static void shuffleValueUseLists(Value *V, std::minstd_rand0 &Gen,
DenseSet<Value *> &Seen) {
if (!Seen.insert(V).second)
return;
if (auto *C = dyn_cast<Constant>(V))
if (!isa<GlobalValue>(C))
for (Value *Op : C->operands())
shuffleValueUseLists(Op, Gen, Seen);
if (V->use_empty() || std::next(V->use_begin()) == V->use_end())
// Nothing to shuffle for 0 or 1 users.
return;
// Generate random numbers between 10 and 99, which will line up nicely in
// debug output. We're not worried about collisons here.
DEBUG(dbgs() << "V = "; V->dump());
std::uniform_int_distribution<short> Dist(10, 99);
SmallDenseMap<const Use *, short, 16> Order;
for (const Use &U : V->uses()) {
auto I = Dist(Gen);
Order[&U] = I;
DEBUG(dbgs() << " - order: " << I << ", U = "; U.getUser()->dump());
}
DEBUG(dbgs() << " => shuffle\n");
V->sortUseList(
[&Order](const Use &L, const Use &R) { return Order[&L] < Order[&R]; });
DEBUG({
for (const Use &U : V->uses())
DEBUG(dbgs() << " - order: " << Order.lookup(&U) << ", U = ";
U.getUser()->dump());
});
/// Sort local variables so that variables appearing inside of helper
/// expressions come first.
static SmallVector<DbgVariable *, 8>
sortLocalVars(SmallVectorImpl<DbgVariable *> &Input) {
SmallVector<DbgVariable *, 8> Result;
SmallVector<PointerIntPair<DbgVariable *, 1>, 8> WorkList;
// Map back from a DIVariable to its containing DbgVariable.
SmallDenseMap<const DILocalVariable *, DbgVariable *> DbgVar;
// Set of DbgVariables in Result.
SmallDenseSet<DbgVariable *, 8> Visited;
// For cycle detection.
SmallDenseSet<DbgVariable *, 8> Visiting;
// Initialize the worklist and the DIVariable lookup table.
for (auto Var : reverse(Input)) {
DbgVar.insert({Var->getVariable(), Var});
WorkList.push_back({Var, 0});
}
// Perform a stable topological sort by doing a DFS.
while (!WorkList.empty()) {
auto Item = WorkList.back();
DbgVariable *Var = Item.getPointer();
bool visitedAllDependencies = Item.getInt();
WorkList.pop_back();
// Dependency is in a different lexical scope or a global.
if (!Var)
continue;
// Already handled.
if (Visited.count(Var))
continue;
// Add to Result if all dependencies are visited.
if (visitedAllDependencies) {
Visited.insert(Var);
Result.push_back(Var);
continue;
}
// Detect cycles.
auto Res = Visiting.insert(Var);
if (!Res.second) {
assert(false && "dependency cycle in local variables");
return Result;
}
// Push dependencies and this node onto the worklist, so that this node is
// visited again after all of its dependencies are handled.
WorkList.push_back({Var, 1});
for (auto *Dependency : dependencies(Var)) {
auto Dep = dyn_cast_or_null<const DILocalVariable>(Dependency);
WorkList.push_back({DbgVar[Dep], 0});
}
}
return Result;
}
// This function calculates the number of iterations after which the given Phi
// becomes an invariant. The pre-calculated values are memorized in the map. The
// function (shortcut is I) is calculated according to the following definition:
// Given %x = phi <Inputs from above the loop>, ..., [%y, %back.edge].
// If %y is a loop invariant, then I(%x) = 1.
// If %y is a Phi from the loop header, I(%x) = I(%y) + 1.
// Otherwise, I(%x) is infinite.
// TODO: Actually if %y is an expression that depends only on Phi %z and some
// loop invariants, we can estimate I(%x) = I(%z) + 1. The example
// looks like:
// %x = phi(0, %a), <-- becomes invariant starting from 3rd iteration.
// %y = phi(0, 5),
// %a = %y + 1.
static unsigned calculateIterationsToInvariance(
PHINode *Phi, Loop *L, BasicBlock *BackEdge,
SmallDenseMap<PHINode *, unsigned> &IterationsToInvariance) {
assert(Phi->getParent() == L->getHeader() &&
"Non-loop Phi should not be checked for turning into invariant.");
assert(BackEdge == L->getLoopLatch() && "Wrong latch?");
// If we already know the answer, take it from the map.
auto I = IterationsToInvariance.find(Phi);
if (I != IterationsToInvariance.end())
return I->second;
// Otherwise we need to analyze the input from the back edge.
Value *Input = Phi->getIncomingValueForBlock(BackEdge);
// Place infinity to map to avoid infinite recursion for cycled Phis. Such
// cycles can never stop on an invariant.
IterationsToInvariance[Phi] = InfiniteIterationsToInvariance;
unsigned ToInvariance = InfiniteIterationsToInvariance;
if (L->isLoopInvariant(Input))
ToInvariance = 1u;
else if (PHINode *IncPhi = dyn_cast<PHINode>(Input)) {
// Only consider Phis in header block.
if (IncPhi->getParent() != L->getHeader())
return InfiniteIterationsToInvariance;
// If the input becomes an invariant after X iterations, then our Phi
// becomes an invariant after X + 1 iterations.
unsigned InputToInvariance = calculateIterationsToInvariance(
IncPhi, L, BackEdge, IterationsToInvariance);
if (InputToInvariance != InfiniteIterationsToInvariance)
ToInvariance = InputToInvariance + 1u;
}
// If we found that this Phi lies in an invariant chain, update the map.
if (ToInvariance != InfiniteIterationsToInvariance)
IterationsToInvariance[Phi] = ToInvariance;
return ToInvariance;
}
static bool IsEquivalentPHI(PHINode *PHI,
SmallDenseMap<BasicBlock*, Value*, 8> &ValueMapping) {
unsigned PHINumValues = PHI->getNumIncomingValues();
if (PHINumValues != ValueMapping.size())
return false;
// Scan the phi to see if it matches.
for (unsigned i = 0, e = PHINumValues; i != e; ++i)
if (ValueMapping[PHI->getIncomingBlock(i)] !=
PHI->getIncomingValue(i)) {
return false;
}
return true;
}
/// Remove dead functions that are not included in DNR (Do Not Remove) list.
bool Inliner::removeDeadFunctions(CallGraph &CG, bool AlwaysInlineOnly) {
SmallVector<CallGraphNode*, 16> FunctionsToRemove;
SmallVector<CallGraphNode *, 16> DeadFunctionsInComdats;
SmallDenseMap<const Comdat *, int, 16> ComdatEntriesAlive;
auto RemoveCGN = [&](CallGraphNode *CGN) {
// Remove any call graph edges from the function to its callees.
CGN->removeAllCalledFunctions();
// Remove any edges from the external node to the function's call graph
// node. These edges might have been made irrelegant due to
// optimization of the program.
CG.getExternalCallingNode()->removeAnyCallEdgeTo(CGN);
// Removing the node for callee from the call graph and delete it.
FunctionsToRemove.push_back(CGN);
};
// Scan for all of the functions, looking for ones that should now be removed
// from the program. Insert the dead ones in the FunctionsToRemove set.
for (CallGraph::iterator I = CG.begin(), E = CG.end(); I != E; ++I) {
CallGraphNode *CGN = I->second;
Function *F = CGN->getFunction();
if (!F || F->isDeclaration())
continue;
// Handle the case when this function is called and we only want to care
// about always-inline functions. This is a bit of a hack to share code
// between here and the InlineAlways pass.
if (AlwaysInlineOnly && !F->hasFnAttribute(Attribute::AlwaysInline))
continue;
// If the only remaining users of the function are dead constants, remove
// them.
F->removeDeadConstantUsers();
if (!F->isDefTriviallyDead())
continue;
// It is unsafe to drop a function with discardable linkage from a COMDAT
// without also dropping the other members of the COMDAT.
// The inliner doesn't visit non-function entities which are in COMDAT
// groups so it is unsafe to do so *unless* the linkage is local.
if (!F->hasLocalLinkage()) {
if (const Comdat *C = F->getComdat()) {
--ComdatEntriesAlive[C];
DeadFunctionsInComdats.push_back(CGN);
continue;
}
}
RemoveCGN(CGN);
}
if (!DeadFunctionsInComdats.empty()) {
// Count up all the entities in COMDAT groups
auto ComdatGroupReferenced = [&](const Comdat *C) {
auto I = ComdatEntriesAlive.find(C);
if (I != ComdatEntriesAlive.end())
++(I->getSecond());
};
for (const Function &F : CG.getModule())
if (const Comdat *C = F.getComdat())
ComdatGroupReferenced(C);
for (const GlobalVariable &GV : CG.getModule().globals())
if (const Comdat *C = GV.getComdat())
ComdatGroupReferenced(C);
for (const GlobalAlias &GA : CG.getModule().aliases())
if (const Comdat *C = GA.getComdat())
ComdatGroupReferenced(C);
for (CallGraphNode *CGN : DeadFunctionsInComdats) {
Function *F = CGN->getFunction();
const Comdat *C = F->getComdat();
int NumAlive = ComdatEntriesAlive[C];
// We can remove functions in a COMDAT group if the entire group is dead.
assert(NumAlive >= 0);
if (NumAlive > 0)
continue;
RemoveCGN(CGN);
}
}
if (FunctionsToRemove.empty())
return false;
// Now that we know which functions to delete, do so. We didn't want to do
// this inline, because that would invalidate our CallGraph::iterator
// objects. :(
//
// Note that it doesn't matter that we are iterating over a non-stable order
// here to do this, it doesn't matter which order the functions are deleted
// in.
array_pod_sort(FunctionsToRemove.begin(), FunctionsToRemove.end());
FunctionsToRemove.erase(std::unique(FunctionsToRemove.begin(),
FunctionsToRemove.end()),
FunctionsToRemove.end());
for (SmallVectorImpl<CallGraphNode *>::iterator I = FunctionsToRemove.begin(),
E = FunctionsToRemove.end();
I != E; ++I) {
delete CG.removeFunctionFromModule(*I);
++NumDeleted;
}
return true;
}
/// Unroll the given loop by Count. The loop must be in LCSSA form. Returns true
/// if unrolling was successful, or false if the loop was unmodified. Unrolling
/// can only fail when the loop's latch block is not terminated by a conditional
/// branch instruction. However, if the trip count (and multiple) are not known,
/// loop unrolling will mostly produce more code that is no faster.
///
/// TripCount is generally defined as the number of times the loop header
/// executes. UnrollLoop relaxes the definition to permit early exits: here
/// TripCount is the iteration on which control exits LatchBlock if no early
/// exits were taken. Note that UnrollLoop assumes that the loop counter test
/// terminates LatchBlock in order to remove unnecesssary instances of the
/// test. In other words, control may exit the loop prior to TripCount
/// iterations via an early branch, but control may not exit the loop from the
/// LatchBlock's terminator prior to TripCount iterations.
///
/// Similarly, TripMultiple divides the number of times that the LatchBlock may
/// execute without exiting the loop.
///
/// The LoopInfo Analysis that is passed will be kept consistent.
///
/// If a LoopPassManager is passed in, and the loop is fully removed, it will be
/// removed from the LoopPassManager as well. LPM can also be NULL.
///
/// This utility preserves LoopInfo. If DominatorTree or ScalarEvolution are
/// available from the Pass it must also preserve those analyses.
bool llvm::UnrollLoop(Loop *L, unsigned Count, unsigned TripCount,
bool AllowRuntime, unsigned TripMultiple, LoopInfo *LI,
Pass *PP, LPPassManager *LPM, AssumptionCache *AC) {
BasicBlock *Preheader = L->getLoopPreheader();
if (!Preheader) {
DEBUG(dbgs() << " Can't unroll; loop preheader-insertion failed.\n");
return false;
}
BasicBlock *LatchBlock = L->getLoopLatch();
if (!LatchBlock) {
DEBUG(dbgs() << " Can't unroll; loop exit-block-insertion failed.\n");
return false;
}
// Loops with indirectbr cannot be cloned.
if (!L->isSafeToClone()) {
DEBUG(dbgs() << " Can't unroll; Loop body cannot be cloned.\n");
return false;
}
BasicBlock *Header = L->getHeader();
BranchInst *BI = dyn_cast<BranchInst>(LatchBlock->getTerminator());
if (!BI || BI->isUnconditional()) {
// The loop-rotate pass can be helpful to avoid this in many cases.
DEBUG(dbgs() <<
" Can't unroll; loop not terminated by a conditional branch.\n");
return false;
}
if (Header->hasAddressTaken()) {
// The loop-rotate pass can be helpful to avoid this in many cases.
DEBUG(dbgs() <<
" Won't unroll loop: address of header block is taken.\n");
return false;
}
if (TripCount != 0)
DEBUG(dbgs() << " Trip Count = " << TripCount << "\n");
if (TripMultiple != 1)
DEBUG(dbgs() << " Trip Multiple = " << TripMultiple << "\n");
// Effectively "DCE" unrolled iterations that are beyond the tripcount
// and will never be executed.
if (TripCount != 0 && Count > TripCount)
Count = TripCount;
// Don't enter the unroll code if there is nothing to do. This way we don't
// need to support "partial unrolling by 1".
if (TripCount == 0 && Count < 2)
return false;
assert(Count > 0);
assert(TripMultiple > 0);
assert(TripCount == 0 || TripCount % TripMultiple == 0);
// Are we eliminating the loop control altogether?
bool CompletelyUnroll = Count == TripCount;
// We assume a run-time trip count if the compiler cannot
// figure out the loop trip count and the unroll-runtime
// flag is specified.
bool RuntimeTripCount = (TripCount == 0 && Count > 0 && AllowRuntime);
if (RuntimeTripCount && !UnrollRuntimeLoopProlog(L, Count, LI, LPM))
return false;
// Notify ScalarEvolution that the loop will be substantially changed,
// if not outright eliminated.
ScalarEvolution *SE =
PP ? PP->getAnalysisIfAvailable<ScalarEvolution>() : nullptr;
if (SE)
SE->forgetLoop(L);
// If we know the trip count, we know the multiple...
unsigned BreakoutTrip = 0;
if (TripCount != 0) {
BreakoutTrip = TripCount % Count;
TripMultiple = 0;
} else {
// Figure out what multiple to use.
BreakoutTrip = TripMultiple =
(unsigned)GreatestCommonDivisor64(Count, TripMultiple);
}
// Report the unrolling decision.
DebugLoc LoopLoc = L->getStartLoc();
Function *F = Header->getParent();
LLVMContext &Ctx = F->getContext();
if (CompletelyUnroll) {
DEBUG(dbgs() << "COMPLETELY UNROLLING loop %" << Header->getName()
<< " with trip count " << TripCount << "!\n");
emitOptimizationRemark(Ctx, DEBUG_TYPE, *F, LoopLoc,
Twine("completely unrolled loop with ") +
Twine(TripCount) + " iterations");
} else {
auto EmitDiag = [&](const Twine &T) {
emitOptimizationRemark(Ctx, DEBUG_TYPE, *F, LoopLoc,
"unrolled loop by a factor of " + Twine(Count) +
T);
};
DEBUG(dbgs() << "UNROLLING loop %" << Header->getName()
<< " by " << Count);
if (TripMultiple == 0 || BreakoutTrip != TripMultiple) {
DEBUG(dbgs() << " with a breakout at trip " << BreakoutTrip);
EmitDiag(" with a breakout at trip " + Twine(BreakoutTrip));
} else if (TripMultiple != 1) {
DEBUG(dbgs() << " with " << TripMultiple << " trips per branch");
EmitDiag(" with " + Twine(TripMultiple) + " trips per branch");
} else if (RuntimeTripCount) {
DEBUG(dbgs() << " with run-time trip count");
EmitDiag(" with run-time trip count");
}
DEBUG(dbgs() << "!\n");
}
bool ContinueOnTrue = L->contains(BI->getSuccessor(0));
BasicBlock *LoopExit = BI->getSuccessor(ContinueOnTrue);
// For the first iteration of the loop, we should use the precloned values for
// PHI nodes. Insert associations now.
ValueToValueMapTy LastValueMap;
std::vector<PHINode*> OrigPHINode;
for (BasicBlock::iterator I = Header->begin(); isa<PHINode>(I); ++I) {
OrigPHINode.push_back(cast<PHINode>(I));
}
std::vector<BasicBlock*> Headers;
std::vector<BasicBlock*> Latches;
Headers.push_back(Header);
Latches.push_back(LatchBlock);
// The current on-the-fly SSA update requires blocks to be processed in
// reverse postorder so that LastValueMap contains the correct value at each
// exit.
LoopBlocksDFS DFS(L);
DFS.perform(LI);
// Stash the DFS iterators before adding blocks to the loop.
LoopBlocksDFS::RPOIterator BlockBegin = DFS.beginRPO();
LoopBlocksDFS::RPOIterator BlockEnd = DFS.endRPO();
for (unsigned It = 1; It != Count; ++It) {
std::vector<BasicBlock*> NewBlocks;
SmallDenseMap<const Loop *, Loop *, 4> NewLoops;
NewLoops[L] = L;
for (LoopBlocksDFS::RPOIterator BB = BlockBegin; BB != BlockEnd; ++BB) {
ValueToValueMapTy VMap;
BasicBlock *New = CloneBasicBlock(*BB, VMap, "." + Twine(It));
Header->getParent()->getBasicBlockList().push_back(New);
// Tell LI about New.
if (*BB == Header) {
assert(LI->getLoopFor(*BB) == L && "Header should not be in a sub-loop");
L->addBasicBlockToLoop(New, *LI);
} else {
// Figure out which loop New is in.
const Loop *OldLoop = LI->getLoopFor(*BB);
assert(OldLoop && "Should (at least) be in the loop being unrolled!");
Loop *&NewLoop = NewLoops[OldLoop];
if (!NewLoop) {
// Found a new sub-loop.
assert(*BB == OldLoop->getHeader() &&
"Header should be first in RPO");
Loop *NewLoopParent = NewLoops.lookup(OldLoop->getParentLoop());
assert(NewLoopParent &&
"Expected parent loop before sub-loop in RPO");
NewLoop = new Loop;
NewLoopParent->addChildLoop(NewLoop);
// Forget the old loop, since its inputs may have changed.
if (SE)
SE->forgetLoop(OldLoop);
}
NewLoop->addBasicBlockToLoop(New, *LI);
}
if (*BB == Header)
// Loop over all of the PHI nodes in the block, changing them to use
// the incoming values from the previous block.
for (unsigned i = 0, e = OrigPHINode.size(); i != e; ++i) {
PHINode *NewPHI = cast<PHINode>(VMap[OrigPHINode[i]]);
Value *InVal = NewPHI->getIncomingValueForBlock(LatchBlock);
if (Instruction *InValI = dyn_cast<Instruction>(InVal))
if (It > 1 && L->contains(InValI))
InVal = LastValueMap[InValI];
VMap[OrigPHINode[i]] = InVal;
New->getInstList().erase(NewPHI);
}
// Update our running map of newest clones
LastValueMap[*BB] = New;
for (ValueToValueMapTy::iterator VI = VMap.begin(), VE = VMap.end();
VI != VE; ++VI)
LastValueMap[VI->first] = VI->second;
// Add phi entries for newly created values to all exit blocks.
for (succ_iterator SI = succ_begin(*BB), SE = succ_end(*BB);
SI != SE; ++SI) {
if (L->contains(*SI))
continue;
for (BasicBlock::iterator BBI = (*SI)->begin();
PHINode *phi = dyn_cast<PHINode>(BBI); ++BBI) {
Value *Incoming = phi->getIncomingValueForBlock(*BB);
ValueToValueMapTy::iterator It = LastValueMap.find(Incoming);
if (It != LastValueMap.end())
Incoming = It->second;
phi->addIncoming(Incoming, New);
}
}
// Keep track of new headers and latches as we create them, so that
// we can insert the proper branches later.
if (*BB == Header)
Headers.push_back(New);
if (*BB == LatchBlock)
Latches.push_back(New);
NewBlocks.push_back(New);
}
// Remap all instructions in the most recent iteration
for (unsigned i = 0; i < NewBlocks.size(); ++i)
for (BasicBlock::iterator I = NewBlocks[i]->begin(),
E = NewBlocks[i]->end(); I != E; ++I)
::RemapInstruction(I, LastValueMap);
}
// Loop over the PHI nodes in the original block, setting incoming values.
for (unsigned i = 0, e = OrigPHINode.size(); i != e; ++i) {
PHINode *PN = OrigPHINode[i];
if (CompletelyUnroll) {
PN->replaceAllUsesWith(PN->getIncomingValueForBlock(Preheader));
Header->getInstList().erase(PN);
}
else if (Count > 1) {
Value *InVal = PN->removeIncomingValue(LatchBlock, false);
// If this value was defined in the loop, take the value defined by the
// last iteration of the loop.
if (Instruction *InValI = dyn_cast<Instruction>(InVal)) {
if (L->contains(InValI))
InVal = LastValueMap[InVal];
}
assert(Latches.back() == LastValueMap[LatchBlock] && "bad last latch");
PN->addIncoming(InVal, Latches.back());
}
}
// Now that all the basic blocks for the unrolled iterations are in place,
// set up the branches to connect them.
for (unsigned i = 0, e = Latches.size(); i != e; ++i) {
// The original branch was replicated in each unrolled iteration.
BranchInst *Term = cast<BranchInst>(Latches[i]->getTerminator());
// The branch destination.
unsigned j = (i + 1) % e;
BasicBlock *Dest = Headers[j];
bool NeedConditional = true;
if (RuntimeTripCount && j != 0) {
NeedConditional = false;
}
// For a complete unroll, make the last iteration end with a branch
// to the exit block.
if (CompletelyUnroll && j == 0) {
Dest = LoopExit;
NeedConditional = false;
}
// If we know the trip count or a multiple of it, we can safely use an
// unconditional branch for some iterations.
if (j != BreakoutTrip && (TripMultiple == 0 || j % TripMultiple != 0)) {
NeedConditional = false;
}
if (NeedConditional) {
// Update the conditional branch's successor for the following
// iteration.
Term->setSuccessor(!ContinueOnTrue, Dest);
} else {
// Remove phi operands at this loop exit
if (Dest != LoopExit) {
BasicBlock *BB = Latches[i];
for (succ_iterator SI = succ_begin(BB), SE = succ_end(BB);
SI != SE; ++SI) {
if (*SI == Headers[i])
continue;
for (BasicBlock::iterator BBI = (*SI)->begin();
PHINode *Phi = dyn_cast<PHINode>(BBI); ++BBI) {
Phi->removeIncomingValue(BB, false);
}
}
}
// Replace the conditional branch with an unconditional one.
BranchInst::Create(Dest, Term);
Term->eraseFromParent();
}
}
// Merge adjacent basic blocks, if possible.
SmallPtrSet<Loop *, 4> ForgottenLoops;
for (unsigned i = 0, e = Latches.size(); i != e; ++i) {
BranchInst *Term = cast<BranchInst>(Latches[i]->getTerminator());
if (Term->isUnconditional()) {
BasicBlock *Dest = Term->getSuccessor(0);
if (BasicBlock *Fold = FoldBlockIntoPredecessor(Dest, LI, LPM,
ForgottenLoops))
std::replace(Latches.begin(), Latches.end(), Dest, Fold);
}
}
// FIXME: We could register any cloned assumptions instead of clearing the
// whole function's cache.
AC->clear();
DominatorTree *DT = nullptr;
if (PP) {
// FIXME: Reconstruct dom info, because it is not preserved properly.
// Incrementally updating domtree after loop unrolling would be easy.
if (DominatorTreeWrapperPass *DTWP =
PP->getAnalysisIfAvailable<DominatorTreeWrapperPass>()) {
DT = &DTWP->getDomTree();
DT->recalculate(*L->getHeader()->getParent());
}
// Simplify any new induction variables in the partially unrolled loop.
if (SE && !CompletelyUnroll) {
SmallVector<WeakVH, 16> DeadInsts;
simplifyLoopIVs(L, SE, LPM, DeadInsts);
// Aggressively clean up dead instructions that simplifyLoopIVs already
// identified. Any remaining should be cleaned up below.
while (!DeadInsts.empty())
if (Instruction *Inst =
dyn_cast_or_null<Instruction>(&*DeadInsts.pop_back_val()))
RecursivelyDeleteTriviallyDeadInstructions(Inst);
}
}
// At this point, the code is well formed. We now do a quick sweep over the
// inserted code, doing constant propagation and dead code elimination as we
// go.
const std::vector<BasicBlock*> &NewLoopBlocks = L->getBlocks();
for (std::vector<BasicBlock*>::const_iterator BB = NewLoopBlocks.begin(),
BBE = NewLoopBlocks.end(); BB != BBE; ++BB)
for (BasicBlock::iterator I = (*BB)->begin(), E = (*BB)->end(); I != E; ) {
Instruction *Inst = I++;
if (isInstructionTriviallyDead(Inst))
(*BB)->getInstList().erase(Inst);
else if (Value *V = SimplifyInstruction(Inst))
if (LI->replacementPreservesLCSSAForm(Inst, V)) {
Inst->replaceAllUsesWith(V);
(*BB)->getInstList().erase(Inst);
}
}
NumCompletelyUnrolled += CompletelyUnroll;
++NumUnrolled;
Loop *OuterL = L->getParentLoop();
// Remove the loop from the LoopPassManager if it's completely removed.
if (CompletelyUnroll && LPM != nullptr)
LPM->deleteLoopFromQueue(L);
// If we have a pass and a DominatorTree we should re-simplify impacted loops
// to ensure subsequent analyses can rely on this form. We want to simplify
// at least one layer outside of the loop that was unrolled so that any
// changes to the parent loop exposed by the unrolling are considered.
if (PP && DT) {
if (!OuterL && !CompletelyUnroll)
OuterL = L;
if (OuterL) {
DataLayoutPass *DLP = PP->getAnalysisIfAvailable<DataLayoutPass>();
const DataLayout *DL = DLP ? &DLP->getDataLayout() : nullptr;
simplifyLoop(OuterL, DT, LI, PP, /*AliasAnalysis*/ nullptr, SE, DL, AC);
// LCSSA must be performed on the outermost affected loop. The unrolled
// loop's last loop latch is guaranteed to be in the outermost loop after
// deleteLoopFromQueue updates LoopInfo.
Loop *LatchLoop = LI->getLoopFor(Latches.back());
if (!OuterL->contains(LatchLoop))
while (OuterL->getParentLoop() != LatchLoop)
OuterL = OuterL->getParentLoop();
formLCSSARecursively(*OuterL, *DT, LI, SE);
}
}
return true;
}
// Sinks \p I from the loop \p L's preheader to its uses. Returns true if
// sinking is successful.
// \p LoopBlockNumber is used to sort the insertion blocks to ensure
// determinism.
static bool sinkInstruction(Loop &L, Instruction &I,
const SmallVectorImpl<BasicBlock *> &ColdLoopBBs,
const SmallDenseMap<BasicBlock *, int, 16> &LoopBlockNumber,
LoopInfo &LI, DominatorTree &DT,
BlockFrequencyInfo &BFI) {
// Compute the set of blocks in loop L which contain a use of I.
SmallPtrSet<BasicBlock *, 2> BBs;
for (auto &U : I.uses()) {
Instruction *UI = cast<Instruction>(U.getUser());
// We cannot sink I to PHI-uses.
if (dyn_cast<PHINode>(UI))
return false;
// We cannot sink I if it has uses outside of the loop.
if (!L.contains(LI.getLoopFor(UI->getParent())))
return false;
BBs.insert(UI->getParent());
}
// findBBsToSinkInto is O(BBs.size() * ColdLoopBBs.size()). We cap the max
// BBs.size() to avoid expensive computation.
// FIXME: Handle code size growth for min_size and opt_size.
if (BBs.size() > MaxNumberOfUseBBsForSinking)
return false;
// Find the set of BBs that we should insert a copy of I.
SmallPtrSet<BasicBlock *, 2> BBsToSinkInto =
findBBsToSinkInto(L, BBs, ColdLoopBBs, DT, BFI);
if (BBsToSinkInto.empty())
return false;
// Copy the final BBs into a vector and sort them using the total ordering
// of the loop block numbers as iterating the set doesn't give a useful
// order. No need to stable sort as the block numbers are a total ordering.
SmallVector<BasicBlock *, 2> SortedBBsToSinkInto;
SortedBBsToSinkInto.insert(SortedBBsToSinkInto.begin(), BBsToSinkInto.begin(),
BBsToSinkInto.end());
std::sort(SortedBBsToSinkInto.begin(), SortedBBsToSinkInto.end(),
[&](BasicBlock *A, BasicBlock *B) {
return *LoopBlockNumber.find(A) < *LoopBlockNumber.find(B);
});
BasicBlock *MoveBB = *SortedBBsToSinkInto.begin();
// FIXME: Optimize the efficiency for cloned value replacement. The current
// implementation is O(SortedBBsToSinkInto.size() * I.num_uses()).
for (BasicBlock *N : SortedBBsToSinkInto) {
if (N == MoveBB)
continue;
// Clone I and replace its uses.
Instruction *IC = I.clone();
IC->setName(I.getName());
IC->insertBefore(&*N->getFirstInsertionPt());
// Replaces uses of I with IC in N
for (Value::use_iterator UI = I.use_begin(), UE = I.use_end(); UI != UE;) {
Use &U = *UI++;
auto *I = cast<Instruction>(U.getUser());
if (I->getParent() == N)
U.set(IC);
}
// Replaces uses of I with IC in blocks dominated by N
replaceDominatedUsesWith(&I, IC, DT, N);
DEBUG(dbgs() << "Sinking a clone of " << I << " To: " << N->getName()
<< '\n');
NumLoopSunkCloned++;
}
DEBUG(dbgs() << "Sinking " << I << " To: " << MoveBB->getName() << '\n');
NumLoopSunk++;
I.moveBefore(&*MoveBB->getFirstInsertionPt());
return true;
}
/// For every instruction from the worklist, check to see if it has any uses
/// that are outside the current loop. If so, insert LCSSA PHI nodes and
/// rewrite the uses.
bool llvm::formLCSSAForInstructions(SmallVectorImpl<Instruction *> &Worklist,
DominatorTree &DT, LoopInfo &LI) {
SmallVector<Use *, 16> UsesToRewrite;
SmallSetVector<PHINode *, 16> PHIsToRemove;
PredIteratorCache PredCache;
bool Changed = false;
// Cache the Loop ExitBlocks across this loop. We expect to get a lot of
// instructions within the same loops, computing the exit blocks is
// expensive, and we're not mutating the loop structure.
SmallDenseMap<Loop*, SmallVector<BasicBlock *,1>> LoopExitBlocks;
while (!Worklist.empty()) {
UsesToRewrite.clear();
Instruction *I = Worklist.pop_back_val();
BasicBlock *InstBB = I->getParent();
Loop *L = LI.getLoopFor(InstBB);
if (!LoopExitBlocks.count(L))
L->getExitBlocks(LoopExitBlocks[L]);
assert(LoopExitBlocks.count(L));
const SmallVectorImpl<BasicBlock *> &ExitBlocks = LoopExitBlocks[L];
if (ExitBlocks.empty())
continue;
// Tokens cannot be used in PHI nodes, so we skip over them.
// We can run into tokens which are live out of a loop with catchswitch
// instructions in Windows EH if the catchswitch has one catchpad which
// is inside the loop and another which is not.
if (I->getType()->isTokenTy())
continue;
for (Use &U : I->uses()) {
Instruction *User = cast<Instruction>(U.getUser());
BasicBlock *UserBB = User->getParent();
if (PHINode *PN = dyn_cast<PHINode>(User))
UserBB = PN->getIncomingBlock(U);
if (InstBB != UserBB && !L->contains(UserBB))
UsesToRewrite.push_back(&U);
}
// If there are no uses outside the loop, exit with no change.
if (UsesToRewrite.empty())
continue;
++NumLCSSA; // We are applying the transformation
// Invoke instructions are special in that their result value is not
// available along their unwind edge. The code below tests to see whether
// DomBB dominates the value, so adjust DomBB to the normal destination
// block, which is effectively where the value is first usable.
BasicBlock *DomBB = InstBB;
if (InvokeInst *Inv = dyn_cast<InvokeInst>(I))
DomBB = Inv->getNormalDest();
DomTreeNode *DomNode = DT.getNode(DomBB);
SmallVector<PHINode *, 16> AddedPHIs;
SmallVector<PHINode *, 8> PostProcessPHIs;
SmallVector<PHINode *, 4> InsertedPHIs;
SSAUpdater SSAUpdate(&InsertedPHIs);
SSAUpdate.Initialize(I->getType(), I->getName());
// Insert the LCSSA phi's into all of the exit blocks dominated by the
// value, and add them to the Phi's map.
for (BasicBlock *ExitBB : ExitBlocks) {
if (!DT.dominates(DomNode, DT.getNode(ExitBB)))
continue;
// If we already inserted something for this BB, don't reprocess it.
if (SSAUpdate.HasValueForBlock(ExitBB))
continue;
PHINode *PN = PHINode::Create(I->getType(), PredCache.size(ExitBB),
I->getName() + ".lcssa", &ExitBB->front());
// Add inputs from inside the loop for this PHI.
for (BasicBlock *Pred : PredCache.get(ExitBB)) {
PN->addIncoming(I, Pred);
// If the exit block has a predecessor not within the loop, arrange for
// the incoming value use corresponding to that predecessor to be
// rewritten in terms of a different LCSSA PHI.
if (!L->contains(Pred))
UsesToRewrite.push_back(
&PN->getOperandUse(PN->getOperandNumForIncomingValue(
PN->getNumIncomingValues() - 1)));
}
AddedPHIs.push_back(PN);
// Remember that this phi makes the value alive in this block.
SSAUpdate.AddAvailableValue(ExitBB, PN);
// LoopSimplify might fail to simplify some loops (e.g. when indirect
// branches are involved). In such situations, it might happen that an
// exit for Loop L1 is the header of a disjoint Loop L2. Thus, when we
// create PHIs in such an exit block, we are also inserting PHIs into L2's
// header. This could break LCSSA form for L2 because these inserted PHIs
// can also have uses outside of L2. Remember all PHIs in such situation
// as to revisit than later on. FIXME: Remove this if indirectbr support
// into LoopSimplify gets improved.
if (auto *OtherLoop = LI.getLoopFor(ExitBB))
if (!L->contains(OtherLoop))
PostProcessPHIs.push_back(PN);
}
// Rewrite all uses outside the loop in terms of the new PHIs we just
// inserted.
for (Use *UseToRewrite : UsesToRewrite) {
// If this use is in an exit block, rewrite to use the newly inserted PHI.
// This is required for correctness because SSAUpdate doesn't handle uses
// in the same block. It assumes the PHI we inserted is at the end of the
// block.
Instruction *User = cast<Instruction>(UseToRewrite->getUser());
BasicBlock *UserBB = User->getParent();
if (PHINode *PN = dyn_cast<PHINode>(User))
UserBB = PN->getIncomingBlock(*UseToRewrite);
if (isa<PHINode>(UserBB->begin()) && isExitBlock(UserBB, ExitBlocks)) {
// Tell the VHs that the uses changed. This updates SCEV's caches.
if (UseToRewrite->get()->hasValueHandle())
ValueHandleBase::ValueIsRAUWd(*UseToRewrite, &UserBB->front());
UseToRewrite->set(&UserBB->front());
continue;
}
// Otherwise, do full PHI insertion.
SSAUpdate.RewriteUse(*UseToRewrite);
}
// SSAUpdater might have inserted phi-nodes inside other loops. We'll need
// to post-process them to keep LCSSA form.
for (PHINode *InsertedPN : InsertedPHIs) {
if (auto *OtherLoop = LI.getLoopFor(InsertedPN->getParent()))
if (!L->contains(OtherLoop))
PostProcessPHIs.push_back(InsertedPN);
}
// Post process PHI instructions that were inserted into another disjoint
// loop and update their exits properly.
for (auto *PostProcessPN : PostProcessPHIs) {
if (PostProcessPN->use_empty())
continue;
// Reprocess each PHI instruction.
Worklist.push_back(PostProcessPN);
}
// Keep track of PHI nodes that we want to remove because they did not have
// any uses rewritten.
for (PHINode *PN : AddedPHIs)
if (PN->use_empty())
PHIsToRemove.insert(PN);
Changed = true;
}
// Remove PHI nodes that did not have any uses rewritten.
for (PHINode *PN : PHIsToRemove) {
assert (PN->use_empty() && "Trying to remove a phi with uses.");
PN->eraseFromParent();
}
return Changed;
}
// Create output section objects and add them to OutputSections.
template <class ELFT> void Writer<ELFT>::createSections() {
// .interp needs to be on the first page in the output file.
if (needsInterpSection())
OutputSections.push_back(Out<ELFT>::Interp);
SmallDenseMap<SectionKey<ELFT::Is64Bits>, OutputSectionBase<ELFT> *> Map;
std::vector<OutputSectionBase<ELFT> *> RegularSections;
for (const std::unique_ptr<ObjectFile<ELFT>> &F : Symtab.getObjectFiles()) {
for (InputSectionBase<ELFT> *C : F->getSections()) {
if (isDiscarded(C))
continue;
const Elf_Shdr *H = C->getSectionHdr();
uintX_t OutFlags = H->sh_flags & ~SHF_GROUP;
// For SHF_MERGE we create different output sections for each sh_entsize.
// This makes each output section simple and keeps a single level
// mapping from input to output.
typename InputSectionBase<ELFT>::Kind K = C->SectionKind;
uintX_t EntSize = K != InputSectionBase<ELFT>::Merge ? 0 : H->sh_entsize;
uint32_t OutType = H->sh_type;
if (OutType == SHT_PROGBITS && C->getSectionName() == ".eh_frame" &&
Config->EMachine == EM_X86_64)
OutType = SHT_X86_64_UNWIND;
SectionKey<ELFT::Is64Bits> Key{getOutputSectionName(C->getSectionName()),
OutType, OutFlags, EntSize};
OutputSectionBase<ELFT> *&Sec = Map[Key];
if (!Sec) {
switch (K) {
case InputSectionBase<ELFT>::Regular:
Sec = new (SecAlloc.Allocate())
OutputSection<ELFT>(Key.Name, Key.Type, Key.Flags);
break;
case InputSectionBase<ELFT>::EHFrame:
Sec = new (EHSecAlloc.Allocate())
EHOutputSection<ELFT>(Key.Name, Key.Type, Key.Flags);
break;
case InputSectionBase<ELFT>::Merge:
Sec = new (MSecAlloc.Allocate())
MergeOutputSection<ELFT>(Key.Name, Key.Type, Key.Flags);
break;
}
OutputSections.push_back(Sec);
RegularSections.push_back(Sec);
}
switch (K) {
case InputSectionBase<ELFT>::Regular:
static_cast<OutputSection<ELFT> *>(Sec)
->addSection(cast<InputSection<ELFT>>(C));
break;
case InputSectionBase<ELFT>::EHFrame:
static_cast<EHOutputSection<ELFT> *>(Sec)
->addSection(cast<EHInputSection<ELFT>>(C));
break;
case InputSectionBase<ELFT>::Merge:
static_cast<MergeOutputSection<ELFT> *>(Sec)
->addSection(cast<MergeInputSection<ELFT>>(C));
break;
}
}
}
Out<ELFT>::Bss = static_cast<OutputSection<ELFT> *>(
Map[{".bss", SHT_NOBITS, SHF_ALLOC | SHF_WRITE, 0}]);
Out<ELFT>::Dynamic->PreInitArraySec = Map.lookup(
{".preinit_array", SHT_PREINIT_ARRAY, SHF_WRITE | SHF_ALLOC, 0});
Out<ELFT>::Dynamic->InitArraySec =
Map.lookup({".init_array", SHT_INIT_ARRAY, SHF_WRITE | SHF_ALLOC, 0});
Out<ELFT>::Dynamic->FiniArraySec =
Map.lookup({".fini_array", SHT_FINI_ARRAY, SHF_WRITE | SHF_ALLOC, 0});
auto AddStartEnd = [&](StringRef Start, StringRef End,
OutputSectionBase<ELFT> *OS) {
if (OS) {
Symtab.addSyntheticSym(Start, *OS, 0);
Symtab.addSyntheticSym(End, *OS, OS->getSize());
} else {
Symtab.addIgnoredSym(Start);
Symtab.addIgnoredSym(End);
}
};
AddStartEnd("__preinit_array_start", "__preinit_array_end",
Out<ELFT>::Dynamic->PreInitArraySec);
AddStartEnd("__init_array_start", "__init_array_end",
Out<ELFT>::Dynamic->InitArraySec);
AddStartEnd("__fini_array_start", "__fini_array_end",
Out<ELFT>::Dynamic->FiniArraySec);
for (OutputSectionBase<ELFT> *Sec : RegularSections)
addStartStopSymbols(Sec);
// __tls_get_addr is defined by the dynamic linker for dynamic ELFs. For
// static linking the linker is required to optimize away any references to
// __tls_get_addr, so it's not defined anywhere. Create a hidden definition
// to avoid the undefined symbol error.
if (!isOutputDynamic())
Symtab.addIgnoredSym("__tls_get_addr");
// If the "_end" symbol is referenced, it is expected to point to the address
// right after the data segment. Usually, this symbol points to the end
// of .bss section or to the end of .data section if .bss section is absent.
// The order of the sections can be affected by linker script,
// so it is hard to predict which section will be the last one.
// So, if this symbol is referenced, we just add the placeholder here
// and update its value later.
if (Symtab.find("_end"))
Symtab.addAbsoluteSym("_end", DefinedAbsolute<ELFT>::End);
// If there is an undefined symbol "end", we should initialize it
// with the same value as "_end". In any other case it should stay intact,
// because it is an allowable name for a user symbol.
if (SymbolBody *B = Symtab.find("end"))
if (B->isUndefined())
Symtab.addAbsoluteSym("end", DefinedAbsolute<ELFT>::End);
// Scan relocations. This must be done after every symbol is declared so that
// we can correctly decide if a dynamic relocation is needed.
for (const std::unique_ptr<ObjectFile<ELFT>> &F : Symtab.getObjectFiles()) {
for (InputSectionBase<ELFT> *C : F->getSections()) {
if (isDiscarded(C))
continue;
if (auto *S = dyn_cast<InputSection<ELFT>>(C))
scanRelocs(*S);
else if (auto *S = dyn_cast<EHInputSection<ELFT>>(C))
if (S->RelocSection)
scanRelocs(*S, *S->RelocSection);
}
}
std::vector<DefinedCommon<ELFT> *> CommonSymbols;
std::vector<SharedSymbol<ELFT> *> SharedCopySymbols;
for (auto &P : Symtab.getSymbols()) {
SymbolBody *Body = P.second->Body;
if (auto *U = dyn_cast<Undefined<ELFT>>(Body))
if (!U->isWeak() && !U->canKeepUndefined())
reportUndefined<ELFT>(Symtab, *Body);
if (auto *C = dyn_cast<DefinedCommon<ELFT>>(Body))
CommonSymbols.push_back(C);
if (auto *SC = dyn_cast<SharedSymbol<ELFT>>(Body))
if (SC->needsCopy())
SharedCopySymbols.push_back(SC);
if (!includeInSymtab<ELFT>(*Body))
continue;
if (Out<ELFT>::SymTab)
Out<ELFT>::SymTab->addSymbol(Body);
if (isOutputDynamic() && includeInDynamicSymtab(*Body))
Out<ELFT>::DynSymTab->addSymbol(Body);
}
addCommonSymbols(CommonSymbols);
addSharedCopySymbols(SharedCopySymbols);
// This order is not the same as the final output order
// because we sort the sections using their attributes below.
if (Out<ELFT>::SymTab)
OutputSections.push_back(Out<ELFT>::SymTab);
OutputSections.push_back(Out<ELFT>::ShStrTab);
if (Out<ELFT>::StrTab)
OutputSections.push_back(Out<ELFT>::StrTab);
if (isOutputDynamic()) {
OutputSections.push_back(Out<ELFT>::DynSymTab);
if (Out<ELFT>::GnuHashTab)
OutputSections.push_back(Out<ELFT>::GnuHashTab);
if (Out<ELFT>::HashTab)
OutputSections.push_back(Out<ELFT>::HashTab);
OutputSections.push_back(Out<ELFT>::Dynamic);
OutputSections.push_back(Out<ELFT>::DynStrTab);
if (Out<ELFT>::RelaDyn->hasRelocs())
OutputSections.push_back(Out<ELFT>::RelaDyn);
if (Out<ELFT>::RelaPlt && Out<ELFT>::RelaPlt->hasRelocs())
OutputSections.push_back(Out<ELFT>::RelaPlt);
// This is a MIPS specific section to hold a space within the data segment
// of executable file which is pointed to by the DT_MIPS_RLD_MAP entry.
// See "Dynamic section" in Chapter 5 in the following document:
// ftp://www.linux-mips.org/pub/linux/mips/doc/ABI/mipsabi.pdf
if (Config->EMachine == EM_MIPS && !Config->Shared) {
Out<ELFT>::MipsRldMap = new (SecAlloc.Allocate())
OutputSection<ELFT>(".rld_map", SHT_PROGBITS, SHF_ALLOC | SHF_WRITE);
Out<ELFT>::MipsRldMap->setSize(ELFT::Is64Bits ? 8 : 4);
Out<ELFT>::MipsRldMap->updateAlign(ELFT::Is64Bits ? 8 : 4);
OutputSections.push_back(Out<ELFT>::MipsRldMap);
}
}
// We add the .got section to the result for dynamic MIPS target because
// its address and properties are mentioned in the .dynamic section.
if (!Out<ELFT>::Got->empty() ||
(isOutputDynamic() && Config->EMachine == EM_MIPS))
OutputSections.push_back(Out<ELFT>::Got);
if (Out<ELFT>::GotPlt && !Out<ELFT>::GotPlt->empty())
OutputSections.push_back(Out<ELFT>::GotPlt);
if (!Out<ELFT>::Plt->empty())
OutputSections.push_back(Out<ELFT>::Plt);
std::stable_sort(OutputSections.begin(), OutputSections.end(),
compareSections<ELFT>);
for (unsigned I = 0, N = OutputSections.size(); I < N; ++I) {
OutputSections[I]->SectionIndex = I + 1;
HasRelro |= (Config->ZRelro && isRelroSection(OutputSections[I]));
}
for (OutputSectionBase<ELFT> *Sec : OutputSections)
Out<ELFT>::ShStrTab->add(Sec->getName());
// Finalizers fix each section's size.
// .dynamic section's finalizer may add strings to .dynstr,
// so finalize that early.
// Likewise, .dynsym is finalized early since that may fill up .gnu.hash.
Out<ELFT>::Dynamic->finalize();
if (isOutputDynamic())
Out<ELFT>::DynSymTab->finalize();
// Fill other section headers.
for (OutputSectionBase<ELFT> *Sec : OutputSections)
Sec->finalize();
// If we have a .opd section (used under PPC64 for function descriptors),
// store a pointer to it here so that we can use it later when processing
// relocations.
Out<ELFT>::Opd = Map.lookup({".opd", SHT_PROGBITS, SHF_WRITE | SHF_ALLOC, 0});
}
