void LowerEmAsyncify::FindContextVariables(AsyncCallEntry & Entry) {
BasicBlock *AfterCallBlock = Entry.AfterCallBlock;
Function & F = *AfterCallBlock->getParent();
// Create a new entry block as if in the callback function
// theck check variables that no longer properly dominate their uses
BasicBlock *EntryBlock = BasicBlock::Create(TheModule->getContext(), "", &F, &F.getEntryBlock());
BranchInst::Create(AfterCallBlock, EntryBlock);
DominatorTreeWrapperPass DTW;
DTW.runOnFunction(F);
DominatorTree& DT = DTW.getDomTree();
// These blocks may be using some values defined at or before AsyncCallBlock
BasicBlockSet Ramifications = FindReachableBlocksFrom(AfterCallBlock);
SmallPtrSet<Value*, 256> ContextVariables;
Values Pending;
// Examine the instructions, find all variables that we need to store in the context
for (BasicBlockSet::iterator RI = Ramifications.begin(), RE = Ramifications.end(); RI != RE; ++RI) {
for (BasicBlock::iterator I = (*RI)->begin(), E = (*RI)->end(); I != E; ++I) {
for (unsigned i = 0, NumOperands = I->getNumOperands(); i < NumOperands; ++i) {
Value *O = I->getOperand(i);
if (Instruction *Inst = dyn_cast<Instruction>(O)) {
if (Inst == Entry.AsyncCallInst) continue; // for the original async call, we will load directly from async return value
if (ContextVariables.count(Inst) != 0) continue; // already examined
if (!DT.dominates(Inst, I->getOperandUse(i))) {
// `I` is using `Inst`, yet `Inst` does not dominate `I` if we arrive directly at AfterCallBlock
// so we need to save `Inst` in the context
ContextVariables.insert(Inst);
Pending.push_back(Inst);
}
} else if (Argument *Arg = dyn_cast<Argument>(O)) {
// count() should be as fast/slow as insert, so just insert here
ContextVariables.insert(Arg);
}
}
}
}
// restore F
EntryBlock->eraseFromParent();
Entry.ContextVariables.clear();
Entry.ContextVariables.reserve(ContextVariables.size());
for (SmallPtrSet<Value*, 256>::iterator I = ContextVariables.begin(), E = ContextVariables.end(); I != E; ++I) {
Entry.ContextVariables.push_back(*I);
}
}
/// Fold the loop tail into the loop exit by speculating the loop tail
/// instructions. Typically, this is a single post-increment. In the case of a
/// simple 2-block loop, hoisting the increment can be much better than
/// duplicating the entire loop header. In the case of loops with early exits,
/// rotation will not work anyway, but simplifyLoopLatch will put the loop in
/// canonical form so downstream passes can handle it.
///
/// I don't believe this invalidates SCEV.
bool LoopRotate::simplifyLoopLatch(Loop *L) {
BasicBlock *Latch = L->getLoopLatch();
if (!Latch || Latch->hasAddressTaken())
return false;
BranchInst *Jmp = dyn_cast<BranchInst>(Latch->getTerminator());
if (!Jmp || !Jmp->isUnconditional())
return false;
BasicBlock *LastExit = Latch->getSinglePredecessor();
if (!LastExit || !L->isLoopExiting(LastExit))
return false;
BranchInst *BI = dyn_cast<BranchInst>(LastExit->getTerminator());
if (!BI)
return false;
if (!shouldSpeculateInstrs(Latch->begin(), Jmp))
return false;
DEBUG(dbgs() << "Folding loop latch " << Latch->getName() << " into "
<< LastExit->getName() << "\n");
// Hoist the instructions from Latch into LastExit.
LastExit->getInstList().splice(BI, Latch->getInstList(), Latch->begin(), Jmp);
unsigned FallThruPath = BI->getSuccessor(0) == Latch ? 0 : 1;
BasicBlock *Header = Jmp->getSuccessor(0);
assert(Header == L->getHeader() && "expected a backward branch");
// Remove Latch from the CFG so that LastExit becomes the new Latch.
BI->setSuccessor(FallThruPath, Header);
Latch->replaceSuccessorsPhiUsesWith(LastExit);
Jmp->eraseFromParent();
// Nuke the Latch block.
assert(Latch->empty() && "unable to evacuate Latch");
LI->removeBlock(Latch);
if (DominatorTreeWrapperPass *DTWP =
getAnalysisIfAvailable<DominatorTreeWrapperPass>())
DTWP->getDomTree().eraseNode(Latch);
Latch->eraseFromParent();
return true;
}
/// MergeBasicBlockIntoOnlyPred - DestBB is a block with one predecessor and its
/// predecessor is known to have one successor (DestBB!). Eliminate the edge
/// between them, moving the instructions in the predecessor into DestBB and
/// deleting the predecessor block.
///
void llvm::MergeBasicBlockIntoOnlyPred(BasicBlock *DestBB) {
// If BB has single-entry PHI nodes, fold them.
while (PHINode *PN = dyn_cast<PHINode>(DestBB->begin())) {
Value *NewVal = PN->getIncomingValue(0);
// Replace self referencing PHI with undef, it must be dead.
if (NewVal == PN) NewVal = UndefValue::get(PN->getType());
PN->replaceAllUsesWith(NewVal);
PN->eraseFromParent();
}
BasicBlock *PredBB = DestBB->getSinglePredecessor();
assert(PredBB && "Block doesn't have a single predecessor!");
// Splice all the instructions from PredBB to DestBB.
PredBB->getTerminator()->eraseFromParent();
DestBB->getInstList().splice(DestBB->begin(), PredBB->getInstList());
// Anything that branched to PredBB now branches to DestBB.
PredBB->replaceAllUsesWith(DestBB);
// Nuke BB.
PredBB->eraseFromParent();
}
/// 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;
}
/// \brief Simplify one loop and queue further loops for simplification.
///
/// FIXME: Currently this accepts both lots of analyses that it uses and a raw
/// Pass pointer. The Pass pointer is used by numerous utilities to update
/// specific analyses. Rather than a pass it would be much cleaner and more
/// explicit if they accepted the analysis directly and then updated it.
static bool simplifyOneLoop(Loop *L, SmallVectorImpl<Loop *> &Worklist,
DominatorTree *DT, LoopInfo *LI,
ScalarEvolution *SE, Pass *PP,
AssumptionCache *AC) {
bool Changed = false;
ReprocessLoop:
// Check to see that no blocks (other than the header) in this loop have
// predecessors that are not in the loop. This is not valid for natural
// loops, but can occur if the blocks are unreachable. Since they are
// unreachable we can just shamelessly delete those CFG edges!
for (Loop::block_iterator BB = L->block_begin(), E = L->block_end();
BB != E; ++BB) {
if (*BB == L->getHeader()) continue;
SmallPtrSet<BasicBlock*, 4> BadPreds;
for (pred_iterator PI = pred_begin(*BB),
PE = pred_end(*BB); PI != PE; ++PI) {
BasicBlock *P = *PI;
if (!L->contains(P))
BadPreds.insert(P);
}
// Delete each unique out-of-loop (and thus dead) predecessor.
for (BasicBlock *P : BadPreds) {
DEBUG(dbgs() << "LoopSimplify: Deleting edge from dead predecessor "
<< P->getName() << "\n");
// Inform each successor of each dead pred.
for (succ_iterator SI = succ_begin(P), SE = succ_end(P); SI != SE; ++SI)
(*SI)->removePredecessor(P);
// Zap the dead pred's terminator and replace it with unreachable.
TerminatorInst *TI = P->getTerminator();
TI->replaceAllUsesWith(UndefValue::get(TI->getType()));
P->getTerminator()->eraseFromParent();
new UnreachableInst(P->getContext(), P);
Changed = true;
}
}
// If there are exiting blocks with branches on undef, resolve the undef in
// the direction which will exit the loop. This will help simplify loop
// trip count computations.
SmallVector<BasicBlock*, 8> ExitingBlocks;
L->getExitingBlocks(ExitingBlocks);
for (SmallVectorImpl<BasicBlock *>::iterator I = ExitingBlocks.begin(),
E = ExitingBlocks.end(); I != E; ++I)
if (BranchInst *BI = dyn_cast<BranchInst>((*I)->getTerminator()))
if (BI->isConditional()) {
if (UndefValue *Cond = dyn_cast<UndefValue>(BI->getCondition())) {
DEBUG(dbgs() << "LoopSimplify: Resolving \"br i1 undef\" to exit in "
<< (*I)->getName() << "\n");
BI->setCondition(ConstantInt::get(Cond->getType(),
!L->contains(BI->getSuccessor(0))));
// This may make the loop analyzable, force SCEV recomputation.
if (SE)
SE->forgetLoop(L);
Changed = true;
}
}
// Does the loop already have a preheader? If so, don't insert one.
BasicBlock *Preheader = L->getLoopPreheader();
if (!Preheader) {
Preheader = InsertPreheaderForLoop(L, PP);
if (Preheader) {
++NumInserted;
Changed = true;
}
}
// Next, check to make sure that all exit nodes of the loop only have
// predecessors that are inside of the loop. This check guarantees that the
// loop preheader/header will dominate the exit blocks. If the exit block has
// predecessors from outside of the loop, split the edge now.
SmallVector<BasicBlock*, 8> ExitBlocks;
L->getExitBlocks(ExitBlocks);
SmallSetVector<BasicBlock *, 8> ExitBlockSet(ExitBlocks.begin(),
ExitBlocks.end());
for (SmallSetVector<BasicBlock *, 8>::iterator I = ExitBlockSet.begin(),
E = ExitBlockSet.end(); I != E; ++I) {
BasicBlock *ExitBlock = *I;
for (pred_iterator PI = pred_begin(ExitBlock), PE = pred_end(ExitBlock);
PI != PE; ++PI)
// Must be exactly this loop: no subloops, parent loops, or non-loop preds
// allowed.
if (!L->contains(*PI)) {
if (rewriteLoopExitBlock(L, ExitBlock, DT, LI, PP)) {
++NumInserted;
Changed = true;
}
break;
}
}
// If the header has more than two predecessors at this point (from the
// preheader and from multiple backedges), we must adjust the loop.
BasicBlock *LoopLatch = L->getLoopLatch();
if (!LoopLatch) {
// If this is really a nested loop, rip it out into a child loop. Don't do
// this for loops with a giant number of backedges, just factor them into a
// common backedge instead.
if (L->getNumBackEdges() < 8) {
if (Loop *OuterL = separateNestedLoop(L, Preheader, DT, LI, SE, PP, AC)) {
++NumNested;
// Enqueue the outer loop as it should be processed next in our
// depth-first nest walk.
Worklist.push_back(OuterL);
// This is a big restructuring change, reprocess the whole loop.
Changed = true;
// GCC doesn't tail recursion eliminate this.
// FIXME: It isn't clear we can't rely on LLVM to TRE this.
goto ReprocessLoop;
}
}
// If we either couldn't, or didn't want to, identify nesting of the loops,
// insert a new block that all backedges target, then make it jump to the
// loop header.
LoopLatch = insertUniqueBackedgeBlock(L, Preheader, DT, LI);
if (LoopLatch) {
++NumInserted;
Changed = true;
}
}
const DataLayout &DL = L->getHeader()->getModule()->getDataLayout();
// Scan over the PHI nodes in the loop header. Since they now have only two
// incoming values (the loop is canonicalized), we may have simplified the PHI
// down to 'X = phi [X, Y]', which should be replaced with 'Y'.
PHINode *PN;
for (BasicBlock::iterator I = L->getHeader()->begin();
(PN = dyn_cast<PHINode>(I++)); )
if (Value *V = SimplifyInstruction(PN, DL, nullptr, DT, AC)) {
if (SE) SE->forgetValue(PN);
PN->replaceAllUsesWith(V);
PN->eraseFromParent();
}
// If this loop has multiple exits and the exits all go to the same
// block, attempt to merge the exits. This helps several passes, such
// as LoopRotation, which do not support loops with multiple exits.
// SimplifyCFG also does this (and this code uses the same utility
// function), however this code is loop-aware, where SimplifyCFG is
// not. That gives it the advantage of being able to hoist
// loop-invariant instructions out of the way to open up more
// opportunities, and the disadvantage of having the responsibility
// to preserve dominator information.
bool UniqueExit = true;
if (!ExitBlocks.empty())
for (unsigned i = 1, e = ExitBlocks.size(); i != e; ++i)
if (ExitBlocks[i] != ExitBlocks[0]) {
UniqueExit = false;
break;
}
if (UniqueExit) {
for (unsigned i = 0, e = ExitingBlocks.size(); i != e; ++i) {
BasicBlock *ExitingBlock = ExitingBlocks[i];
if (!ExitingBlock->getSinglePredecessor()) continue;
BranchInst *BI = dyn_cast<BranchInst>(ExitingBlock->getTerminator());
if (!BI || !BI->isConditional()) continue;
CmpInst *CI = dyn_cast<CmpInst>(BI->getCondition());
if (!CI || CI->getParent() != ExitingBlock) continue;
// Attempt to hoist out all instructions except for the
// comparison and the branch.
bool AllInvariant = true;
bool AnyInvariant = false;
for (BasicBlock::iterator I = ExitingBlock->begin(); &*I != BI; ) {
Instruction *Inst = I++;
// Skip debug info intrinsics.
if (isa<DbgInfoIntrinsic>(Inst))
continue;
if (Inst == CI)
continue;
if (!L->makeLoopInvariant(Inst, AnyInvariant,
Preheader ? Preheader->getTerminator()
: nullptr)) {
AllInvariant = false;
break;
}
}
if (AnyInvariant) {
Changed = true;
// The loop disposition of all SCEV expressions that depend on any
// hoisted values have also changed.
if (SE)
SE->forgetLoopDispositions(L);
}
if (!AllInvariant) continue;
// The block has now been cleared of all instructions except for
// a comparison and a conditional branch. SimplifyCFG may be able
// to fold it now.
if (!FoldBranchToCommonDest(BI))
continue;
// Success. The block is now dead, so remove it from the loop,
// update the dominator tree and delete it.
DEBUG(dbgs() << "LoopSimplify: Eliminating exiting block "
<< ExitingBlock->getName() << "\n");
// Notify ScalarEvolution before deleting this block. Currently assume the
// parent loop doesn't change (spliting edges doesn't count). If blocks,
// CFG edges, or other values in the parent loop change, then we need call
// to forgetLoop() for the parent instead.
if (SE)
SE->forgetLoop(L);
assert(pred_begin(ExitingBlock) == pred_end(ExitingBlock));
Changed = true;
LI->removeBlock(ExitingBlock);
DomTreeNode *Node = DT->getNode(ExitingBlock);
const std::vector<DomTreeNodeBase<BasicBlock> *> &Children =
Node->getChildren();
while (!Children.empty()) {
DomTreeNode *Child = Children.front();
DT->changeImmediateDominator(Child, Node->getIDom());
}
DT->eraseNode(ExitingBlock);
BI->getSuccessor(0)->removePredecessor(ExitingBlock);
BI->getSuccessor(1)->removePredecessor(ExitingBlock);
ExitingBlock->eraseFromParent();
}
}
return Changed;
}
/// 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,
bool ModuleLevelChanges,
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, ModuleLevelChanges,
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) {
Value *V = VMap[BI];
BasicBlock *NewBB = cast_or_null<BasicBlock>(V);
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();
DebugLoc TheCallDL;
if (TheCall)
TheCallDL = TheCall->getDebugLoc();
// 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)
PHIToResolve.push_back(cast<PHINode>(OldI));
}
// Otherwise, remap the rest of the instructions normally.
for (; I != NewBB->end(); ++I)
RemapInstruction(I, VMap,
ModuleLevelChanges ? RF_None : RF_NoModuleLevelChanges);
}
// 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;
}
}
// 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();
// 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.
}
}
/// removeBlocks - Remove basic block DeadBB and all blocks dominated by DeadBB.
/// This routine is used to remove split condition's dead branch, dominated by
/// DeadBB. LiveBB dominates split conidition's other branch.
void LoopIndexSplit::removeBlocks(BasicBlock *DeadBB, Loop *LP,
BasicBlock *LiveBB) {
// First update DeadBB's dominance frontier.
SmallVector<BasicBlock *, 8> FrontierBBs;
DominanceFrontier::iterator DeadBBDF = DF->find(DeadBB);
if (DeadBBDF != DF->end()) {
SmallVector<BasicBlock *, 8> PredBlocks;
DominanceFrontier::DomSetType DeadBBSet = DeadBBDF->second;
for (DominanceFrontier::DomSetType::iterator DeadBBSetI = DeadBBSet.begin(),
DeadBBSetE = DeadBBSet.end(); DeadBBSetI != DeadBBSetE; ++DeadBBSetI)
{
BasicBlock *FrontierBB = *DeadBBSetI;
FrontierBBs.push_back(FrontierBB);
// Rremove any PHI incoming edge from blocks dominated by DeadBB.
PredBlocks.clear();
for(pred_iterator PI = pred_begin(FrontierBB), PE = pred_end(FrontierBB);
PI != PE; ++PI) {
BasicBlock *P = *PI;
if (P == DeadBB || DT->dominates(DeadBB, P))
PredBlocks.push_back(P);
}
for(BasicBlock::iterator FBI = FrontierBB->begin(), FBE = FrontierBB->end();
FBI != FBE; ++FBI) {
if (PHINode *PN = dyn_cast<PHINode>(FBI)) {
for(SmallVector<BasicBlock *, 8>::iterator PI = PredBlocks.begin(),
PE = PredBlocks.end(); PI != PE; ++PI) {
BasicBlock *P = *PI;
PN->removeIncomingValue(P);
}
}
else
break;
}
}
}
// Now remove DeadBB and all nodes dominated by DeadBB in df order.
SmallVector<BasicBlock *, 32> WorkList;
DomTreeNode *DN = DT->getNode(DeadBB);
for (df_iterator<DomTreeNode*> DI = df_begin(DN),
E = df_end(DN); DI != E; ++DI) {
BasicBlock *BB = DI->getBlock();
WorkList.push_back(BB);
BB->replaceAllUsesWith(UndefValue::get(
Type::getLabelTy(DeadBB->getContext())));
}
while (!WorkList.empty()) {
BasicBlock *BB = WorkList.back(); WorkList.pop_back();
LPM->deleteSimpleAnalysisValue(BB, LP);
for(BasicBlock::iterator BBI = BB->begin(), BBE = BB->end();
BBI != BBE; ) {
Instruction *I = BBI;
++BBI;
I->replaceAllUsesWith(UndefValue::get(I->getType()));
LPM->deleteSimpleAnalysisValue(I, LP);
I->eraseFromParent();
}
DT->eraseNode(BB);
DF->removeBlock(BB);
LI->removeBlock(BB);
BB->eraseFromParent();
}
// Update Frontier BBs' dominator info.
while (!FrontierBBs.empty()) {
BasicBlock *FBB = FrontierBBs.back(); FrontierBBs.pop_back();
BasicBlock *NewDominator = FBB->getSinglePredecessor();
if (!NewDominator) {
pred_iterator PI = pred_begin(FBB), PE = pred_end(FBB);
NewDominator = *PI;
++PI;
if (NewDominator != LiveBB) {
for(; PI != PE; ++PI) {
BasicBlock *P = *PI;
if (P == LiveBB) {
NewDominator = LiveBB;
break;
}
NewDominator = DT->findNearestCommonDominator(NewDominator, P);
}
}
}
assert (NewDominator && "Unable to fix dominator info.");
DT->changeImmediateDominator(FBB, NewDominator);
DF->changeImmediateDominator(FBB, NewDominator, DT);
}
}
/// SimplifyCode - Okay, now that we have simplified some instructions in the
/// loop, walk over it and constant prop, dce, and fold control flow where
/// possible. Note that this is effectively a very simple loop-structure-aware
/// optimizer. During processing of this loop, L could very well be deleted, so
/// it must not be used.
///
/// FIXME: When the loop optimizer is more mature, separate this out to a new
/// pass.
///
void LoopUnswitch::SimplifyCode(std::vector<Instruction*> &Worklist, Loop *L) {
while (!Worklist.empty()) {
Instruction *I = Worklist.back();
Worklist.pop_back();
// Simple DCE.
if (isInstructionTriviallyDead(I)) {
DEBUG(dbgs() << "Remove dead instruction '" << *I);
// Add uses to the worklist, which may be dead now.
for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i)
if (Instruction *Use = dyn_cast<Instruction>(I->getOperand(i)))
Worklist.push_back(Use);
LPM->deleteSimpleAnalysisValue(I, L);
RemoveFromWorklist(I, Worklist);
I->eraseFromParent();
++NumSimplify;
continue;
}
// See if instruction simplification can hack this up. This is common for
// things like "select false, X, Y" after unswitching made the condition be
// 'false'.
if (Value *V = SimplifyInstruction(I, 0, DT))
if (LI->replacementPreservesLCSSAForm(I, V)) {
ReplaceUsesOfWith(I, V, Worklist, L, LPM);
continue;
}
// Special case hacks that appear commonly in unswitched code.
if (BranchInst *BI = dyn_cast<BranchInst>(I)) {
if (BI->isUnconditional()) {
// If BI's parent is the only pred of the successor, fold the two blocks
// together.
BasicBlock *Pred = BI->getParent();
BasicBlock *Succ = BI->getSuccessor(0);
BasicBlock *SinglePred = Succ->getSinglePredecessor();
if (!SinglePred) continue; // Nothing to do.
assert(SinglePred == Pred && "CFG broken");
DEBUG(dbgs() << "Merging blocks: " << Pred->getName() << " <- "
<< Succ->getName() << "\n");
// Resolve any single entry PHI nodes in Succ.
while (PHINode *PN = dyn_cast<PHINode>(Succ->begin()))
ReplaceUsesOfWith(PN, PN->getIncomingValue(0), Worklist, L, LPM);
// Move all of the successor contents from Succ to Pred.
Pred->getInstList().splice(BI, Succ->getInstList(), Succ->begin(),
Succ->end());
LPM->deleteSimpleAnalysisValue(BI, L);
BI->eraseFromParent();
RemoveFromWorklist(BI, Worklist);
// If Succ has any successors with PHI nodes, update them to have
// entries coming from Pred instead of Succ.
Succ->replaceAllUsesWith(Pred);
// Remove Succ from the loop tree.
LI->removeBlock(Succ);
LPM->deleteSimpleAnalysisValue(Succ, L);
Succ->eraseFromParent();
++NumSimplify;
continue;
}
if (ConstantInt *CB = dyn_cast<ConstantInt>(BI->getCondition())){
// Conditional branch. Turn it into an unconditional branch, then
// remove dead blocks.
continue; // FIXME: Enable.
DEBUG(dbgs() << "Folded branch: " << *BI);
BasicBlock *DeadSucc = BI->getSuccessor(CB->getZExtValue());
BasicBlock *LiveSucc = BI->getSuccessor(!CB->getZExtValue());
DeadSucc->removePredecessor(BI->getParent(), true);
Worklist.push_back(BranchInst::Create(LiveSucc, BI));
LPM->deleteSimpleAnalysisValue(BI, L);
BI->eraseFromParent();
RemoveFromWorklist(BI, Worklist);
++NumSimplify;
RemoveBlockIfDead(DeadSucc, Worklist, L);
}
continue;
}
}
}
// InlineFunction - This function inlines the called function into the basic
// block of the caller. This returns false if it is not possible to inline this
// call. The program is still in a well defined state if this occurs though.
//
// Note that this only does one level of inlining. For example, if the
// instruction 'call B' is inlined, and 'B' calls 'C', then the call to 'C' now
// exists in the instruction stream. Similiarly this will inline a recursive
// function by one level.
//
bool llvm::InlineFunction(CallSite CS, CallGraph *CG, const TargetData *TD) {
Instruction *TheCall = CS.getInstruction();
assert(TheCall->getParent() && TheCall->getParent()->getParent() &&
"Instruction not in function!");
const Function *CalledFunc = CS.getCalledFunction();
if (CalledFunc == 0 || // Can't inline external function or indirect
CalledFunc->isDeclaration() || // call, or call to a vararg function!
CalledFunc->getFunctionType()->isVarArg()) return false;
// If the call to the callee is not a tail call, we must clear the 'tail'
// flags on any calls that we inline.
bool MustClearTailCallFlags =
!(isa<CallInst>(TheCall) && cast<CallInst>(TheCall)->isTailCall());
// If the call to the callee cannot throw, set the 'nounwind' flag on any
// calls that we inline.
bool MarkNoUnwind = CS.doesNotThrow();
BasicBlock *OrigBB = TheCall->getParent();
Function *Caller = OrigBB->getParent();
// GC poses two hazards to inlining, which only occur when the callee has GC:
// 1. If the caller has no GC, then the callee's GC must be propagated to the
// caller.
// 2. If the caller has a differing GC, it is invalid to inline.
if (CalledFunc->hasGC()) {
if (!Caller->hasGC())
Caller->setGC(CalledFunc->getGC());
else if (CalledFunc->getGC() != Caller->getGC())
return false;
}
// Get an iterator to the last basic block in the function, which will have
// the new function inlined after it.
//
Function::iterator LastBlock = &Caller->back();
// Make sure to capture all of the return instructions from the cloned
// function.
std::vector<ReturnInst*> Returns;
ClonedCodeInfo InlinedFunctionInfo;
Function::iterator FirstNewBlock;
{ // Scope to destroy ValueMap after cloning.
DenseMap<const Value*, Value*> ValueMap;
assert(CalledFunc->arg_size() == CS.arg_size() &&
"No varargs calls can be inlined!");
// Calculate the vector of arguments to pass into the function cloner, which
// matches up the formal to the actual argument values.
CallSite::arg_iterator AI = CS.arg_begin();
unsigned ArgNo = 0;
for (Function::const_arg_iterator I = CalledFunc->arg_begin(),
E = CalledFunc->arg_end(); I != E; ++I, ++AI, ++ArgNo) {
Value *ActualArg = *AI;
// When byval arguments actually inlined, we need to make the copy implied
// by them explicit. However, we don't do this if the callee is readonly
// or readnone, because the copy would be unneeded: the callee doesn't
// modify the struct.
if (CalledFunc->paramHasAttr(ArgNo+1, Attribute::ByVal) &&
!CalledFunc->onlyReadsMemory()) {
const Type *AggTy = cast<PointerType>(I->getType())->getElementType();
const Type *VoidPtrTy = PointerType::getUnqual(Type::Int8Ty);
// Create the alloca. If we have TargetData, use nice alignment.
unsigned Align = 1;
if (TD) Align = TD->getPrefTypeAlignment(AggTy);
Value *NewAlloca = new AllocaInst(AggTy, 0, Align, I->getName(),
Caller->begin()->begin());
// Emit a memcpy.
const Type *Tys[] = { Type::Int64Ty };
Function *MemCpyFn = Intrinsic::getDeclaration(Caller->getParent(),
Intrinsic::memcpy,
Tys, 1);
Value *DestCast = new BitCastInst(NewAlloca, VoidPtrTy, "tmp", TheCall);
Value *SrcCast = new BitCastInst(*AI, VoidPtrTy, "tmp", TheCall);
Value *Size;
if (TD == 0)
Size = ConstantExpr::getSizeOf(AggTy);
else
Size = ConstantInt::get(Type::Int64Ty, TD->getTypeStoreSize(AggTy));
// Always generate a memcpy of alignment 1 here because we don't know
// the alignment of the src pointer. Other optimizations can infer
// better alignment.
Value *CallArgs[] = {
DestCast, SrcCast, Size, ConstantInt::get(Type::Int32Ty, 1)
};
CallInst *TheMemCpy =
CallInst::Create(MemCpyFn, CallArgs, CallArgs+4, "", TheCall);
// If we have a call graph, update it.
if (CG) {
CallGraphNode *MemCpyCGN = CG->getOrInsertFunction(MemCpyFn);
CallGraphNode *CallerNode = (*CG)[Caller];
CallerNode->addCalledFunction(TheMemCpy, MemCpyCGN);
}
// Uses of the argument in the function should use our new alloca
// instead.
ActualArg = NewAlloca;
}
ValueMap[I] = ActualArg;
}
// We want the inliner to prune the code as it copies. We would LOVE to
// have no dead or constant instructions leftover after inlining occurs
// (which can happen, e.g., because an argument was constant), but we'll be
// happy with whatever the cloner can do.
CloneAndPruneFunctionInto(Caller, CalledFunc, ValueMap, Returns, ".i",
&InlinedFunctionInfo, TD);
// Remember the first block that is newly cloned over.
FirstNewBlock = LastBlock; ++FirstNewBlock;
// Update the callgraph if requested.
if (CG)
UpdateCallGraphAfterInlining(CS, FirstNewBlock, ValueMap, *CG);
}
// If there are any alloca instructions in the block that used to be the entry
// block for the callee, move them to the entry block of the caller. First
// calculate which instruction they should be inserted before. We insert the
// instructions at the end of the current alloca list.
//
{
BasicBlock::iterator InsertPoint = Caller->begin()->begin();
for (BasicBlock::iterator I = FirstNewBlock->begin(),
E = FirstNewBlock->end(); I != E; )
if (AllocaInst *AI = dyn_cast<AllocaInst>(I++)) {
// If the alloca is now dead, remove it. This often occurs due to code
// specialization.
if (AI->use_empty()) {
AI->eraseFromParent();
continue;
}
if (isa<Constant>(AI->getArraySize())) {
// Scan for the block of allocas that we can move over, and move them
// all at once.
while (isa<AllocaInst>(I) &&
isa<Constant>(cast<AllocaInst>(I)->getArraySize()))
++I;
// Transfer all of the allocas over in a block. Using splice means
// that the instructions aren't removed from the symbol table, then
// reinserted.
Caller->getEntryBlock().getInstList().splice(
InsertPoint,
FirstNewBlock->getInstList(),
AI, I);
}
}
}
// If the inlined code contained dynamic alloca instructions, wrap the inlined
// code with llvm.stacksave/llvm.stackrestore intrinsics.
if (InlinedFunctionInfo.ContainsDynamicAllocas) {
Module *M = Caller->getParent();
// Get the two intrinsics we care about.
Constant *StackSave, *StackRestore;
StackSave = Intrinsic::getDeclaration(M, Intrinsic::stacksave);
StackRestore = Intrinsic::getDeclaration(M, Intrinsic::stackrestore);
// If we are preserving the callgraph, add edges to the stacksave/restore
// functions for the calls we insert.
CallGraphNode *StackSaveCGN = 0, *StackRestoreCGN = 0, *CallerNode = 0;
if (CG) {
// We know that StackSave/StackRestore are Function*'s, because they are
// intrinsics which must have the right types.
StackSaveCGN = CG->getOrInsertFunction(cast<Function>(StackSave));
StackRestoreCGN = CG->getOrInsertFunction(cast<Function>(StackRestore));
CallerNode = (*CG)[Caller];
}
// Insert the llvm.stacksave.
CallInst *SavedPtr = CallInst::Create(StackSave, "savedstack",
FirstNewBlock->begin());
if (CG) CallerNode->addCalledFunction(SavedPtr, StackSaveCGN);
// Insert a call to llvm.stackrestore before any return instructions in the
// inlined function.
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
CallInst *CI = CallInst::Create(StackRestore, SavedPtr, "", Returns[i]);
if (CG) CallerNode->addCalledFunction(CI, StackRestoreCGN);
}
// Count the number of StackRestore calls we insert.
unsigned NumStackRestores = Returns.size();
// If we are inlining an invoke instruction, insert restores before each
// unwind. These unwinds will be rewritten into branches later.
if (InlinedFunctionInfo.ContainsUnwinds && isa<InvokeInst>(TheCall)) {
for (Function::iterator BB = FirstNewBlock, E = Caller->end();
BB != E; ++BB)
if (UnwindInst *UI = dyn_cast<UnwindInst>(BB->getTerminator())) {
CallInst::Create(StackRestore, SavedPtr, "", UI);
++NumStackRestores;
}
}
}
// If we are inlining tail call instruction through a call site that isn't
// marked 'tail', we must remove the tail marker for any calls in the inlined
// code. Also, calls inlined through a 'nounwind' call site should be marked
// 'nounwind'.
if (InlinedFunctionInfo.ContainsCalls &&
(MustClearTailCallFlags || MarkNoUnwind)) {
for (Function::iterator BB = FirstNewBlock, E = Caller->end();
BB != E; ++BB)
for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
if (CallInst *CI = dyn_cast<CallInst>(I)) {
if (MustClearTailCallFlags)
CI->setTailCall(false);
if (MarkNoUnwind)
CI->setDoesNotThrow();
}
}
// If we are inlining through a 'nounwind' call site then any inlined 'unwind'
// instructions are unreachable.
if (InlinedFunctionInfo.ContainsUnwinds && MarkNoUnwind)
for (Function::iterator BB = FirstNewBlock, E = Caller->end();
BB != E; ++BB) {
TerminatorInst *Term = BB->getTerminator();
if (isa<UnwindInst>(Term)) {
new UnreachableInst(Term);
BB->getInstList().erase(Term);
}
}
// If we are inlining for an invoke instruction, we must make sure to rewrite
// any inlined 'unwind' instructions into branches to the invoke exception
// destination, and call instructions into invoke instructions.
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall))
HandleInlinedInvoke(II, FirstNewBlock, InlinedFunctionInfo);
// If we cloned in _exactly one_ basic block, and if that block ends in a
// return instruction, we splice the body of the inlined callee directly into
// the calling basic block.
if (Returns.size() == 1 && std::distance(FirstNewBlock, Caller->end()) == 1) {
// Move all of the instructions right before the call.
OrigBB->getInstList().splice(TheCall, FirstNewBlock->getInstList(),
FirstNewBlock->begin(), FirstNewBlock->end());
// Remove the cloned basic block.
Caller->getBasicBlockList().pop_back();
// If the call site was an invoke instruction, add a branch to the normal
// destination.
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall))
BranchInst::Create(II->getNormalDest(), TheCall);
// If the return instruction returned a value, replace uses of the call with
// uses of the returned value.
if (!TheCall->use_empty()) {
ReturnInst *R = Returns[0];
TheCall->replaceAllUsesWith(R->getReturnValue());
}
// Since we are now done with the Call/Invoke, we can delete it.
TheCall->eraseFromParent();
// Since we are now done with the return instruction, delete it also.
Returns[0]->eraseFromParent();
// We are now done with the inlining.
return true;
}
// Otherwise, we have the normal case, of more than one block to inline or
// multiple return sites.
// We want to clone the entire callee function into the hole between the
// "starter" and "ender" blocks. How we accomplish this depends on whether
// this is an invoke instruction or a call instruction.
BasicBlock *AfterCallBB;
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall)) {
// Add an unconditional branch to make this look like the CallInst case...
BranchInst *NewBr = BranchInst::Create(II->getNormalDest(), TheCall);
// Split the basic block. This guarantees that no PHI nodes will have to be
// updated due to new incoming edges, and make the invoke case more
// symmetric to the call case.
AfterCallBB = OrigBB->splitBasicBlock(NewBr,
CalledFunc->getName()+".exit");
} else { // It's a call
// If this is a call instruction, we need to split the basic block that
// the call lives in.
//
AfterCallBB = OrigBB->splitBasicBlock(TheCall,
CalledFunc->getName()+".exit");
}
// Change the branch that used to go to AfterCallBB to branch to the first
// basic block of the inlined function.
//
TerminatorInst *Br = OrigBB->getTerminator();
assert(Br && Br->getOpcode() == Instruction::Br &&
"splitBasicBlock broken!");
Br->setOperand(0, FirstNewBlock);
// Now that the function is correct, make it a little bit nicer. In
// particular, move the basic blocks inserted from the end of the function
// into the space made by splitting the source basic block.
Caller->getBasicBlockList().splice(AfterCallBB, Caller->getBasicBlockList(),
FirstNewBlock, Caller->end());
// Handle all of the return instructions that we just cloned in, and eliminate
// any users of the original call/invoke instruction.
const Type *RTy = CalledFunc->getReturnType();
if (Returns.size() > 1) {
// The PHI node should go at the front of the new basic block to merge all
// possible incoming values.
PHINode *PHI = 0;
if (!TheCall->use_empty()) {
PHI = PHINode::Create(RTy, TheCall->getName(),
AfterCallBB->begin());
// Anything that used the result of the function call should now use the
// PHI node as their operand.
TheCall->replaceAllUsesWith(PHI);
}
// Loop over all of the return instructions adding entries to the PHI node
// as appropriate.
if (PHI) {
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
ReturnInst *RI = Returns[i];
assert(RI->getReturnValue()->getType() == PHI->getType() &&
"Ret value not consistent in function!");
PHI->addIncoming(RI->getReturnValue(), RI->getParent());
}
}
// Add a branch to the merge points and remove return instructions.
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
ReturnInst *RI = Returns[i];
BranchInst::Create(AfterCallBB, RI);
RI->eraseFromParent();
}
} else if (!Returns.empty()) {
// Otherwise, if there is exactly one return value, just replace anything
// using the return value of the call with the computed value.
if (!TheCall->use_empty())
TheCall->replaceAllUsesWith(Returns[0]->getReturnValue());
// Splice the code from the return block into the block that it will return
// to, which contains the code that was after the call.
BasicBlock *ReturnBB = Returns[0]->getParent();
AfterCallBB->getInstList().splice(AfterCallBB->begin(),
ReturnBB->getInstList());
// Update PHI nodes that use the ReturnBB to use the AfterCallBB.
ReturnBB->replaceAllUsesWith(AfterCallBB);
// Delete the return instruction now and empty ReturnBB now.
Returns[0]->eraseFromParent();
ReturnBB->eraseFromParent();
} else if (!TheCall->use_empty()) {
// No returns, but something is using the return value of the call. Just
// nuke the result.
TheCall->replaceAllUsesWith(UndefValue::get(TheCall->getType()));
}
// Since we are now done with the Call/Invoke, we can delete it.
TheCall->eraseFromParent();
// We should always be able to fold the entry block of the function into the
// single predecessor of the block...
assert(cast<BranchInst>(Br)->isUnconditional() && "splitBasicBlock broken!");
BasicBlock *CalleeEntry = cast<BranchInst>(Br)->getSuccessor(0);
// Splice the code entry block into calling block, right before the
// unconditional branch.
OrigBB->getInstList().splice(Br, CalleeEntry->getInstList());
CalleeEntry->replaceAllUsesWith(OrigBB); // Update PHI nodes
// Remove the unconditional branch.
OrigBB->getInstList().erase(Br);
// Now we can remove the CalleeEntry block, which is now empty.
Caller->getBasicBlockList().erase(CalleeEntry);
return true;
}
/// For each (predecessor, conditions from predecessors) pair, it will split the
/// basic block containing the call site, hook it up to the predecessor and
/// replace the call instruction with new call instructions, which contain
/// constraints based on the conditions from their predecessors.
/// For example, in the IR below with an OR condition, the call-site can
/// be split. In this case, Preds for Tail is [(Header, a == null),
/// (TBB, a != null, b == null)]. Tail is replaced by 2 split blocks, containing
/// CallInst1, which has constraints based on the conditions from Head and
/// CallInst2, which has constraints based on the conditions coming from TBB.
///
/// From :
///
/// Header:
/// %c = icmp eq i32* %a, null
/// br i1 %c %Tail, %TBB
/// TBB:
/// %c2 = icmp eq i32* %b, null
/// br i1 %c %Tail, %End
/// Tail:
/// %ca = call i1 @callee (i32* %a, i32* %b)
///
/// to :
///
/// Header: // PredBB1 is Header
/// %c = icmp eq i32* %a, null
/// br i1 %c %Tail-split1, %TBB
/// TBB: // PredBB2 is TBB
/// %c2 = icmp eq i32* %b, null
/// br i1 %c %Tail-split2, %End
/// Tail-split1:
/// %ca1 = call @callee (i32* null, i32* %b) // CallInst1
/// br %Tail
/// Tail-split2:
/// %ca2 = call @callee (i32* nonnull %a, i32* null) // CallInst2
/// br %Tail
/// Tail:
/// %p = phi i1 [%ca1, %Tail-split1],[%ca2, %Tail-split2]
///
/// Note that in case any arguments at the call-site are constrained by its
/// predecessors, new call-sites with more constrained arguments will be
/// created in createCallSitesOnPredicatedArgument().
static void splitCallSite(
CallSite CS,
const SmallVectorImpl<std::pair<BasicBlock *, ConditionsTy>> &Preds) {
Instruction *Instr = CS.getInstruction();
BasicBlock *TailBB = Instr->getParent();
bool IsMustTailCall = CS.isMustTailCall();
PHINode *CallPN = nullptr;
// `musttail` calls must be followed by optional `bitcast`, and `ret`. The
// split blocks will be terminated right after that so there're no users for
// this phi in a `TailBB`.
if (!IsMustTailCall && Instr->getNumUses())
CallPN = PHINode::Create(Instr->getType(), Preds.size(), "phi.call");
DEBUG(dbgs() << "split call-site : " << *Instr << " into \n");
assert(Preds.size() == 2 && "The ValueToValueMaps array has size 2.");
// ValueToValueMapTy is neither copy nor moveable, so we use a simple array
// here.
ValueToValueMapTy ValueToValueMaps[2];
for (unsigned i = 0; i < Preds.size(); i++) {
BasicBlock *PredBB = Preds[i].first;
BasicBlock *SplitBlock = DuplicateInstructionsInSplitBetween(
TailBB, PredBB, &*std::next(Instr->getIterator()), ValueToValueMaps[i]);
assert(SplitBlock && "Unexpected new basic block split.");
Instruction *NewCI =
&*std::prev(SplitBlock->getTerminator()->getIterator());
CallSite NewCS(NewCI);
addConditions(NewCS, Preds[i].second);
// Handle PHIs used as arguments in the call-site.
for (PHINode &PN : TailBB->phis()) {
unsigned ArgNo = 0;
for (auto &CI : CS.args()) {
if (&*CI == &PN) {
NewCS.setArgument(ArgNo, PN.getIncomingValueForBlock(SplitBlock));
}
++ArgNo;
}
}
DEBUG(dbgs() << " " << *NewCI << " in " << SplitBlock->getName()
<< "\n");
if (CallPN)
CallPN->addIncoming(NewCI, SplitBlock);
// Clone and place bitcast and return instructions before `TI`
if (IsMustTailCall)
copyMustTailReturn(SplitBlock, Instr, NewCI);
}
NumCallSiteSplit++;
// FIXME: remove TI in `copyMustTailReturn`
if (IsMustTailCall) {
// Remove superfluous `br` terminators from the end of the Split blocks
// NOTE: Removing terminator removes the SplitBlock from the TailBB's
// predecessors. Therefore we must get complete list of Splits before
// attempting removal.
SmallVector<BasicBlock *, 2> Splits(predecessors((TailBB)));
assert(Splits.size() == 2 && "Expected exactly 2 splits!");
for (unsigned i = 0; i < Splits.size(); i++)
Splits[i]->getTerminator()->eraseFromParent();
// Erase the tail block once done with musttail patching
TailBB->eraseFromParent();
return;
}
auto *OriginalBegin = &*TailBB->begin();
// Replace users of the original call with a PHI mering call-sites split.
if (CallPN) {
CallPN->insertBefore(OriginalBegin);
Instr->replaceAllUsesWith(CallPN);
}
// Remove instructions moved to split blocks from TailBB, from the duplicated
// call instruction to the beginning of the basic block. If an instruction
// has any uses, add a new PHI node to combine the values coming from the
// split blocks. The new PHI nodes are placed before the first original
// instruction, so we do not end up deleting them. By using reverse-order, we
// do not introduce unnecessary PHI nodes for def-use chains from the call
// instruction to the beginning of the block.
auto I = Instr->getReverseIterator();
while (I != TailBB->rend()) {
Instruction *CurrentI = &*I++;
if (!CurrentI->use_empty()) {
// If an existing PHI has users after the call, there is no need to create
// a new one.
if (isa<PHINode>(CurrentI))
continue;
PHINode *NewPN = PHINode::Create(CurrentI->getType(), Preds.size());
for (auto &Mapping : ValueToValueMaps)
NewPN->addIncoming(Mapping[CurrentI],
cast<Instruction>(Mapping[CurrentI])->getParent());
NewPN->insertBefore(&*TailBB->begin());
CurrentI->replaceAllUsesWith(NewPN);
}
CurrentI->eraseFromParent();
// We are done once we handled the first original instruction in TailBB.
if (CurrentI == OriginalBegin)
break;
}
}
/// 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) {
assert(NameSuffix && "NameSuffix cannot be null!");
ValueMapTypeRemapper *TypeMapper = nullptr;
ValueMaterializer *Materializer = nullptr;
#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);
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);
}
/// Check whether \param BB is the merge block of a if-region. If yes, check
/// whether there exists an adjacent if-region upstream, the two if-regions
/// contain identical instuctions and can be legally merged. \returns true if
/// the two if-regions are merged.
///
/// From:
/// if (a)
/// statement;
/// if (b)
/// statement;
///
/// To:
/// if (a || b)
/// statement;
///
bool FlattenCFGOpt::MergeIfRegion(BasicBlock *BB, IRBuilder<> &Builder,
Pass *P) {
BasicBlock *IfTrue2, *IfFalse2;
Value *IfCond2 = GetIfCondition(BB, IfTrue2, IfFalse2);
Instruction *CInst2 = dyn_cast_or_null<Instruction>(IfCond2);
if (!CInst2)
return false;
BasicBlock *SecondEntryBlock = CInst2->getParent();
if (SecondEntryBlock->hasAddressTaken())
return false;
BasicBlock *IfTrue1, *IfFalse1;
Value *IfCond1 = GetIfCondition(SecondEntryBlock, IfTrue1, IfFalse1);
Instruction *CInst1 = dyn_cast_or_null<Instruction>(IfCond1);
if (!CInst1)
return false;
BasicBlock *FirstEntryBlock = CInst1->getParent();
// Either then-path or else-path should be empty.
if ((IfTrue1 != FirstEntryBlock) && (IfFalse1 != FirstEntryBlock))
return false;
if ((IfTrue2 != SecondEntryBlock) && (IfFalse2 != SecondEntryBlock))
return false;
TerminatorInst *PTI2 = SecondEntryBlock->getTerminator();
Instruction *PBI2 = SecondEntryBlock->begin();
if (!CompareIfRegionBlock(FirstEntryBlock, SecondEntryBlock, IfTrue1,
IfTrue2))
return false;
if (!CompareIfRegionBlock(FirstEntryBlock, SecondEntryBlock, IfFalse1,
IfFalse2))
return false;
// Check whether \param SecondEntryBlock has side-effect and is safe to
// speculate.
for (BasicBlock::iterator BI = PBI2, BE = PTI2; BI != BE; ++BI) {
Instruction *CI = BI;
if (isa<PHINode>(CI) || CI->mayHaveSideEffects() ||
!isSafeToSpeculativelyExecute(CI))
return false;
}
// Merge \param SecondEntryBlock into \param FirstEntryBlock.
FirstEntryBlock->getInstList().pop_back();
FirstEntryBlock->getInstList()
.splice(FirstEntryBlock->end(), SecondEntryBlock->getInstList());
BranchInst *PBI = dyn_cast<BranchInst>(FirstEntryBlock->getTerminator());
Value *CC = PBI->getCondition();
BasicBlock *SaveInsertBB = Builder.GetInsertBlock();
BasicBlock::iterator SaveInsertPt = Builder.GetInsertPoint();
Builder.SetInsertPoint(PBI);
Value *NC = Builder.CreateOr(CInst1, CC);
PBI->replaceUsesOfWith(CC, NC);
Builder.SetInsertPoint(SaveInsertBB, SaveInsertPt);
// Remove IfTrue1
if (IfTrue1 != FirstEntryBlock) {
IfTrue1->dropAllReferences();
IfTrue1->eraseFromParent();
}
// Remove IfFalse1
if (IfFalse1 != FirstEntryBlock) {
IfFalse1->dropAllReferences();
IfFalse1->eraseFromParent();
}
// Remove \param SecondEntryBlock
SecondEntryBlock->dropAllReferences();
SecondEntryBlock->eraseFromParent();
DEBUG(dbgs() << "If conditions merged into:\n" << *FirstEntryBlock);
return true;
}
/// InlineFunction - This function inlines the called function into the basic
/// block of the caller. This returns false if it is not possible to inline
/// this call. The program is still in a well defined state if this occurs
/// though.
///
/// Note that this only does one level of inlining. For example, if the
/// instruction 'call B' is inlined, and 'B' calls 'C', then the call to 'C' now
/// exists in the instruction stream. Similarly this will inline a recursive
/// function by one level.
bool llvm::InlineFunction(CallSite CS, InlineFunctionInfo &IFI,
bool InsertLifetime) {
Instruction *TheCall = CS.getInstruction();
assert(TheCall->getParent() && TheCall->getParent()->getParent() &&
"Instruction not in function!");
// If IFI has any state in it, zap it before we fill it in.
IFI.reset();
const Function *CalledFunc = CS.getCalledFunction();
if (CalledFunc == 0 || // Can't inline external function or indirect
CalledFunc->isDeclaration() || // call, or call to a vararg function!
CalledFunc->getFunctionType()->isVarArg()) return false;
// If the call to the callee is not a tail call, we must clear the 'tail'
// flags on any calls that we inline.
bool MustClearTailCallFlags =
!(isa<CallInst>(TheCall) && cast<CallInst>(TheCall)->isTailCall());
// If the call to the callee cannot throw, set the 'nounwind' flag on any
// calls that we inline.
bool MarkNoUnwind = CS.doesNotThrow();
BasicBlock *OrigBB = TheCall->getParent();
Function *Caller = OrigBB->getParent();
// GC poses two hazards to inlining, which only occur when the callee has GC:
// 1. If the caller has no GC, then the callee's GC must be propagated to the
// caller.
// 2. If the caller has a differing GC, it is invalid to inline.
if (CalledFunc->hasGC()) {
if (!Caller->hasGC())
Caller->setGC(CalledFunc->getGC());
else if (CalledFunc->getGC() != Caller->getGC())
return false;
}
// Get the personality function from the callee if it contains a landing pad.
Value *CalleePersonality = 0;
for (Function::const_iterator I = CalledFunc->begin(), E = CalledFunc->end();
I != E; ++I)
if (const InvokeInst *II = dyn_cast<InvokeInst>(I->getTerminator())) {
const BasicBlock *BB = II->getUnwindDest();
const LandingPadInst *LP = BB->getLandingPadInst();
CalleePersonality = LP->getPersonalityFn();
break;
}
// Find the personality function used by the landing pads of the caller. If it
// exists, then check to see that it matches the personality function used in
// the callee.
if (CalleePersonality) {
for (Function::const_iterator I = Caller->begin(), E = Caller->end();
I != E; ++I)
if (const InvokeInst *II = dyn_cast<InvokeInst>(I->getTerminator())) {
const BasicBlock *BB = II->getUnwindDest();
const LandingPadInst *LP = BB->getLandingPadInst();
// If the personality functions match, then we can perform the
// inlining. Otherwise, we can't inline.
// TODO: This isn't 100% true. Some personality functions are proper
// supersets of others and can be used in place of the other.
if (LP->getPersonalityFn() != CalleePersonality)
return false;
break;
}
}
// Get an iterator to the last basic block in the function, which will have
// the new function inlined after it.
Function::iterator LastBlock = &Caller->back();
// Make sure to capture all of the return instructions from the cloned
// function.
SmallVector<ReturnInst*, 8> Returns;
ClonedCodeInfo InlinedFunctionInfo;
Function::iterator FirstNewBlock;
{ // Scope to destroy VMap after cloning.
ValueToValueMapTy VMap;
assert(CalledFunc->arg_size() == CS.arg_size() &&
"No varargs calls can be inlined!");
// Calculate the vector of arguments to pass into the function cloner, which
// matches up the formal to the actual argument values.
CallSite::arg_iterator AI = CS.arg_begin();
unsigned ArgNo = 0;
for (Function::const_arg_iterator I = CalledFunc->arg_begin(),
E = CalledFunc->arg_end(); I != E; ++I, ++AI, ++ArgNo) {
Value *ActualArg = *AI;
const Argument *Arg = I;
// When byval arguments actually inlined, we need to make the copy implied
// by them explicit. However, we don't do this if the callee is readonly
// or readnone, because the copy would be unneeded: the callee doesn't
// modify the struct.
if (CS.isByValArgument(ArgNo)) {
ActualArg = HandleByValArgument(ActualArg, Arg, TheCall, CalledFunc, IFI,
CalledFunc->getParamAlignment(ArgNo+1));
// Calls that we inline may use the new alloca, so we need to clear
// their 'tail' flags if HandleByValArgument introduced a new alloca and
// the callee has calls.
MustClearTailCallFlags |= ActualArg != *AI;
}
VMap[I] = ActualArg;
}
// We want the inliner to prune the code as it copies. We would LOVE to
// have no dead or constant instructions leftover after inlining occurs
// (which can happen, e.g., because an argument was constant), but we'll be
// happy with whatever the cloner can do.
CloneAndPruneFunctionInto(Caller, CalledFunc, VMap,
/*ModuleLevelChanges=*/false, Returns, ".i",
&InlinedFunctionInfo, IFI.TD, TheCall);
// Remember the first block that is newly cloned over.
FirstNewBlock = LastBlock; ++FirstNewBlock;
// Update the callgraph if requested.
if (IFI.CG)
UpdateCallGraphAfterInlining(CS, FirstNewBlock, VMap, IFI);
// Update inlined instructions' line number information.
fixupLineNumbers(Caller, FirstNewBlock, TheCall);
}
// If there are any alloca instructions in the block that used to be the entry
// block for the callee, move them to the entry block of the caller. First
// calculate which instruction they should be inserted before. We insert the
// instructions at the end of the current alloca list.
{
BasicBlock::iterator InsertPoint = Caller->begin()->begin();
for (BasicBlock::iterator I = FirstNewBlock->begin(),
E = FirstNewBlock->end(); I != E; ) {
AllocaInst *AI = dyn_cast<AllocaInst>(I++);
if (AI == 0) continue;
// If the alloca is now dead, remove it. This often occurs due to code
// specialization.
if (AI->use_empty()) {
AI->eraseFromParent();
continue;
}
if (!isa<Constant>(AI->getArraySize()))
continue;
// Keep track of the static allocas that we inline into the caller.
IFI.StaticAllocas.push_back(AI);
// Scan for the block of allocas that we can move over, and move them
// all at once.
while (isa<AllocaInst>(I) &&
isa<Constant>(cast<AllocaInst>(I)->getArraySize())) {
IFI.StaticAllocas.push_back(cast<AllocaInst>(I));
++I;
}
// Transfer all of the allocas over in a block. Using splice means
// that the instructions aren't removed from the symbol table, then
// reinserted.
Caller->getEntryBlock().getInstList().splice(InsertPoint,
FirstNewBlock->getInstList(),
AI, I);
}
}
// Leave lifetime markers for the static alloca's, scoping them to the
// function we just inlined.
if (InsertLifetime && !IFI.StaticAllocas.empty()) {
IRBuilder<> builder(FirstNewBlock->begin());
for (unsigned ai = 0, ae = IFI.StaticAllocas.size(); ai != ae; ++ai) {
AllocaInst *AI = IFI.StaticAllocas[ai];
// If the alloca is already scoped to something smaller than the whole
// function then there's no need to add redundant, less accurate markers.
if (hasLifetimeMarkers(AI))
continue;
// Try to determine the size of the allocation.
ConstantInt *AllocaSize = 0;
if (ConstantInt *AIArraySize =
dyn_cast<ConstantInt>(AI->getArraySize())) {
if (IFI.TD) {
Type *AllocaType = AI->getAllocatedType();
uint64_t AllocaTypeSize = IFI.TD->getTypeAllocSize(AllocaType);
uint64_t AllocaArraySize = AIArraySize->getLimitedValue();
assert(AllocaArraySize > 0 && "array size of AllocaInst is zero");
// Check that array size doesn't saturate uint64_t and doesn't
// overflow when it's multiplied by type size.
if (AllocaArraySize != ~0ULL &&
UINT64_MAX / AllocaArraySize >= AllocaTypeSize) {
AllocaSize = ConstantInt::get(Type::getInt64Ty(AI->getContext()),
AllocaArraySize * AllocaTypeSize);
}
}
}
builder.CreateLifetimeStart(AI, AllocaSize);
for (unsigned ri = 0, re = Returns.size(); ri != re; ++ri) {
IRBuilder<> builder(Returns[ri]);
builder.CreateLifetimeEnd(AI, AllocaSize);
}
}
}
// If the inlined code contained dynamic alloca instructions, wrap the inlined
// code with llvm.stacksave/llvm.stackrestore intrinsics.
if (InlinedFunctionInfo.ContainsDynamicAllocas) {
Module *M = Caller->getParent();
// Get the two intrinsics we care about.
Function *StackSave = Intrinsic::getDeclaration(M, Intrinsic::stacksave);
Function *StackRestore=Intrinsic::getDeclaration(M,Intrinsic::stackrestore);
// Insert the llvm.stacksave.
CallInst *SavedPtr = IRBuilder<>(FirstNewBlock, FirstNewBlock->begin())
.CreateCall(StackSave, "savedstack");
// Insert a call to llvm.stackrestore before any return instructions in the
// inlined function.
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
IRBuilder<>(Returns[i]).CreateCall(StackRestore, SavedPtr);
}
}
// If we are inlining tail call instruction through a call site that isn't
// marked 'tail', we must remove the tail marker for any calls in the inlined
// code. Also, calls inlined through a 'nounwind' call site should be marked
// 'nounwind'.
if (InlinedFunctionInfo.ContainsCalls &&
(MustClearTailCallFlags || MarkNoUnwind)) {
for (Function::iterator BB = FirstNewBlock, E = Caller->end();
BB != E; ++BB)
for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
if (CallInst *CI = dyn_cast<CallInst>(I)) {
if (MustClearTailCallFlags)
CI->setTailCall(false);
if (MarkNoUnwind)
CI->setDoesNotThrow();
}
}
// If we are inlining for an invoke instruction, we must make sure to rewrite
// any call instructions into invoke instructions.
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall))
HandleInlinedInvoke(II, FirstNewBlock, InlinedFunctionInfo);
// If we cloned in _exactly one_ basic block, and if that block ends in a
// return instruction, we splice the body of the inlined callee directly into
// the calling basic block.
if (Returns.size() == 1 && std::distance(FirstNewBlock, Caller->end()) == 1) {
// Move all of the instructions right before the call.
OrigBB->getInstList().splice(TheCall, FirstNewBlock->getInstList(),
FirstNewBlock->begin(), FirstNewBlock->end());
// Remove the cloned basic block.
Caller->getBasicBlockList().pop_back();
// If the call site was an invoke instruction, add a branch to the normal
// destination.
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall)) {
BranchInst *NewBr = BranchInst::Create(II->getNormalDest(), TheCall);
NewBr->setDebugLoc(Returns[0]->getDebugLoc());
}
// If the return instruction returned a value, replace uses of the call with
// uses of the returned value.
if (!TheCall->use_empty()) {
ReturnInst *R = Returns[0];
if (TheCall == R->getReturnValue())
TheCall->replaceAllUsesWith(UndefValue::get(TheCall->getType()));
else
TheCall->replaceAllUsesWith(R->getReturnValue());
}
// Since we are now done with the Call/Invoke, we can delete it.
TheCall->eraseFromParent();
// Since we are now done with the return instruction, delete it also.
Returns[0]->eraseFromParent();
// We are now done with the inlining.
return true;
}
// Otherwise, we have the normal case, of more than one block to inline or
// multiple return sites.
// We want to clone the entire callee function into the hole between the
// "starter" and "ender" blocks. How we accomplish this depends on whether
// this is an invoke instruction or a call instruction.
BasicBlock *AfterCallBB;
BranchInst *CreatedBranchToNormalDest = NULL;
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall)) {
// Add an unconditional branch to make this look like the CallInst case...
CreatedBranchToNormalDest = BranchInst::Create(II->getNormalDest(), TheCall);
// Split the basic block. This guarantees that no PHI nodes will have to be
// updated due to new incoming edges, and make the invoke case more
// symmetric to the call case.
AfterCallBB = OrigBB->splitBasicBlock(CreatedBranchToNormalDest,
CalledFunc->getName()+".exit");
} else { // It's a call
// If this is a call instruction, we need to split the basic block that
// the call lives in.
//
AfterCallBB = OrigBB->splitBasicBlock(TheCall,
CalledFunc->getName()+".exit");
}
// Change the branch that used to go to AfterCallBB to branch to the first
// basic block of the inlined function.
//
TerminatorInst *Br = OrigBB->getTerminator();
assert(Br && Br->getOpcode() == Instruction::Br &&
"splitBasicBlock broken!");
Br->setOperand(0, FirstNewBlock);
// Now that the function is correct, make it a little bit nicer. In
// particular, move the basic blocks inserted from the end of the function
// into the space made by splitting the source basic block.
Caller->getBasicBlockList().splice(AfterCallBB, Caller->getBasicBlockList(),
FirstNewBlock, Caller->end());
// Handle all of the return instructions that we just cloned in, and eliminate
// any users of the original call/invoke instruction.
Type *RTy = CalledFunc->getReturnType();
PHINode *PHI = 0;
if (Returns.size() > 1) {
// The PHI node should go at the front of the new basic block to merge all
// possible incoming values.
if (!TheCall->use_empty()) {
PHI = PHINode::Create(RTy, Returns.size(), TheCall->getName(),
AfterCallBB->begin());
// Anything that used the result of the function call should now use the
// PHI node as their operand.
TheCall->replaceAllUsesWith(PHI);
}
// Loop over all of the return instructions adding entries to the PHI node
// as appropriate.
if (PHI) {
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
ReturnInst *RI = Returns[i];
assert(RI->getReturnValue()->getType() == PHI->getType() &&
"Ret value not consistent in function!");
PHI->addIncoming(RI->getReturnValue(), RI->getParent());
}
}
// Add a branch to the merge points and remove return instructions.
DebugLoc Loc;
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
ReturnInst *RI = Returns[i];
BranchInst* BI = BranchInst::Create(AfterCallBB, RI);
Loc = RI->getDebugLoc();
BI->setDebugLoc(Loc);
RI->eraseFromParent();
}
// We need to set the debug location to *somewhere* inside the
// inlined function. The line number may be nonsensical, but the
// instruction will at least be associated with the right
// function.
if (CreatedBranchToNormalDest)
CreatedBranchToNormalDest->setDebugLoc(Loc);
} else if (!Returns.empty()) {
// Otherwise, if there is exactly one return value, just replace anything
// using the return value of the call with the computed value.
if (!TheCall->use_empty()) {
if (TheCall == Returns[0]->getReturnValue())
TheCall->replaceAllUsesWith(UndefValue::get(TheCall->getType()));
else
TheCall->replaceAllUsesWith(Returns[0]->getReturnValue());
}
// Update PHI nodes that use the ReturnBB to use the AfterCallBB.
BasicBlock *ReturnBB = Returns[0]->getParent();
ReturnBB->replaceAllUsesWith(AfterCallBB);
// Splice the code from the return block into the block that it will return
// to, which contains the code that was after the call.
AfterCallBB->getInstList().splice(AfterCallBB->begin(),
ReturnBB->getInstList());
if (CreatedBranchToNormalDest)
CreatedBranchToNormalDest->setDebugLoc(Returns[0]->getDebugLoc());
// Delete the return instruction now and empty ReturnBB now.
Returns[0]->eraseFromParent();
ReturnBB->eraseFromParent();
} else if (!TheCall->use_empty()) {
// No returns, but something is using the return value of the call. Just
// nuke the result.
TheCall->replaceAllUsesWith(UndefValue::get(TheCall->getType()));
}
// Since we are now done with the Call/Invoke, we can delete it.
TheCall->eraseFromParent();
// We should always be able to fold the entry block of the function into the
// single predecessor of the block...
assert(cast<BranchInst>(Br)->isUnconditional() && "splitBasicBlock broken!");
BasicBlock *CalleeEntry = cast<BranchInst>(Br)->getSuccessor(0);
// Splice the code entry block into calling block, right before the
// unconditional branch.
CalleeEntry->replaceAllUsesWith(OrigBB); // Update PHI nodes
OrigBB->getInstList().splice(Br, CalleeEntry->getInstList());
// Remove the unconditional branch.
OrigBB->getInstList().erase(Br);
// Now we can remove the CalleeEntry block, which is now empty.
Caller->getBasicBlockList().erase(CalleeEntry);
// If we inserted a phi node, check to see if it has a single value (e.g. all
// the entries are the same or undef). If so, remove the PHI so it doesn't
// block other optimizations.
if (PHI) {
if (Value *V = SimplifyInstruction(PHI, IFI.TD)) {
PHI->replaceAllUsesWith(V);
PHI->eraseFromParent();
}
}
return true;
}
// InlineFunction - This function inlines the called function into the basic
// block of the caller. This returns false if it is not possible to inline this
// call. The program is still in a well defined state if this occurs though.
//
// Note that this only does one level of inlining. For example, if the
// instruction 'call B' is inlined, and 'B' calls 'C', then the call to 'C' now
// exists in the instruction stream. Similiarly this will inline a recursive
// function by one level.
//
bool llvm::InlineFunction(CallSite CS, CallGraph *CG, const TargetData *TD) {
Instruction *TheCall = CS.getInstruction();
assert(TheCall->getParent() && TheCall->getParent()->getParent() &&
"Instruction not in function!");
const Function *CalledFunc = CS.getCalledFunction();
if (CalledFunc == 0 || // Can't inline external function or indirect
CalledFunc->isDeclaration() || // call, or call to a vararg function!
CalledFunc->getFunctionType()->isVarArg()) return false;
// If the call to the callee is a non-tail call, we must clear the 'tail'
// flags on any calls that we inline.
bool MustClearTailCallFlags =
isa<CallInst>(TheCall) && !cast<CallInst>(TheCall)->isTailCall();
BasicBlock *OrigBB = TheCall->getParent();
Function *Caller = OrigBB->getParent();
// Get an iterator to the last basic block in the function, which will have
// the new function inlined after it.
//
Function::iterator LastBlock = &Caller->back();
// Make sure to capture all of the return instructions from the cloned
// function.
std::vector<ReturnInst*> Returns;
ClonedCodeInfo InlinedFunctionInfo;
Function::iterator FirstNewBlock;
{ // Scope to destroy ValueMap after cloning.
DenseMap<const Value*, Value*> ValueMap;
// Calculate the vector of arguments to pass into the function cloner, which
// matches up the formal to the actual argument values.
assert(std::distance(CalledFunc->arg_begin(), CalledFunc->arg_end()) ==
std::distance(CS.arg_begin(), CS.arg_end()) &&
"No varargs calls can be inlined!");
CallSite::arg_iterator AI = CS.arg_begin();
for (Function::const_arg_iterator I = CalledFunc->arg_begin(),
E = CalledFunc->arg_end(); I != E; ++I, ++AI)
ValueMap[I] = *AI;
// We want the inliner to prune the code as it copies. We would LOVE to
// have no dead or constant instructions leftover after inlining occurs
// (which can happen, e.g., because an argument was constant), but we'll be
// happy with whatever the cloner can do.
CloneAndPruneFunctionInto(Caller, CalledFunc, ValueMap, Returns, ".i",
&InlinedFunctionInfo, TD);
// Remember the first block that is newly cloned over.
FirstNewBlock = LastBlock; ++FirstNewBlock;
// Update the callgraph if requested.
if (CG)
UpdateCallGraphAfterInlining(Caller, CalledFunc, FirstNewBlock, ValueMap,
*CG);
}
// If there are any alloca instructions in the block that used to be the entry
// block for the callee, move them to the entry block of the caller. First
// calculate which instruction they should be inserted before. We insert the
// instructions at the end of the current alloca list.
//
{
BasicBlock::iterator InsertPoint = Caller->begin()->begin();
for (BasicBlock::iterator I = FirstNewBlock->begin(),
E = FirstNewBlock->end(); I != E; )
if (AllocaInst *AI = dyn_cast<AllocaInst>(I++)) {
// If the alloca is now dead, remove it. This often occurs due to code
// specialization.
if (AI->use_empty()) {
AI->eraseFromParent();
continue;
}
if (isa<Constant>(AI->getArraySize())) {
// Scan for the block of allocas that we can move over, and move them
// all at once.
while (isa<AllocaInst>(I) &&
isa<Constant>(cast<AllocaInst>(I)->getArraySize()))
++I;
// Transfer all of the allocas over in a block. Using splice means
// that the instructions aren't removed from the symbol table, then
// reinserted.
Caller->getEntryBlock().getInstList().splice(
InsertPoint,
FirstNewBlock->getInstList(),
AI, I);
}
}
}
// If the inlined code contained dynamic alloca instructions, wrap the inlined
// code with llvm.stacksave/llvm.stackrestore intrinsics.
if (InlinedFunctionInfo.ContainsDynamicAllocas) {
Module *M = Caller->getParent();
const Type *BytePtr = PointerType::get(Type::Int8Ty);
// Get the two intrinsics we care about.
Constant *StackSave, *StackRestore;
StackSave = M->getOrInsertFunction("llvm.stacksave", BytePtr, NULL);
StackRestore = M->getOrInsertFunction("llvm.stackrestore", Type::VoidTy,
BytePtr, NULL);
// If we are preserving the callgraph, add edges to the stacksave/restore
// functions for the calls we insert.
CallGraphNode *StackSaveCGN = 0, *StackRestoreCGN = 0, *CallerNode = 0;
if (CG) {
// We know that StackSave/StackRestore are Function*'s, because they are
// intrinsics which must have the right types.
StackSaveCGN = CG->getOrInsertFunction(cast<Function>(StackSave));
StackRestoreCGN = CG->getOrInsertFunction(cast<Function>(StackRestore));
CallerNode = (*CG)[Caller];
}
// Insert the llvm.stacksave.
CallInst *SavedPtr = new CallInst(StackSave, "savedstack",
FirstNewBlock->begin());
if (CG) CallerNode->addCalledFunction(SavedPtr, StackSaveCGN);
// Insert a call to llvm.stackrestore before any return instructions in the
// inlined function.
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
CallInst *CI = new CallInst(StackRestore, SavedPtr, "", Returns[i]);
if (CG) CallerNode->addCalledFunction(CI, StackRestoreCGN);
}
// Count the number of StackRestore calls we insert.
unsigned NumStackRestores = Returns.size();
// If we are inlining an invoke instruction, insert restores before each
// unwind. These unwinds will be rewritten into branches later.
if (InlinedFunctionInfo.ContainsUnwinds && isa<InvokeInst>(TheCall)) {
for (Function::iterator BB = FirstNewBlock, E = Caller->end();
BB != E; ++BB)
if (UnwindInst *UI = dyn_cast<UnwindInst>(BB->getTerminator())) {
new CallInst(StackRestore, SavedPtr, "", UI);
++NumStackRestores;
}
}
}
// If we are inlining tail call instruction through a call site that isn't
// marked 'tail', we must remove the tail marker for any calls in the inlined
// code.
if (MustClearTailCallFlags && InlinedFunctionInfo.ContainsCalls) {
for (Function::iterator BB = FirstNewBlock, E = Caller->end();
BB != E; ++BB)
for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
if (CallInst *CI = dyn_cast<CallInst>(I))
CI->setTailCall(false);
}
// If we are inlining for an invoke instruction, we must make sure to rewrite
// any inlined 'unwind' instructions into branches to the invoke exception
// destination, and call instructions into invoke instructions.
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall))
HandleInlinedInvoke(II, FirstNewBlock, InlinedFunctionInfo);
// If we cloned in _exactly one_ basic block, and if that block ends in a
// return instruction, we splice the body of the inlined callee directly into
// the calling basic block.
if (Returns.size() == 1 && std::distance(FirstNewBlock, Caller->end()) == 1) {
// Move all of the instructions right before the call.
OrigBB->getInstList().splice(TheCall, FirstNewBlock->getInstList(),
FirstNewBlock->begin(), FirstNewBlock->end());
// Remove the cloned basic block.
Caller->getBasicBlockList().pop_back();
// If the call site was an invoke instruction, add a branch to the normal
// destination.
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall))
new BranchInst(II->getNormalDest(), TheCall);
// If the return instruction returned a value, replace uses of the call with
// uses of the returned value.
if (!TheCall->use_empty())
TheCall->replaceAllUsesWith(Returns[0]->getReturnValue());
// Since we are now done with the Call/Invoke, we can delete it.
TheCall->getParent()->getInstList().erase(TheCall);
// Since we are now done with the return instruction, delete it also.
Returns[0]->getParent()->getInstList().erase(Returns[0]);
// We are now done with the inlining.
return true;
}
// Otherwise, we have the normal case, of more than one block to inline or
// multiple return sites.
// We want to clone the entire callee function into the hole between the
// "starter" and "ender" blocks. How we accomplish this depends on whether
// this is an invoke instruction or a call instruction.
BasicBlock *AfterCallBB;
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall)) {
// Add an unconditional branch to make this look like the CallInst case...
BranchInst *NewBr = new BranchInst(II->getNormalDest(), TheCall);
// Split the basic block. This guarantees that no PHI nodes will have to be
// updated due to new incoming edges, and make the invoke case more
// symmetric to the call case.
AfterCallBB = OrigBB->splitBasicBlock(NewBr,
CalledFunc->getName()+".exit");
} else { // It's a call
// If this is a call instruction, we need to split the basic block that
// the call lives in.
//
AfterCallBB = OrigBB->splitBasicBlock(TheCall,
CalledFunc->getName()+".exit");
}
// Change the branch that used to go to AfterCallBB to branch to the first
// basic block of the inlined function.
//
TerminatorInst *Br = OrigBB->getTerminator();
assert(Br && Br->getOpcode() == Instruction::Br &&
"splitBasicBlock broken!");
Br->setOperand(0, FirstNewBlock);
// Now that the function is correct, make it a little bit nicer. In
// particular, move the basic blocks inserted from the end of the function
// into the space made by splitting the source basic block.
//
Caller->getBasicBlockList().splice(AfterCallBB, Caller->getBasicBlockList(),
FirstNewBlock, Caller->end());
// Handle all of the return instructions that we just cloned in, and eliminate
// any users of the original call/invoke instruction.
if (Returns.size() > 1) {
// The PHI node should go at the front of the new basic block to merge all
// possible incoming values.
//
PHINode *PHI = 0;
if (!TheCall->use_empty()) {
PHI = new PHINode(CalledFunc->getReturnType(),
TheCall->getName(), AfterCallBB->begin());
// Anything that used the result of the function call should now use the
// PHI node as their operand.
//
TheCall->replaceAllUsesWith(PHI);
}
// Loop over all of the return instructions, turning them into unconditional
// branches to the merge point now, and adding entries to the PHI node as
// appropriate.
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
ReturnInst *RI = Returns[i];
if (PHI) {
assert(RI->getReturnValue() && "Ret should have value!");
assert(RI->getReturnValue()->getType() == PHI->getType() &&
"Ret value not consistent in function!");
PHI->addIncoming(RI->getReturnValue(), RI->getParent());
}
// Add a branch to the merge point where the PHI node lives if it exists.
new BranchInst(AfterCallBB, RI);
// Delete the return instruction now
RI->getParent()->getInstList().erase(RI);
}
} else if (!Returns.empty()) {
// Otherwise, if there is exactly one return value, just replace anything
// using the return value of the call with the computed value.
if (!TheCall->use_empty())
TheCall->replaceAllUsesWith(Returns[0]->getReturnValue());
// Splice the code from the return block into the block that it will return
// to, which contains the code that was after the call.
BasicBlock *ReturnBB = Returns[0]->getParent();
AfterCallBB->getInstList().splice(AfterCallBB->begin(),
ReturnBB->getInstList());
// Update PHI nodes that use the ReturnBB to use the AfterCallBB.
ReturnBB->replaceAllUsesWith(AfterCallBB);
// Delete the return instruction now and empty ReturnBB now.
Returns[0]->eraseFromParent();
ReturnBB->eraseFromParent();
} else if (!TheCall->use_empty()) {
// No returns, but something is using the return value of the call. Just
// nuke the result.
TheCall->replaceAllUsesWith(UndefValue::get(TheCall->getType()));
}
// Since we are now done with the Call/Invoke, we can delete it.
TheCall->eraseFromParent();
// We should always be able to fold the entry block of the function into the
// single predecessor of the block...
assert(cast<BranchInst>(Br)->isUnconditional() && "splitBasicBlock broken!");
BasicBlock *CalleeEntry = cast<BranchInst>(Br)->getSuccessor(0);
// Splice the code entry block into calling block, right before the
// unconditional branch.
OrigBB->getInstList().splice(Br, CalleeEntry->getInstList());
CalleeEntry->replaceAllUsesWith(OrigBB); // Update PHI nodes
// Remove the unconditional branch.
OrigBB->getInstList().erase(Br);
// Now we can remove the CalleeEntry block, which is now empty.
Caller->getBasicBlockList().erase(CalleeEntry);
return true;
}
void DSWP::cleanup(Loop *L, LPPassManager &LPM) {
// Move some instructions that may not have been inserted in the right
// place, delete the old loop, and clean up our aux data structures for this
// loop.
/*
* move the produce instructions, which have been inserted after the branch,
* in front of it
*/
for (int i = 0; i < MAX_THREAD; i++) {
for (Function::iterator bi = allFunc[i]->begin(), be = allFunc[i]->end(); bi != be; ++bi) {
BasicBlock *bb = bi;
TerminatorInst *term = NULL;
for (BasicBlock::iterator ii = bb->begin(), ie = bb->end(); ii != ie; ++ii) {
Instruction *inst = ii;
if (isa<TerminatorInst>(inst)) {
term = dyn_cast<TerminatorInst>(inst);
break;
}
}
if (term == NULL) {
error("term cannot be null");
}
while (true) {
Instruction *last = &bb->getInstList().back();
if (isa<TerminatorInst>(last))
break;
last->moveBefore(term);
}
}
}
/*
* move the phi nodes to the top of the block
*/
for (int i = 0; i < MAX_THREAD; i++) {
for (Function::iterator bi = allFunc[i]->begin(), be = allFunc[i]->end(); bi != be; ++bi) {
BasicBlock *bb = bi;
Instruction *first_nonphi = bb->getFirstNonPHI();
BasicBlock::iterator ii = bb->begin(), ie = bb->end();
// advance the iterator up to one past first_nonphi
while (&(*ii) != first_nonphi) {
++ii;
}
++ii;
// move any phi nodes after the first nonphi to before it
for (BasicBlock::iterator i_next; ii != ie; ii = i_next) {
i_next = ii;
++i_next;
Instruction *inst = ii;
if (isa<PHINode>(inst)) {
inst->moveBefore(first_nonphi);
}
}
}
}
cout << "begin to delete loop" << endl;
for (Loop::block_iterator bi = L->block_begin(), be = L->block_end(); bi != be; ++bi) {
BasicBlock *BB = *bi;
for (BasicBlock::iterator ii = BB->begin(), i_next, ie = BB->end(); ii != ie; ii = i_next) {
i_next = ii;
++i_next;
Instruction &inst = *ii;
inst.replaceAllUsesWith(UndefValue::get(inst.getType()));
inst.eraseFromParent();
}
}
// Delete the basic blocks only afterwards
// so that backwards branch instructions don't break
for (Loop::block_iterator bi = L->block_begin(), be = L->block_end(); bi != be; ++bi) {
BasicBlock *BB = *bi;
BB->eraseFromParent();
}
LPM.deleteLoopFromQueue(L);
}
bool partition::runOnLoop(Loop* L, LPPassManager &LPM) {
errs() << "*************************** Loop encountered: ************************" << '\n' << L->getHeader()->getName() << '\n';
if (function->getName() != "main")
return false;
IntegerType* int32Ty = Type::getInt32Ty(*context);
IntegerType* int64Ty = Type::getInt64Ty(*context);
PointerType* voidPtrTy = Type::getInt8PtrTy(*context);
FunctionType* funcTy = FunctionType::get(int32Ty, false);
Constant* func1_c;
Function* func1;
func1_c = module->getOrInsertFunction("func1", funcTy);
func1 = cast<Function>(func1_c);
Function* pro = module->getFunction("produce");
Function* con = module->getFunction("consume");
BasicBlock* func1EntryBlock = BasicBlock::Create(*context, "entry.func1", func1);
AllocaInst* i_var = new AllocaInst(int32Ty, NULL, 4, "i", func1EntryBlock);
Value* liveIn;
BasicBlock *forCond, *forBody, *forInc;
ValueToValueMapTy VMap;
std::map<BasicBlock *, BasicBlock *> BlockMap;
for (Loop::block_iterator BB = L->block_begin(), BBe = L->block_end(); BB != BBe; ++BB) {
BasicBlock* func1Block = CloneBasicBlock(*BB, VMap, ".func1", func1);
BlockMap[*BB] = func1Block;
if ((*BB)->getName() == "for.cond")
forCond = func1Block;
if ((*BB)->getName() == "for.body")
forBody = func1Block;
if ((*BB)->getName() == "for.inc")
forInc = func1Block;
for (BasicBlock::iterator it = func1Block->begin(), ite = func1Block->end(); it != ite; ++it) {
for (User::op_iterator oit = it->op_begin(), oite = it->op_end(); oit != oite; ++oit) {
if (VMap[*oit] != NULL) {
*oit = VMap[*oit];
} else {
Constant* cons = dyn_cast<Constant>(*oit);
BranchInst* br = dyn_cast<BranchInst>(it);
if (cons == NULL && br == NULL) {
liveIn = *oit;
*oit = i_var;
}
}
}
}
if ((*BB)->getName() == "for.body") {
Instruction* term = (*BB)->getTerminator();
term->removeFromParent();
for (int i = 0; i < 7; i++) {
(*BB)->back().eraseFromParent();
}
term->insertAfter(&(*BB)->back());
(*BB)->front().eraseFromParent();
LoadInst* load = new LoadInst(liveIn, "", false, 4, term);
std::vector<Value *> produce_args;
CastInst* cast = CastInst::CreateIntegerCast(load, int64Ty, true);
cast->insertAfter(load);
produce_args.push_back(cast);
ConstantInt* val = ConstantInt::get(int32Ty, (uint32_t) 3);
produce_args.push_back(val);
CallInst::Create(pro, ArrayRef<Value*>(produce_args), "", term);
produce_args.pop_back();
val = ConstantInt::get(int32Ty, (uint32_t) 2);
produce_args.push_back(val);
CallInst::Create(pro, ArrayRef<Value*>(produce_args), "", term);
}
}
// set branch instructions to restructure the CFG in created function
BasicBlock* func1EndBlock = BasicBlock::Create(*context, "if.end.func1", func1);
BasicBlock* garbageBB = BasicBlock::Create(*context, "garbage", func1);
ConstantInt* retVal_g = ConstantInt::get(int32Ty, (uint32_t) 0);
ReturnInst* ret_g = ReturnInst::Create(*context, retVal_g, garbageBB);
for (Function::iterator fit = func1->begin(), fite = func1->end(); fit != fite; ++fit) {
if (fit->getTerminator() == NULL || fit->getName() == "garbage")
continue;
BranchInst* br = dyn_cast<BranchInst>(fit->getTerminator());
int numSuccessors = br->getNumSuccessors();
for (int i = 0; i < numSuccessors; i++) {
BasicBlock* successor = br->getSuccessor(i);
if (BlockMap[successor] != NULL) {
br->setSuccessor(i, BlockMap[successor]);
}
else {
br->setSuccessor(i, func1EndBlock);
}
}
/*
if (fit->getName() == "for.body.func1") {
for (int i = 0; i < 4; i++) {
BasicBlock::iterator it = fit->begin();
it->moveBefore(ret_g);
}
}
*/
}
garbageBB->eraseFromParent();
BranchInst* br = dyn_cast<BranchInst>(forBody->getTerminator());
br->setSuccessor(0, forCond);
forInc->eraseFromParent();
// Create return instruction for func1EndBlock and set a branch from loop header to func1EndBlock
ConstantInt* retVal = ConstantInt::get(int32Ty, (uint32_t) 0);
ReturnInst* ret1 = ReturnInst::Create(*context, retVal, func1EndBlock);
BasicBlock* loopHeader = BlockMap.at(L->getHeader());
BranchInst* brInst = BranchInst::Create(loopHeader, func1EntryBlock);
// add produce function call
std::vector<Value *> produce_args;
ConstantInt* val = ConstantInt::get(int64Ty, (uint64_t) 0);
produce_args.push_back(val);
val = ConstantInt::get(int32Ty, (uint32_t) 5);
produce_args.push_back(val);
CallInst::Create(pro, ArrayRef<Value*>(produce_args), "", func1EndBlock->getTerminator());
// add consume function call
int q_id = 2;
for (Value::use_iterator uit = i_var->use_begin(), uite = i_var->use_end(); uit != uite; ++uit) {
std::vector<Value *> consume_args;
ConstantInt* val = ConstantInt::get(int32Ty, (uint32_t) q_id);
consume_args.push_back(val);
CallInst* call = CallInst::Create(con, ArrayRef<Value*>(consume_args));
Instruction* inst = dyn_cast<Instruction>(*uit);
call->insertAfter(inst);
CastInst* cast = CastInst::CreateIntegerCast(call, int32Ty, true);
cast->insertAfter(call);
(*uit)->replaceAllUsesWith(cast);
inst->eraseFromParent();
q_id++;
}
i_var->eraseFromParent();
// add produce and consume function calls to main thread
// transmit the function pointer to created function by a produce call
BasicBlock* loopPreheader = L->getLoopPreheader();
produce_args.clear();
CastInst* cast = CastInst::CreatePointerCast(func1, int64Ty);
cast->insertBefore(loopPreheader->getTerminator());
produce_args.push_back(cast);
val = ConstantInt::get(int32Ty, (uint32_t) 0);
produce_args.push_back(val);
CallInst::Create(pro, ArrayRef<Value*>(produce_args), "", loopPreheader->getTerminator());
// transmit induction variable to created function by a produce call
Instruction* load = &L->getHeader()->front();
produce_args.clear();
cast = CastInst::CreateIntegerCast(load, int64Ty, true);
cast->insertAfter(load);
produce_args.push_back(cast);
val = ConstantInt::get(int32Ty, (uint32_t) 4);
produce_args.push_back(val);
CallInst::Create(pro, ArrayRef<Value*>(produce_args))->insertAfter(cast);
return true;
}
