/*PreserveOnlyFirst*/ false, /*TripMultiple*/ 1,
/*PeelCount*/ 0, /*UnrollRemainder*/ false, LI, SE, DT, AC,
/*ORE*/ nullptr, PreserveLCSSA);
}
if (ResultLoop && UnrollResult != LoopUnrollResult::FullyUnrolled)
*ResultLoop = remainderLoop;
NumRuntimeUnrolled++;
return true;
}
/// InsertUniqueBackedgeBlock - This method is called when the specified loop
/// has more than one backedge in it. If this occurs, revector all of these
/// backedges to target a new basic block and have that block branch to the loop
/// header. This ensures that loops have exactly one backedge.
///
BasicBlock *
LoopSimplify::InsertUniqueBackedgeBlock(Loop *L, BasicBlock *Preheader) {
assert(L->getNumBackEdges() > 1 && "Must have > 1 backedge!");
// Get information about the loop
BasicBlock *Header = L->getHeader();
Function *F = Header->getParent();
// Unique backedge insertion currently depends on having a preheader.
if (!Preheader)
return 0;
// Figure out which basic blocks contain back-edges to the loop header.
std::vector<BasicBlock*> BackedgeBlocks;
for (pred_iterator I = pred_begin(Header), E = pred_end(Header); I != E; ++I){
BasicBlock *P = *I;
// Indirectbr edges cannot be split, so we must fail if we find one.
if (isa<IndirectBrInst>(P->getTerminator()))
return 0;
if (P != Preheader) BackedgeBlocks.push_back(P);
}
// Create and insert the new backedge block...
BasicBlock *BEBlock = BasicBlock::Create(Header->getContext(),
Header->getName()+".backedge", F);
BranchInst *BETerminator = BranchInst::Create(Header, BEBlock);
DEBUG(dbgs() << "LoopSimplify: Inserting unique backedge block "
<< BEBlock->getName() << "\n");
// Move the new backedge block to right after the last backedge block.
Function::iterator InsertPos = BackedgeBlocks.back(); ++InsertPos;
F->getBasicBlockList().splice(InsertPos, F->getBasicBlockList(), BEBlock);
// Now that the block has been inserted into the function, create PHI nodes in
// the backedge block which correspond to any PHI nodes in the header block.
for (BasicBlock::iterator I = Header->begin(); isa<PHINode>(I); ++I) {
PHINode *PN = cast<PHINode>(I);
PHINode *NewPN = PHINode::Create(PN->getType(), PN->getName()+".be",
BETerminator);
NewPN->reserveOperandSpace(BackedgeBlocks.size());
if (AA) AA->copyValue(PN, NewPN);
// Loop over the PHI node, moving all entries except the one for the
// preheader over to the new PHI node.
unsigned PreheaderIdx = ~0U;
bool HasUniqueIncomingValue = true;
Value *UniqueValue = 0;
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) {
BasicBlock *IBB = PN->getIncomingBlock(i);
Value *IV = PN->getIncomingValue(i);
if (IBB == Preheader) {
PreheaderIdx = i;
} else {
NewPN->addIncoming(IV, IBB);
if (HasUniqueIncomingValue) {
if (UniqueValue == 0)
UniqueValue = IV;
else if (UniqueValue != IV)
HasUniqueIncomingValue = false;
}
}
}
// Delete all of the incoming values from the old PN except the preheader's
assert(PreheaderIdx != ~0U && "PHI has no preheader entry??");
if (PreheaderIdx != 0) {
PN->setIncomingValue(0, PN->getIncomingValue(PreheaderIdx));
PN->setIncomingBlock(0, PN->getIncomingBlock(PreheaderIdx));
}
// Nuke all entries except the zero'th.
for (unsigned i = 0, e = PN->getNumIncomingValues()-1; i != e; ++i)
PN->removeIncomingValue(e-i, false);
// Finally, add the newly constructed PHI node as the entry for the BEBlock.
PN->addIncoming(NewPN, BEBlock);
// As an optimization, if all incoming values in the new PhiNode (which is a
// subset of the incoming values of the old PHI node) have the same value,
// eliminate the PHI Node.
if (HasUniqueIncomingValue) {
NewPN->replaceAllUsesWith(UniqueValue);
if (AA) AA->deleteValue(NewPN);
BEBlock->getInstList().erase(NewPN);
}
}
// Now that all of the PHI nodes have been inserted and adjusted, modify the
// backedge blocks to just to the BEBlock instead of the header.
for (unsigned i = 0, e = BackedgeBlocks.size(); i != e; ++i) {
TerminatorInst *TI = BackedgeBlocks[i]->getTerminator();
for (unsigned Op = 0, e = TI->getNumSuccessors(); Op != e; ++Op)
if (TI->getSuccessor(Op) == Header)
TI->setSuccessor(Op, BEBlock);
}
//===--- Update all analyses which we must preserve now -----------------===//
// Update Loop Information - we know that this block is now in the current
// loop and all parent loops.
L->addBasicBlockToLoop(BEBlock, LI->getBase());
// Update dominator information
DT->splitBlock(BEBlock);
return BEBlock;
}
void RegionExtractor::findInputsOutputs(ValueSet &Inputs,
ValueSet &Outputs) const {
for (SetVector<BasicBlock *>::const_iterator I = Blocks.begin(),
E = Blocks.end();
I != E; ++I) {
BasicBlock *BB = *I;
// If a used value is defined outside the region, it's an input. If an
// instruction is used outside the region, it's an output.
for (BasicBlock::iterator II = BB->begin(), IE = BB->end();
II != IE; ++II) {
for (User::op_iterator OI = II->op_begin(), OE = II->op_end();
OI != OE; ++OI)
if (definedInCaller(Blocks, *OI))
Inputs.insert(*OI);
#if LLVM_VERSION_MINOR == 5
for (User *U : II->users())
if (!definedInRegion(Blocks, U)) {
#else
for (Value::use_iterator UI = II->use_begin(), UE = II->use_end();
UI != UE; ++UI)
if (!definedInRegion(Blocks, *UI)) {
#endif
Outputs.insert(II);
break;
}
}
}
}
/// severSplitPHINodes - If a PHI node has multiple inputs from outside of the
/// region, we need to split the entry block of the region so that the PHI node
/// is easier to deal with.
void RegionExtractor::severSplitPHINodes(BasicBlock *&Header) {
unsigned NumPredsFromRegion = 0;
unsigned NumPredsOutsideRegion = 0;
if (Header != &Header->getParent()->getEntryBlock()) {
PHINode *PN = dyn_cast<PHINode>(Header->begin());
if (!PN) return; // No PHI nodes.
// If the header node contains any PHI nodes, check to see if there is more
// than one entry from outside the region. If so, we need to sever the
// header block into two.
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
if (Blocks.count(PN->getIncomingBlock(i)))
++NumPredsFromRegion;
else
++NumPredsOutsideRegion;
// If there is one (or fewer) predecessor from outside the region, we don't
// need to do anything special.
if (NumPredsOutsideRegion <= 1) return;
}
// Otherwise, we need to split the header block into two pieces: one
// containing PHI nodes merging values from outside of the region, and a
// second that contains all of the code for the block and merges back any
// incoming values from inside of the region.
BasicBlock::iterator AfterPHIs = Header->getFirstNonPHI();
BasicBlock *NewBB = Header->splitBasicBlock(AfterPHIs,
Header->getName()+".ce");
// We only want to code extract the second block now, and it becomes the new
// header of the region.
BasicBlock *OldPred = Header;
Blocks.remove(OldPred);
Blocks.insert(NewBB);
Header = NewBB;
// Okay, update dominator sets. The blocks that dominate the new one are the
// blocks that dominate TIBB plus the new block itself.
if (DT)
DT->splitBlock(NewBB);
// Okay, now we need to adjust the PHI nodes and any branches from within the
// region to go to the new header block instead of the old header block.
if (NumPredsFromRegion) {
PHINode *PN = cast<PHINode>(OldPred->begin());
// Loop over all of the predecessors of OldPred that are in the region,
// changing them to branch to NewBB instead.
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
if (Blocks.count(PN->getIncomingBlock(i))) {
TerminatorInst *TI = PN->getIncomingBlock(i)->getTerminator();
TI->replaceUsesOfWith(OldPred, NewBB);
}
// Okay, everything within the region is now branching to the right block, we
// just have to update the PHI nodes now, inserting PHI nodes into NewBB.
for (AfterPHIs = OldPred->begin(); isa<PHINode>(AfterPHIs); ++AfterPHIs) {
PHINode *PN = cast<PHINode>(AfterPHIs);
// Create a new PHI node in the new region, which has an incoming value
// from OldPred of PN.
PHINode *NewPN = PHINode::Create(PN->getType(), 1 + NumPredsFromRegion,
PN->getName()+".ce", NewBB->begin());
NewPN->addIncoming(PN, OldPred);
// Loop over all of the incoming value in PN, moving them to NewPN if they
// are from the extracted region.
for (unsigned i = 0; i != PN->getNumIncomingValues(); ++i) {
if (Blocks.count(PN->getIncomingBlock(i))) {
NewPN->addIncoming(PN->getIncomingValue(i), PN->getIncomingBlock(i));
PN->removeIncomingValue(i);
--i;
}
}
}
}
}
/// This works like CloneAndPruneFunctionInto, except that it does not clone the
/// entire function. Instead it starts at an instruction provided by the caller
/// and copies (and prunes) only the code reachable from that instruction.
void llvm::CloneAndPruneIntoFromInst(Function *NewFunc, const Function *OldFunc,
const Instruction *StartingInst,
ValueToValueMapTy &VMap,
bool ModuleLevelChanges,
SmallVectorImpl<ReturnInst *> &Returns,
const char *NameSuffix,
ClonedCodeInfo *CodeInfo,
CloningDirector *Director) {
assert(NameSuffix && "NameSuffix cannot be null!");
ValueMapTypeRemapper *TypeMapper = nullptr;
ValueMaterializer *Materializer = nullptr;
if (Director) {
TypeMapper = Director->getTypeRemapper();
Materializer = Director->getValueMaterializer();
}
#ifndef NDEBUG
// If the cloning starts at the beginning of the function, verify that
// the function arguments are mapped.
if (!StartingInst)
for (const Argument &II : OldFunc->args())
assert(VMap.count(&II) && "No mapping from source argument specified!");
#endif
PruningFunctionCloner PFC(NewFunc, OldFunc, VMap, ModuleLevelChanges,
NameSuffix, CodeInfo, Director);
const BasicBlock *StartingBB;
if (StartingInst)
StartingBB = StartingInst->getParent();
else {
StartingBB = &OldFunc->getEntryBlock();
StartingInst = &StartingBB->front();
}
// Clone the entry block, and anything recursively reachable from it.
std::vector<const BasicBlock*> CloneWorklist;
PFC.CloneBlock(StartingBB, StartingInst->getIterator(), CloneWorklist);
while (!CloneWorklist.empty()) {
const BasicBlock *BB = CloneWorklist.back();
CloneWorklist.pop_back();
PFC.CloneBlock(BB, BB->begin(), CloneWorklist);
}
// Loop over all of the basic blocks in the old function. If the block was
// reachable, we have cloned it and the old block is now in the value map:
// insert it into the new function in the right order. If not, ignore it.
//
// Defer PHI resolution until rest of function is resolved.
SmallVector<const PHINode*, 16> PHIToResolve;
for (const BasicBlock &BI : *OldFunc) {
Value *V = VMap[&BI];
BasicBlock *NewBB = cast_or_null<BasicBlock>(V);
if (!NewBB) continue; // Dead block.
// Add the new block to the new function.
NewFunc->getBasicBlockList().push_back(NewBB);
// Handle PHI nodes specially, as we have to remove references to dead
// blocks.
for (BasicBlock::const_iterator I = BI.begin(), E = BI.end(); I != E; ++I) {
// PHI nodes may have been remapped to non-PHI nodes by the caller or
// during the cloning process.
if (const PHINode *PN = dyn_cast<PHINode>(I)) {
if (isa<PHINode>(VMap[PN]))
PHIToResolve.push_back(PN);
else
break;
} else {
break;
}
}
// Finally, remap the terminator instructions, as those can't be remapped
// until all BBs are mapped.
RemapInstruction(NewBB->getTerminator(), VMap,
ModuleLevelChanges ? RF_None : RF_NoModuleLevelChanges,
TypeMapper, Materializer);
}
// Defer PHI resolution until rest of function is resolved, PHI resolution
// requires the CFG to be up-to-date.
for (unsigned phino = 0, e = PHIToResolve.size(); phino != e; ) {
const PHINode *OPN = PHIToResolve[phino];
unsigned NumPreds = OPN->getNumIncomingValues();
const BasicBlock *OldBB = OPN->getParent();
BasicBlock *NewBB = cast<BasicBlock>(VMap[OldBB]);
// Map operands for blocks that are live and remove operands for blocks
// that are dead.
for (; phino != PHIToResolve.size() &&
PHIToResolve[phino]->getParent() == OldBB; ++phino) {
OPN = PHIToResolve[phino];
PHINode *PN = cast<PHINode>(VMap[OPN]);
for (unsigned pred = 0, e = NumPreds; pred != e; ++pred) {
Value *V = VMap[PN->getIncomingBlock(pred)];
if (BasicBlock *MappedBlock = cast_or_null<BasicBlock>(V)) {
Value *InVal = MapValue(PN->getIncomingValue(pred),
VMap,
ModuleLevelChanges ? RF_None : RF_NoModuleLevelChanges);
assert(InVal && "Unknown input value?");
PN->setIncomingValue(pred, InVal);
PN->setIncomingBlock(pred, MappedBlock);
} else {
PN->removeIncomingValue(pred, false);
--pred, --e; // Revisit the next entry.
}
}
}
// The loop above has removed PHI entries for those blocks that are dead
// and has updated others. However, if a block is live (i.e. copied over)
// but its terminator has been changed to not go to this block, then our
// phi nodes will have invalid entries. Update the PHI nodes in this
// case.
PHINode *PN = cast<PHINode>(NewBB->begin());
NumPreds = std::distance(pred_begin(NewBB), pred_end(NewBB));
if (NumPreds != PN->getNumIncomingValues()) {
assert(NumPreds < PN->getNumIncomingValues());
// Count how many times each predecessor comes to this block.
std::map<BasicBlock*, unsigned> PredCount;
for (pred_iterator PI = pred_begin(NewBB), E = pred_end(NewBB);
PI != E; ++PI)
--PredCount[*PI];
// Figure out how many entries to remove from each PHI.
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
++PredCount[PN->getIncomingBlock(i)];
// At this point, the excess predecessor entries are positive in the
// map. Loop over all of the PHIs and remove excess predecessor
// entries.
BasicBlock::iterator I = NewBB->begin();
for (; (PN = dyn_cast<PHINode>(I)); ++I) {
for (std::map<BasicBlock*, unsigned>::iterator PCI =PredCount.begin(),
E = PredCount.end(); PCI != E; ++PCI) {
BasicBlock *Pred = PCI->first;
for (unsigned NumToRemove = PCI->second; NumToRemove; --NumToRemove)
PN->removeIncomingValue(Pred, false);
}
}
}
// If the loops above have made these phi nodes have 0 or 1 operand,
// replace them with undef or the input value. We must do this for
// correctness, because 0-operand phis are not valid.
PN = cast<PHINode>(NewBB->begin());
if (PN->getNumIncomingValues() == 0) {
BasicBlock::iterator I = NewBB->begin();
BasicBlock::const_iterator OldI = OldBB->begin();
while ((PN = dyn_cast<PHINode>(I++))) {
Value *NV = UndefValue::get(PN->getType());
PN->replaceAllUsesWith(NV);
assert(VMap[&*OldI] == PN && "VMap mismatch");
VMap[&*OldI] = NV;
PN->eraseFromParent();
++OldI;
}
}
}
// Make a second pass over the PHINodes now that all of them have been
// remapped into the new function, simplifying the PHINode and performing any
// recursive simplifications exposed. This will transparently update the
// WeakVH in the VMap. Notably, we rely on that so that if we coalesce
// two PHINodes, the iteration over the old PHIs remains valid, and the
// mapping will just map us to the new node (which may not even be a PHI
// node).
for (unsigned Idx = 0, Size = PHIToResolve.size(); Idx != Size; ++Idx)
if (PHINode *PN = dyn_cast<PHINode>(VMap[PHIToResolve[Idx]]))
recursivelySimplifyInstruction(PN);
// Now that the inlined function body has been fully constructed, go through
// and zap unconditional fall-through branches. This happens all the time when
// specializing code: code specialization turns conditional branches into
// uncond branches, and this code folds them.
Function::iterator Begin = cast<BasicBlock>(VMap[StartingBB])->getIterator();
Function::iterator I = Begin;
while (I != NewFunc->end()) {
// Check if this block has become dead during inlining or other
// simplifications. Note that the first block will appear dead, as it has
// not yet been wired up properly.
if (I != Begin && (pred_begin(&*I) == pred_end(&*I) ||
I->getSinglePredecessor() == &*I)) {
BasicBlock *DeadBB = &*I++;
DeleteDeadBlock(DeadBB);
continue;
}
// We need to simplify conditional branches and switches with a constant
// operand. We try to prune these out when cloning, but if the
// simplification required looking through PHI nodes, those are only
// available after forming the full basic block. That may leave some here,
// and we still want to prune the dead code as early as possible.
ConstantFoldTerminator(&*I);
BranchInst *BI = dyn_cast<BranchInst>(I->getTerminator());
if (!BI || BI->isConditional()) { ++I; continue; }
BasicBlock *Dest = BI->getSuccessor(0);
if (!Dest->getSinglePredecessor()) {
++I; continue;
}
// We shouldn't be able to get single-entry PHI nodes here, as instsimplify
// above should have zapped all of them..
assert(!isa<PHINode>(Dest->begin()));
// We know all single-entry PHI nodes in the inlined function have been
// removed, so we just need to splice the blocks.
BI->eraseFromParent();
// Make all PHI nodes that referred to Dest now refer to I as their source.
Dest->replaceAllUsesWith(&*I);
// Move all the instructions in the succ to the pred.
I->getInstList().splice(I->end(), Dest->getInstList());
// Remove the dest block.
Dest->eraseFromParent();
// Do not increment I, iteratively merge all things this block branches to.
}
// Make a final pass over the basic blocks from the old function to gather
// any return instructions which survived folding. We have to do this here
// because we can iteratively remove and merge returns above.
for (Function::iterator I = cast<BasicBlock>(VMap[StartingBB])->getIterator(),
E = NewFunc->end();
I != E; ++I)
if (ReturnInst *RI = dyn_cast<ReturnInst>(I->getTerminator()))
Returns.push_back(RI);
}
/// \brief Recursively handle the condition leading to a loop
Value *SIAnnotateControlFlow::handleLoopCondition(Value *Cond, PHINode *Broken,
llvm::Loop *L, BranchInst *Term) {
// Only search through PHI nodes which are inside the loop. If we try this
// with PHI nodes that are outside of the loop, we end up inserting new PHI
// nodes outside of the loop which depend on values defined inside the loop.
// This will break the module with
// 'Instruction does not dominate all users!' errors.
PHINode *Phi = nullptr;
if ((Phi = dyn_cast<PHINode>(Cond)) && L->contains(Phi)) {
BasicBlock *Parent = Phi->getParent();
PHINode *NewPhi = PHINode::Create(Int64, 0, "", &Parent->front());
Value *Ret = NewPhi;
// Handle all non-constant incoming values first
for (unsigned i = 0, e = Phi->getNumIncomingValues(); i != e; ++i) {
Value *Incoming = Phi->getIncomingValue(i);
BasicBlock *From = Phi->getIncomingBlock(i);
if (isa<ConstantInt>(Incoming)) {
NewPhi->addIncoming(Broken, From);
continue;
}
Phi->setIncomingValue(i, BoolFalse);
Value *PhiArg = handleLoopCondition(Incoming, Broken, L, Term);
NewPhi->addIncoming(PhiArg, From);
}
BasicBlock *IDom = DT->getNode(Parent)->getIDom()->getBlock();
for (unsigned i = 0, e = Phi->getNumIncomingValues(); i != e; ++i) {
Value *Incoming = Phi->getIncomingValue(i);
if (Incoming != BoolTrue)
continue;
BasicBlock *From = Phi->getIncomingBlock(i);
if (From == IDom) {
// We're in the following situation:
// IDom/From
// | \
// | If-block
// | /
// Parent
// where we want to break out of the loop if the If-block is not taken.
// Due to the depth-first traversal, there should be an end.cf
// intrinsic in Parent, and we insert an else.break before it.
//
// Note that the end.cf need not be the first non-phi instruction
// of parent, particularly when we're dealing with a multi-level
// break, but it should occur within a group of intrinsic calls
// at the beginning of the block.
CallInst *OldEnd = dyn_cast<CallInst>(Parent->getFirstInsertionPt());
while (OldEnd && OldEnd->getCalledFunction() != EndCf)
OldEnd = dyn_cast<CallInst>(OldEnd->getNextNode());
if (OldEnd && OldEnd->getCalledFunction() == EndCf) {
Value *Args[] = { OldEnd->getArgOperand(0), NewPhi };
Ret = CallInst::Create(ElseBreak, Args, "", OldEnd);
continue;
}
}
TerminatorInst *Insert = From->getTerminator();
Value *PhiArg = CallInst::Create(Break, Broken, "", Insert);
NewPhi->setIncomingValue(i, PhiArg);
}
eraseIfUnused(Phi);
return Ret;
} else if (Instruction *Inst = dyn_cast<Instruction>(Cond)) {
BasicBlock *Parent = Inst->getParent();
Instruction *Insert;
if (L->contains(Inst)) {
Insert = Parent->getTerminator();
} else {
Insert = L->getHeader()->getFirstNonPHIOrDbgOrLifetime();
}
Value *Args[] = { Cond, Broken };
return CallInst::Create(IfBreak, Args, "", Insert);
// Insert IfBreak before TERM for constant COND.
} else if (isa<ConstantInt>(Cond)) {
Value *Args[] = { Cond, Broken };
return CallInst::Create(IfBreak, Args, "", Term);
} else {
llvm_unreachable("Unhandled loop condition!");
}
return nullptr;
}
// First thing we need to do is scan the whole function for values that are
// live across unwind edges. Each value that is live across an unwind edge
// we spill into a stack location, guaranteeing that there is nothing live
// across the unwind edge. This process also splits all critical edges
// coming out of invoke's.
void LowerInvoke::
splitLiveRangesLiveAcrossInvokes(SmallVectorImpl<InvokeInst*> &Invokes) {
// First step, split all critical edges from invoke instructions.
for (unsigned i = 0, e = Invokes.size(); i != e; ++i) {
InvokeInst *II = Invokes[i];
SplitCriticalEdge(II, 0, this);
SplitCriticalEdge(II, 1, this);
assert(!isa<PHINode>(II->getNormalDest()) &&
!isa<PHINode>(II->getUnwindDest()) &&
"critical edge splitting left single entry phi nodes?");
}
Function *F = Invokes.back()->getParent()->getParent();
// To avoid having to handle incoming arguments specially, we lower each arg
// to a copy instruction in the entry block. This ensures that the argument
// value itself cannot be live across the entry block.
BasicBlock::iterator AfterAllocaInsertPt = F->begin()->begin();
while (isa<AllocaInst>(AfterAllocaInsertPt) &&
isa<ConstantInt>(cast<AllocaInst>(AfterAllocaInsertPt)->getArraySize()))
++AfterAllocaInsertPt;
for (Function::arg_iterator AI = F->arg_begin(), E = F->arg_end();
AI != E; ++AI) {
Type *Ty = AI->getType();
// Aggregate types can't be cast, but are legal argument types, so we have
// to handle them differently. We use an extract/insert pair as a
// lightweight method to achieve the same goal.
if (isa<StructType>(Ty) || isa<ArrayType>(Ty) || isa<VectorType>(Ty)) {
Instruction *EI = ExtractValueInst::Create(AI, 0, "",AfterAllocaInsertPt);
Instruction *NI = InsertValueInst::Create(AI, EI, 0);
NI->insertAfter(EI);
AI->replaceAllUsesWith(NI);
// Set the operand of the instructions back to the AllocaInst.
EI->setOperand(0, AI);
NI->setOperand(0, AI);
} else {
// This is always a no-op cast because we're casting AI to AI->getType()
// so src and destination types are identical. BitCast is the only
// possibility.
CastInst *NC = new BitCastInst(
AI, AI->getType(), AI->getName()+".tmp", AfterAllocaInsertPt);
AI->replaceAllUsesWith(NC);
// Set the operand of the cast instruction back to the AllocaInst.
// Normally it's forbidden to replace a CastInst's operand because it
// could cause the opcode to reflect an illegal conversion. However,
// we're replacing it here with the same value it was constructed with.
// We do this because the above replaceAllUsesWith() clobbered the
// operand, but we want this one to remain.
NC->setOperand(0, AI);
}
}
// Finally, scan the code looking for instructions with bad live ranges.
for (Function::iterator BB = F->begin(), E = F->end(); BB != E; ++BB)
for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++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);
}
// Scan all of the uses and see if the live range is live across an unwind
// edge. If we find a use live across an invoke edge, create an alloca
// and spill the value.
// Find all of the blocks that this value is live in.
std::set<BasicBlock*> 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)) {
NeedsSpill = true;
}
}
// If we decided we need a spill, do it.
if (NeedsSpill) {
++NumSpilled;
DemoteRegToStack(*Inst, true);
}
}
}
/// DupRetToEnableTailCallOpts - Look for opportunities to duplicate return
/// instructions to the predecessor to enable tail call optimizations. The
/// case it is currently looking for is:
/// bb0:
/// %tmp0 = tail call i32 @f0()
/// br label %return
/// bb1:
/// %tmp1 = tail call i32 @f1()
/// br label %return
/// bb2:
/// %tmp2 = tail call i32 @f2()
/// br label %return
/// return:
/// %retval = phi i32 [ %tmp0, %bb0 ], [ %tmp1, %bb1 ], [ %tmp2, %bb2 ]
/// ret i32 %retval
///
/// =>
///
/// bb0:
/// %tmp0 = tail call i32 @f0()
/// ret i32 %tmp0
/// bb1:
/// %tmp1 = tail call i32 @f1()
/// ret i32 %tmp1
/// bb2:
/// %tmp2 = tail call i32 @f2()
/// ret i32 %tmp2
///
bool CodeGenPrepare::DupRetToEnableTailCallOpts(ReturnInst *RI) {
if (!TLI)
return false;
Value *V = RI->getReturnValue();
PHINode *PN = V ? dyn_cast<PHINode>(V) : NULL;
if (V && !PN)
return false;
BasicBlock *BB = RI->getParent();
if (PN && PN->getParent() != BB)
return false;
// It's not safe to eliminate the sign / zero extension of the return value.
// See llvm::isInTailCallPosition().
const Function *F = BB->getParent();
Attributes CallerRetAttr = F->getAttributes().getRetAttributes();
if ((CallerRetAttr & Attribute::ZExt) || (CallerRetAttr & Attribute::SExt))
return false;
// Make sure there are no instructions between the PHI and return, or that the
// return is the first instruction in the block.
if (PN) {
BasicBlock::iterator BI = BB->begin();
do {
++BI;
}
while (isa<DbgInfoIntrinsic>(BI));
if (&*BI != RI)
return false;
} else {
BasicBlock::iterator BI = BB->begin();
while (isa<DbgInfoIntrinsic>(BI)) ++BI;
if (&*BI != RI)
return false;
}
/// Only dup the ReturnInst if the CallInst is likely to be emitted as a tail
/// call.
SmallVector<CallInst*, 4> TailCalls;
if (PN) {
for (unsigned I = 0, E = PN->getNumIncomingValues(); I != E; ++I) {
CallInst *CI = dyn_cast<CallInst>(PN->getIncomingValue(I));
// Make sure the phi value is indeed produced by the tail call.
if (CI && CI->hasOneUse() && CI->getParent() == PN->getIncomingBlock(I) &&
TLI->mayBeEmittedAsTailCall(CI))
TailCalls.push_back(CI);
}
} else {
SmallPtrSet<BasicBlock*, 4> VisitedBBs;
for (pred_iterator PI = pred_begin(BB), PE = pred_end(BB); PI != PE; ++PI) {
if (!VisitedBBs.insert(*PI))
continue;
BasicBlock::InstListType &InstList = (*PI)->getInstList();
BasicBlock::InstListType::reverse_iterator RI = InstList.rbegin();
BasicBlock::InstListType::reverse_iterator RE = InstList.rend();
do {
++RI;
}
while (RI != RE && isa<DbgInfoIntrinsic>(&*RI));
if (RI == RE)
continue;
CallInst *CI = dyn_cast<CallInst>(&*RI);
if (CI && CI->use_empty() && TLI->mayBeEmittedAsTailCall(CI))
TailCalls.push_back(CI);
}
}
bool Changed = false;
for (unsigned i = 0, e = TailCalls.size(); i != e; ++i) {
CallInst *CI = TailCalls[i];
CallSite CS(CI);
// Conservatively require the attributes of the call to match those of the
// return. Ignore noalias because it doesn't affect the call sequence.
Attributes CalleeRetAttr = CS.getAttributes().getRetAttributes();
if ((CalleeRetAttr ^ CallerRetAttr) & ~Attribute::NoAlias)
continue;
// Make sure the call instruction is followed by an unconditional branch to
// the return block.
BasicBlock *CallBB = CI->getParent();
BranchInst *BI = dyn_cast<BranchInst>(CallBB->getTerminator());
if (!BI || !BI->isUnconditional() || BI->getSuccessor(0) != BB)
continue;
// Duplicate the return into CallBB.
(void)FoldReturnIntoUncondBranch(RI, BB, CallBB);
ModifiedDT = Changed = true;
++NumRetsDup;
}
// If we eliminated all predecessors of the block, delete the block now.
if (Changed && pred_begin(BB) == pred_end(BB))
BB->eraseFromParent();
return Changed;
}
/// This method is used to notify a BasicBlock that the
/// specified Predecessor of the block is no longer able to reach it. This is
/// actually not used to update the Predecessor list, but is actually used to
/// update the PHI nodes that reside in the block. Note that this should be
/// called while the predecessor still refers to this block.
///
void BasicBlock::removePredecessor(BasicBlock *Pred,
bool DontDeleteUselessPHIs) {
assert((hasNUsesOrMore(16)||// Reduce cost of this assertion for complex CFGs.
find(pred_begin(this), pred_end(this), Pred) != pred_end(this)) &&
"removePredecessor: BB is not a predecessor!");
if (InstList.empty()) return;
PHINode *APN = dyn_cast<PHINode>(&front());
if (!APN) return; // Quick exit.
// If there are exactly two predecessors, then we want to nuke the PHI nodes
// altogether. However, we cannot do this, if this in this case:
//
// Loop:
// %x = phi [X, Loop]
// %x2 = add %x, 1 ;; This would become %x2 = add %x2, 1
// br Loop ;; %x2 does not dominate all uses
//
// This is because the PHI node input is actually taken from the predecessor
// basic block. The only case this can happen is with a self loop, so we
// check for this case explicitly now.
//
unsigned max_idx = APN->getNumIncomingValues();
assert(max_idx != 0 && "PHI Node in block with 0 predecessors!?!?!");
if (max_idx == 2) {
BasicBlock *Other = APN->getIncomingBlock(APN->getIncomingBlock(0) == Pred);
// Disable PHI elimination!
if (this == Other) max_idx = 3;
}
// <= Two predecessors BEFORE I remove one?
if (max_idx <= 2 && !DontDeleteUselessPHIs) {
// Yup, loop through and nuke the PHI nodes
while (PHINode *PN = dyn_cast<PHINode>(&front())) {
// Remove the predecessor first.
PN->removeIncomingValue(Pred, !DontDeleteUselessPHIs);
// If the PHI _HAD_ two uses, replace PHI node with its now *single* value
if (max_idx == 2) {
if (PN->getIncomingValue(0) != PN)
PN->replaceAllUsesWith(PN->getIncomingValue(0));
else
// We are left with an infinite loop with no entries: kill the PHI.
PN->replaceAllUsesWith(UndefValue::get(PN->getType()));
getInstList().pop_front(); // Remove the PHI node
}
// If the PHI node already only had one entry, it got deleted by
// removeIncomingValue.
}
} else {
// Okay, now we know that we need to remove predecessor #pred_idx from all
// PHI nodes. Iterate over each PHI node fixing them up
PHINode *PN;
for (iterator II = begin(); (PN = dyn_cast<PHINode>(II)); ) {
++II;
PN->removeIncomingValue(Pred, false);
// If all incoming values to the Phi are the same, we can replace the Phi
// with that value.
Value* PNV = nullptr;
if (!DontDeleteUselessPHIs && (PNV = PN->hasConstantValue()))
if (PNV != PN) {
PN->replaceAllUsesWith(PNV);
PN->eraseFromParent();
}
}
}
}
/// RewriteLoopExitValues - Check to see if this loop has a computable
/// loop-invariant execution count. If so, this means that we can compute the
/// final value of any expressions that are recurrent in the loop, and
/// substitute the exit values from the loop into any instructions outside of
/// the loop that use the final values of the current expressions.
///
/// This is mostly redundant with the regular IndVarSimplify activities that
/// happen later, except that it's more powerful in some cases, because it's
/// able to brute-force evaluate arbitrary instructions as long as they have
/// constant operands at the beginning of the loop.
void IndVarSimplify::RewriteLoopExitValues(Loop *L,
SCEVExpander &Rewriter) {
// Verify the input to the pass in already in LCSSA form.
assert(L->isLCSSAForm(*DT));
SmallVector<BasicBlock*, 8> ExitBlocks;
L->getUniqueExitBlocks(ExitBlocks);
// Find all values that are computed inside the loop, but used outside of it.
// Because of LCSSA, these values will only occur in LCSSA PHI Nodes. Scan
// the exit blocks of the loop to find them.
for (unsigned i = 0, e = ExitBlocks.size(); i != e; ++i) {
BasicBlock *ExitBB = ExitBlocks[i];
// If there are no PHI nodes in this exit block, then no values defined
// inside the loop are used on this path, skip it.
PHINode *PN = dyn_cast<PHINode>(ExitBB->begin());
if (!PN) continue;
unsigned NumPreds = PN->getNumIncomingValues();
// Iterate over all of the PHI nodes.
BasicBlock::iterator BBI = ExitBB->begin();
while ((PN = dyn_cast<PHINode>(BBI++))) {
if (PN->use_empty())
continue; // dead use, don't replace it
// SCEV only supports integer expressions for now.
if (!PN->getType()->isIntegerTy() && !PN->getType()->isPointerTy())
continue;
// It's necessary to tell ScalarEvolution about this explicitly so that
// it can walk the def-use list and forget all SCEVs, as it may not be
// watching the PHI itself. Once the new exit value is in place, there
// may not be a def-use connection between the loop and every instruction
// which got a SCEVAddRecExpr for that loop.
SE->forgetValue(PN);
// Iterate over all of the values in all the PHI nodes.
for (unsigned i = 0; i != NumPreds; ++i) {
// If the value being merged in is not integer or is not defined
// in the loop, skip it.
Value *InVal = PN->getIncomingValue(i);
if (!isa<Instruction>(InVal))
continue;
// If this pred is for a subloop, not L itself, skip it.
if (LI->getLoopFor(PN->getIncomingBlock(i)) != L)
continue; // The Block is in a subloop, skip it.
// Check that InVal is defined in the loop.
Instruction *Inst = cast<Instruction>(InVal);
if (!L->contains(Inst))
continue;
// Okay, this instruction has a user outside of the current loop
// and varies predictably *inside* the loop. Evaluate the value it
// contains when the loop exits, if possible.
const SCEV *ExitValue = SE->getSCEVAtScope(Inst, L->getParentLoop());
if (!ExitValue->isLoopInvariant(L))
continue;
Changed = true;
++NumReplaced;
Value *ExitVal = Rewriter.expandCodeFor(ExitValue, PN->getType(), Inst);
DEBUG(dbgs() << "INDVARS: RLEV: AfterLoopVal = " << *ExitVal << '\n'
<< " LoopVal = " << *Inst << "\n");
PN->setIncomingValue(i, ExitVal);
// If this instruction is dead now, delete it.
RecursivelyDeleteTriviallyDeadInstructions(Inst);
if (NumPreds == 1) {
// Completely replace a single-pred PHI. This is safe, because the
// NewVal won't be variant in the loop, so we don't need an LCSSA phi
// node anymore.
PN->replaceAllUsesWith(ExitVal);
RecursivelyDeleteTriviallyDeadInstructions(PN);
}
}
if (NumPreds != 1) {
// Clone the PHI and delete the original one. This lets IVUsers and
// any other maps purge the original user from their records.
PHINode *NewPN = cast<PHINode>(PN->clone());
NewPN->takeName(PN);
NewPN->insertBefore(PN);
PN->replaceAllUsesWith(NewPN);
PN->eraseFromParent();
}
}
}
// The insertion point instruction may have been deleted; clear it out
// so that the rewriter doesn't trip over it later.
Rewriter.clearInsertPoint();
}
/// \brief If this loop has multiple backedges, try to pull one of them out into
/// a nested loop.
///
/// This is important for code that looks like
/// this:
///
/// Loop:
/// ...
/// br cond, Loop, Next
/// ...
/// br cond2, Loop, Out
///
/// To identify this common case, we look at the PHI nodes in the header of the
/// loop. PHI nodes with unchanging values on one backedge correspond to values
/// that change in the "outer" loop, but not in the "inner" loop.
///
/// If we are able to separate out a loop, return the new outer loop that was
/// created.
///
static Loop *separateNestedLoop(Loop *L, BasicBlock *Preheader,
AliasAnalysis *AA, DominatorTree *DT,
LoopInfo *LI, ScalarEvolution *SE, Pass *PP) {
// Don't try to separate loops without a preheader.
if (!Preheader)
return nullptr;
// The header is not a landing pad; preheader insertion should ensure this.
assert(!L->getHeader()->isLandingPad() &&
"Can't insert backedge to landing pad");
PHINode *PN = findPHIToPartitionLoops(L, AA, DT);
if (!PN) return nullptr; // No known way to partition.
// Pull out all predecessors that have varying values in the loop. This
// handles the case when a PHI node has multiple instances of itself as
// arguments.
SmallVector<BasicBlock*, 8> OuterLoopPreds;
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) {
if (PN->getIncomingValue(i) != PN ||
!L->contains(PN->getIncomingBlock(i))) {
// We can't split indirectbr edges.
if (isa<IndirectBrInst>(PN->getIncomingBlock(i)->getTerminator()))
return nullptr;
OuterLoopPreds.push_back(PN->getIncomingBlock(i));
}
}
DEBUG(dbgs() << "LoopSimplify: Splitting out a new outer loop\n");
// If ScalarEvolution is around and knows anything about values in
// this loop, tell it to forget them, because we're about to
// substantially change it.
if (SE)
SE->forgetLoop(L);
BasicBlock *Header = L->getHeader();
BasicBlock *NewBB =
SplitBlockPredecessors(Header, OuterLoopPreds, ".outer", PP);
// Make sure that NewBB is put someplace intelligent, which doesn't mess up
// code layout too horribly.
placeSplitBlockCarefully(NewBB, OuterLoopPreds, L);
// Create the new outer loop.
Loop *NewOuter = new Loop();
// Change the parent loop to use the outer loop as its child now.
if (Loop *Parent = L->getParentLoop())
Parent->replaceChildLoopWith(L, NewOuter);
else
LI->changeTopLevelLoop(L, NewOuter);
// L is now a subloop of our outer loop.
NewOuter->addChildLoop(L);
for (Loop::block_iterator I = L->block_begin(), E = L->block_end();
I != E; ++I)
NewOuter->addBlockEntry(*I);
// Now reset the header in L, which had been moved by
// SplitBlockPredecessors for the outer loop.
L->moveToHeader(Header);
// Determine which blocks should stay in L and which should be moved out to
// the Outer loop now.
std::set<BasicBlock*> BlocksInL;
for (pred_iterator PI=pred_begin(Header), E = pred_end(Header); PI!=E; ++PI) {
BasicBlock *P = *PI;
if (DT->dominates(Header, P))
addBlockAndPredsToSet(P, Header, BlocksInL);
}
// Scan all of the loop children of L, moving them to OuterLoop if they are
// not part of the inner loop.
const std::vector<Loop*> &SubLoops = L->getSubLoops();
for (size_t I = 0; I != SubLoops.size(); )
if (BlocksInL.count(SubLoops[I]->getHeader()))
++I; // Loop remains in L
else
NewOuter->addChildLoop(L->removeChildLoop(SubLoops.begin() + I));
// Now that we know which blocks are in L and which need to be moved to
// OuterLoop, move any blocks that need it.
for (unsigned i = 0; i != L->getBlocks().size(); ++i) {
BasicBlock *BB = L->getBlocks()[i];
if (!BlocksInL.count(BB)) {
// Move this block to the parent, updating the exit blocks sets
L->removeBlockFromLoop(BB);
if ((*LI)[BB] == L)
LI->changeLoopFor(BB, NewOuter);
--i;
}
}
return NewOuter;
}
/// SplitCriticalEdge - If this edge is a critical edge, insert a new node to
/// split the critical edge. This will update DominatorTree and
/// DominatorFrontier information if it is available, thus calling this pass
/// will not invalidate either of them. This returns the new block if the edge
/// was split, null otherwise.
///
/// If MergeIdenticalEdges is true (not the default), *all* edges from TI to the
/// specified successor will be merged into the same critical edge block.
/// This is most commonly interesting with switch instructions, which may
/// have many edges to any one destination. This ensures that all edges to that
/// dest go to one block instead of each going to a different block, but isn't
/// the standard definition of a "critical edge".
///
/// It is invalid to call this function on a critical edge that starts at an
/// IndirectBrInst. Splitting these edges will almost always create an invalid
/// program because the address of the new block won't be the one that is jumped
/// to.
///
BasicBlock *llvm::SplitCriticalEdge(TerminatorInst *TI, unsigned SuccNum,
Pass *P, bool MergeIdenticalEdges) {
if (!isCriticalEdge(TI, SuccNum, MergeIdenticalEdges)) return 0;
assert(!isa<IndirectBrInst>(TI) &&
"Cannot split critical edge from IndirectBrInst");
BasicBlock *TIBB = TI->getParent();
BasicBlock *DestBB = TI->getSuccessor(SuccNum);
// Create a new basic block, linking it into the CFG.
BasicBlock *NewBB = BasicBlock::Create(TI->getContext(),
TIBB->getName() + "." + DestBB->getName() + "_crit_edge");
// Create our unconditional branch.
BranchInst::Create(DestBB, NewBB);
// Branch to the new block, breaking the edge.
TI->setSuccessor(SuccNum, NewBB);
// Insert the block into the function... right after the block TI lives in.
Function &F = *TIBB->getParent();
Function::iterator FBBI = TIBB;
F.getBasicBlockList().insert(++FBBI, NewBB);
// If there are any PHI nodes in DestBB, we need to update them so that they
// merge incoming values from NewBB instead of from TIBB.
if (PHINode *APHI = dyn_cast<PHINode>(DestBB->begin())) {
// This conceptually does:
// foreach (PHINode *PN in DestBB)
// PN->setIncomingBlock(PN->getIncomingBlock(TIBB), NewBB);
// but is optimized for two cases.
if (APHI->getNumIncomingValues() <= 8) { // Small # preds case.
unsigned BBIdx = 0;
for (BasicBlock::iterator I = DestBB->begin(); isa<PHINode>(I); ++I) {
// We no longer enter through TIBB, now we come in through NewBB.
// Revector exactly one entry in the PHI node that used to come from
// TIBB to come from NewBB.
PHINode *PN = cast<PHINode>(I);
// Reuse the previous value of BBIdx if it lines up. In cases where we
// have multiple phi nodes with *lots* of predecessors, this is a speed
// win because we don't have to scan the PHI looking for TIBB. This
// happens because the BB list of PHI nodes are usually in the same
// order.
if (PN->getIncomingBlock(BBIdx) != TIBB)
BBIdx = PN->getBasicBlockIndex(TIBB);
PN->setIncomingBlock(BBIdx, NewBB);
}
} else {
// However, the foreach loop is slow for blocks with lots of predecessors
// because PHINode::getIncomingBlock is O(n) in # preds. Instead, walk
// the user list of TIBB to find the PHI nodes.
SmallPtrSet<PHINode*, 16> UpdatedPHIs;
for (Value::use_iterator UI = TIBB->use_begin(), E = TIBB->use_end();
UI != E; ) {
Value::use_iterator Use = UI++;
if (PHINode *PN = dyn_cast<PHINode>(*Use)) {
// Remove one entry from each PHI.
if (PN->getParent() == DestBB && UpdatedPHIs.insert(PN))
PN->setOperand(Use.getOperandNo(), NewBB);
}
}
}
}
// If there are any other edges from TIBB to DestBB, update those to go
// through the split block, making those edges non-critical as well (and
// reducing the number of phi entries in the DestBB if relevant).
if (MergeIdenticalEdges) {
for (unsigned i = SuccNum+1, e = TI->getNumSuccessors(); i != e; ++i) {
if (TI->getSuccessor(i) != DestBB) continue;
// Remove an entry for TIBB from DestBB phi nodes.
DestBB->removePredecessor(TIBB);
// We found another edge to DestBB, go to NewBB instead.
TI->setSuccessor(i, NewBB);
}
}
// If we don't have a pass object, we can't update anything...
if (P == 0) return NewBB;
DominatorTree *DT = P->getAnalysisIfAvailable<DominatorTree>();
DominanceFrontier *DF = P->getAnalysisIfAvailable<DominanceFrontier>();
LoopInfo *LI = P->getAnalysisIfAvailable<LoopInfo>();
ProfileInfo *PI = P->getAnalysisIfAvailable<ProfileInfo>();
// If we have nothing to update, just return.
if (DT == 0 && DF == 0 && LI == 0 && PI == 0)
return NewBB;
// Now update analysis information. Since the only predecessor of NewBB is
// the TIBB, TIBB clearly dominates NewBB. TIBB usually doesn't dominate
// anything, as there are other successors of DestBB. However, if all other
// predecessors of DestBB are already dominated by DestBB (e.g. DestBB is a
// loop header) then NewBB dominates DestBB.
SmallVector<BasicBlock*, 8> OtherPreds;
// If there is a PHI in the block, loop over predecessors with it, which is
// faster than iterating pred_begin/end.
if (PHINode *PN = dyn_cast<PHINode>(DestBB->begin())) {
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
if (PN->getIncomingBlock(i) != NewBB)
OtherPreds.push_back(PN->getIncomingBlock(i));
} else {
for (pred_iterator I = pred_begin(DestBB), E = pred_end(DestBB);
I != E; ++I) {
BasicBlock *P = *I;
if (P != NewBB)
OtherPreds.push_back(P);
}
}
bool NewBBDominatesDestBB = true;
// Should we update DominatorTree information?
if (DT) {
DomTreeNode *TINode = DT->getNode(TIBB);
// The new block is not the immediate dominator for any other nodes, but
// TINode is the immediate dominator for the new node.
//
if (TINode) { // Don't break unreachable code!
DomTreeNode *NewBBNode = DT->addNewBlock(NewBB, TIBB);
DomTreeNode *DestBBNode = 0;
// If NewBBDominatesDestBB hasn't been computed yet, do so with DT.
if (!OtherPreds.empty()) {
DestBBNode = DT->getNode(DestBB);
while (!OtherPreds.empty() && NewBBDominatesDestBB) {
if (DomTreeNode *OPNode = DT->getNode(OtherPreds.back()))
NewBBDominatesDestBB = DT->dominates(DestBBNode, OPNode);
OtherPreds.pop_back();
}
OtherPreds.clear();
}
// If NewBBDominatesDestBB, then NewBB dominates DestBB, otherwise it
// doesn't dominate anything.
if (NewBBDominatesDestBB) {
if (!DestBBNode) DestBBNode = DT->getNode(DestBB);
DT->changeImmediateDominator(DestBBNode, NewBBNode);
}
}
}
// Should we update DominanceFrontier information?
if (DF) {
// If NewBBDominatesDestBB hasn't been computed yet, do so with DF.
if (!OtherPreds.empty()) {
// FIXME: IMPLEMENT THIS!
llvm_unreachable("Requiring domfrontiers but not idom/domtree/domset."
" not implemented yet!");
}
// Since the new block is dominated by its only predecessor TIBB,
// it cannot be in any block's dominance frontier. If NewBB dominates
// DestBB, its dominance frontier is the same as DestBB's, otherwise it is
// just {DestBB}.
DominanceFrontier::DomSetType NewDFSet;
if (NewBBDominatesDestBB) {
DominanceFrontier::iterator I = DF->find(DestBB);
if (I != DF->end()) {
DF->addBasicBlock(NewBB, I->second);
if (I->second.count(DestBB)) {
// However NewBB's frontier does not include DestBB.
DominanceFrontier::iterator NF = DF->find(NewBB);
DF->removeFromFrontier(NF, DestBB);
}
}
else
DF->addBasicBlock(NewBB, DominanceFrontier::DomSetType());
} else {
DominanceFrontier::DomSetType NewDFSet;
NewDFSet.insert(DestBB);
DF->addBasicBlock(NewBB, NewDFSet);
}
}
// Update LoopInfo if it is around.
if (LI) {
if (Loop *TIL = LI->getLoopFor(TIBB)) {
// If one or the other blocks were not in a loop, the new block is not
// either, and thus LI doesn't need to be updated.
if (Loop *DestLoop = LI->getLoopFor(DestBB)) {
if (TIL == DestLoop) {
// Both in the same loop, the NewBB joins loop.
DestLoop->addBasicBlockToLoop(NewBB, LI->getBase());
} else if (TIL->contains(DestLoop)) {
// Edge from an outer loop to an inner loop. Add to the outer loop.
TIL->addBasicBlockToLoop(NewBB, LI->getBase());
} else if (DestLoop->contains(TIL)) {
// Edge from an inner loop to an outer loop. Add to the outer loop.
DestLoop->addBasicBlockToLoop(NewBB, LI->getBase());
} else {
// Edge from two loops with no containment relation. Because these
// are natural loops, we know that the destination block must be the
// header of its loop (adding a branch into a loop elsewhere would
// create an irreducible loop).
assert(DestLoop->getHeader() == DestBB &&
"Should not create irreducible loops!");
if (Loop *P = DestLoop->getParentLoop())
P->addBasicBlockToLoop(NewBB, LI->getBase());
}
}
// If TIBB is in a loop and DestBB is outside of that loop, split the
// other exit blocks of the loop that also have predecessors outside
// the loop, to maintain a LoopSimplify guarantee.
if (!TIL->contains(DestBB) &&
P->mustPreserveAnalysisID(LoopSimplifyID)) {
assert(!TIL->contains(NewBB) &&
"Split point for loop exit is contained in loop!");
// Update LCSSA form in the newly created exit block.
if (P->mustPreserveAnalysisID(LCSSAID)) {
SmallVector<BasicBlock *, 1> OrigPred;
OrigPred.push_back(TIBB);
CreatePHIsForSplitLoopExit(OrigPred, NewBB, DestBB);
}
// For each unique exit block...
SmallVector<BasicBlock *, 4> ExitBlocks;
TIL->getExitBlocks(ExitBlocks);
for (unsigned i = 0, e = ExitBlocks.size(); i != e; ++i) {
// Collect all the preds that are inside the loop, and note
// whether there are any preds outside the loop.
SmallVector<BasicBlock *, 4> Preds;
bool HasPredOutsideOfLoop = false;
BasicBlock *Exit = ExitBlocks[i];
for (pred_iterator I = pred_begin(Exit), E = pred_end(Exit);
I != E; ++I) {
BasicBlock *P = *I;
if (TIL->contains(P))
Preds.push_back(P);
else
HasPredOutsideOfLoop = true;
}
// If there are any preds not in the loop, we'll need to split
// the edges. The Preds.empty() check is needed because a block
// may appear multiple times in the list. We can't use
// getUniqueExitBlocks above because that depends on LoopSimplify
// form, which we're in the process of restoring!
if (!Preds.empty() && HasPredOutsideOfLoop) {
BasicBlock *NewExitBB =
SplitBlockPredecessors(Exit, Preds.data(), Preds.size(),
"split", P);
if (P->mustPreserveAnalysisID(LCSSAID))
CreatePHIsForSplitLoopExit(Preds, NewExitBB, Exit);
}
}
}
// LCSSA form was updated above for the case where LoopSimplify is
// available, which means that all predecessors of loop exit blocks
// are within the loop. Without LoopSimplify form, it would be
// necessary to insert a new phi.
assert((!P->mustPreserveAnalysisID(LCSSAID) ||
P->mustPreserveAnalysisID(LoopSimplifyID)) &&
"SplitCriticalEdge doesn't know how to update LCCSA form "
"without LoopSimplify!");
}
}
// Update ProfileInfo if it is around.
if (PI)
PI->splitEdge(TIBB, DestBB, NewBB, MergeIdenticalEdges);
return NewBB;
}
