// Tries to remove a sanity check; returns true if it worked.
bool AsapPass::optimizeCheckAway(llvm::Instruction *Inst) {
BranchInst *BI = cast<BranchInst>(Inst);
assert(BI->isConditional() && "Sanity check must be conditional branch.");
unsigned int RegularBranch = getRegularBranch(BI, SCI);
bool Changed = false;
if (RegularBranch == 0) {
BI->setCondition(ConstantInt::getTrue(Inst->getContext()));
Changed = true;
} else if (RegularBranch == 1) {
BI->setCondition(ConstantInt::getFalse(Inst->getContext()));
Changed = true;
} else {
// This can happen, e.g., in the following case:
// array[-1] = a + b;
// is transformed into
// if (a + b overflows)
// report_overflow()
// else
// report_index_out_of_bounds();
// In this case, removing the sanity check does not help much, so we
// just do nothing.
// Thanks to Will Dietz for his explanation at
// http://lists.cs.uiuc.edu/pipermail/llvmdev/2014-April/071958.html
dbgs() << "Warning: Sanity check with no regular branch found.\n";
dbgs() << "The sanity check has been kept intact.\n";
}
if (PrintRemovedChecks && Changed) {
DebugLoc DL = getSanityCheckDebugLoc(BI, RegularBranch);
printDebugLoc(DL, BI->getContext(), dbgs());
dbgs() << ": SanityCheck with cost ";
dbgs() << *BI->getMetadata("cost")->getOperand(0);
if (MDNode *IA = DL.getInlinedAt()) {
dbgs() << " (inlined at ";
printDebugLoc(DebugLoc(IA), BI->getContext(), dbgs());
dbgs() << ")";
}
BasicBlock *Succ = BI->getSuccessor(RegularBranch == 0 ? 1 : 0);
if (const CallInst *CI = SCI->findSanityCheckCall(Succ)) {
dbgs() << " " << CI->getCalledFunction()->getName();
}
dbgs() << "\n";
}
return Changed;
}
/// \brief Insert the missing branch conditions
void StructurizeCFG::insertConditions(bool Loops) {
BranchVector &Conds = Loops ? LoopConds : Conditions;
Value *Default = Loops ? BoolTrue : BoolFalse;
SSAUpdater PhiInserter;
for (BranchVector::iterator I = Conds.begin(),
E = Conds.end(); I != E; ++I) {
BranchInst *Term = *I;
assert(Term->isConditional());
BasicBlock *Parent = Term->getParent();
BasicBlock *SuccTrue = Term->getSuccessor(0);
BasicBlock *SuccFalse = Term->getSuccessor(1);
PhiInserter.Initialize(Boolean, "");
PhiInserter.AddAvailableValue(&Func->getEntryBlock(), Default);
PhiInserter.AddAvailableValue(Loops ? SuccFalse : Parent, Default);
BBPredicates &Preds = Loops ? LoopPreds[SuccFalse] : Predicates[SuccTrue];
NearestCommonDominator Dominator(DT);
Dominator.addBlock(Parent, false);
Value *ParentValue = 0;
for (BBPredicates::iterator PI = Preds.begin(), PE = Preds.end();
PI != PE; ++PI) {
if (PI->first == Parent) {
ParentValue = PI->second;
break;
}
PhiInserter.AddAvailableValue(PI->first, PI->second);
Dominator.addBlock(PI->first);
}
if (ParentValue) {
Term->setCondition(ParentValue);
} else {
if (!Dominator.wasResultExplicitMentioned())
PhiInserter.AddAvailableValue(Dominator.getResult(), Default);
Term->setCondition(PhiInserter.GetValueInMiddleOfBlock(Parent));
}
}
}
bool RangedAddressSanitizer::runOnFunction(Function &F) {
if (getenv("FASAN_DISABLE")) {
SPM_DEBUG( dbgs() << "FASan : disabled\n" );
return false;
}
DL_ = &getAnalysis<DataLayout>();
DT_ = &getAnalysis<DominatorTree>();
LI_ = &getAnalysis<LoopInfo>();
RI_ = &getAnalysis<ReduceIndexation>();
#ifdef ENABLE_REUSE
RE_ = &getAnalysis<RelativeExecutions>();
#endif
RMM_ = &getAnalysis<RelativeMinMax>();
Module_ = F.getParent();
Context_ = &Module_->getContext();
Type *VoidTy = Type::getVoidTy(*Context_);
IntegerType *IntTy = IntegerType::getInt64Ty(*Context_);
IntegerType *BoolTy = IntegerType::getInt1Ty(*Context_);
PointerType *IntPtrTy = PointerType::getUnqual(IntTy);
PointerType *VoidPtrTy = PointerType::getInt8PtrTy(*Context_);
SPM_DEBUG( F.dump() );
outs() << "[IterationInfo]\n";
for (Loop * loop : *LI_) {
ii_visitLoop(loop);
}
outs() << "[EndOfIterationInfo]\n";
#if 0 // disabled initialization,shutdown sequence for FASan
if (F.getName() == "main") {
SPM_DEBUG(dbgs() << "RangedAddressSanitizer: inserting hwloc calls into "
"main function\n");
FunctionType *FnType = FunctionType::get(VoidTy, ArrayRef<Type*>(), false);
IRBuilder<> IRB(&(*F.getEntryBlock().begin()));
Constant *Init = Module_->getOrInsertFunction("__spm_init", FnType);
IRB.CreateCall(Init);
Constant *End = Module_->getOrInsertFunction("__spm_end", FnType);
for (auto &BB : F) {
TerminatorInst *TI = BB.getTerminator();
if (isa<ReturnInst>(TI)) {
IRB.SetInsertPoint(TI);
IRB.CreateCall(End);
}
}
}
#endif
if (!ClFunc.empty() && F.getName() != ClFunc) {
SPM_DEBUG(dbgs() << "RangedAddressSanitizer: skipping function "
<< F.getName() << "\n");
return false;
}
Calls_.clear();
SPM_DEBUG(dbgs() << "RangedAddressSanitizer: processing function "
<< F.getName() << "\n");
std::vector<Type*> ReuseFnFormals = { VoidPtrTy, IntTy, IntTy, IntTy };
FunctionType *ReuseFnType = FunctionType::get(BoolTy, ReuseFnFormals, false);
ReuseFn_ =
F.getParent()->getOrInsertFunction("__fasan_check", ReuseFnType);
ReuseFnDestroy_ =
F.getParent()->getOrInsertFunction("__spm_give", ReuseFnType);
// Visit all loops in bottom-up order (innter-most loops first)
std::set<BasicBlock*> Processed;
auto Entry = DT_->getRootNode();
for (auto ET = po_begin(Entry), EE = po_end(Entry); ET != EE; ++ET) {
BasicBlock *Header = (*ET)->getBlock();
if (LI_->isLoopHeader(Header)) {
SPM_DEBUG(dbgs() << "RangedAddressSanitizer: processing loop at "
<< Header->getName() << "\n");
Loop *L = LI_->getLoopFor(Header);
if (L->getNumBackEdges() != 1 ||
std::distance(pred_begin(Header), pred_end(Header)) != 2) {
SPM_DEBUG(dbgs() << "RangedAddressSanitizer: loop has multiple "
<< "backedges or multiple incoming outer blocks\n");
continue;
}
SPM_DEBUG(dbgs() << "RangedAddressSanitizer: processing loop at "
<< Header->getName() << "\n");
// visit all memory acccesses in this loop
for (auto BB = L->block_begin(), BE = L->block_end(); BB != BE; ++BB) {
if (!Processed.count(*BB)) {
Processed.insert(*BB);
for (auto &I : *(*BB))
generateCallFor(L, &I);
}
}
}
}
// FAsan logic goes here
std::map<const BasicBlock*,BasicBlock*> clonedBlockMap; // keeps track of cloned regions to avoid redundant cloning
std::vector<CallInst*> ToInline;
for (auto &CI : Calls_) {
BasicBlock * Preheader = CI.Preheader;
// TODO decide whether it is worthwhile to optimize for this case
// insert range check
IRBuilder<> IRB(Preheader->getTerminator());
Value *VoidArray = IRB.CreateBitCast(CI.Array, VoidPtrTy);
std::vector<Value*> Args = { VoidArray, CI.Min, CI.Max, CI.Reuse };
CallInst *CR = IRB.CreateCall(ReuseFn_, Args);
ToInline.push_back(CR);
// verify if this loop was already instrumented
TerminatorInst * preHeaderTerm = CR->getParent()->getTerminator();
BranchInst * preHeaderBranch = dyn_cast<BranchInst>(preHeaderTerm);
if (preHeaderBranch && preHeaderBranch->isConditional()) {
// discover the structure of the instrumented code (safe and default region)
// abort, if this does not look like instrumented code
BasicBlock * firstTarget = preHeaderBranch->getSuccessor(0);
BasicBlock * secondTarget = preHeaderBranch->getSuccessor(1);
BasicBlock * safeHeader, * defHeader;
if (clonedBlockMap.count(firstTarget)) {
defHeader = firstTarget;
safeHeader = clonedBlockMap[firstTarget];
assert(safeHeader == secondTarget);
} else {
assert(clonedBlockMap.count(secondTarget));
defHeader = secondTarget;
safeHeader = clonedBlockMap[secondTarget];
assert(safeHeader == firstTarget);
}
SPM_DEBUG( dbgs() << "FASan: (Unsupported) second array in safe region controlled by " << * preHeaderBranch << "\n" );
Loop * defLoop = LI_->getLoopFor(defHeader);
assert(defLoop && "default region is not a loop!");
Loop::block_iterator itBodyBlock,S,E;
S = defLoop->block_begin();
E = defLoop->block_end();
// mark accesses in cloned region as safe
for (itBodyBlock = S;itBodyBlock != E; ++itBodyBlock) {
BasicBlock * defBodyBlock = *itBodyBlock;
BasicBlock * safeBodyBlock = clonedBlockMap[defBodyBlock];
for(auto & inst : *safeBodyBlock) {
markSafeArrayUse(&inst, CI.Array);
}
}
// add conjunctive test
Value * oldCond = preHeaderBranch->getCondition();
Value * joinedCond = IRB.CreateAnd(oldCond, CR, "allsafe");
preHeaderBranch->setCondition(joinedCond);
} else {
// get loop
Loop* finalLoop = CI.FinalLoop;
Loop::block_iterator itBodyBlock,S,E;
S = finalLoop->block_begin();
E = finalLoop->block_end();
// clone loop body (cloned loop will run unchecked)
ValueToValueMapTy cloneMap;
BasicBlock * clonedHeader = 0;
std::vector<BasicBlock*> clonedBlocks;
for (itBodyBlock = S;itBodyBlock != E; ++itBodyBlock) {
const BasicBlock * bodyBlock = *itBodyBlock;
BasicBlock * clonedBlock = CloneBasicBlock(bodyBlock, cloneMap, "_checked", &F, 0);
cloneMap[bodyBlock] = clonedBlock;
clonedBlockMap[bodyBlock] = clonedBlock;
clonedBlocks.push_back(clonedBlock);
if (bodyBlock == finalLoop->getHeader()) {
clonedHeader = clonedBlock;
SPM_DEBUG( dbgs() << "FASan: loop header case at " << bodyBlock->getName() << "\n" );
} else {
SPM_DEBUG( dbgs() << "FASan: non-header block at " << bodyBlock->getName() << "\n" );
}
}
if (!clonedHeader) {
// TODO run clean-up code
SPM_DEBUG( dbgs() << "FASan: could not find header!\n");
abort();
}
// Remap uses inside cloned region (mark pointers in the region as unguarded)
for (BasicBlock * block : clonedBlocks) {
for(auto & inst : *block) {
RemapInstruction(&inst, cloneMap, RF_IgnoreMissingEntries);
markSafeArrayUse(&inst, CI.Array);
}
}
// TODO fix PHI-nodes in exit blocks
// Rewire terminator of the range check to branch to the cloned region
TerminatorInst * checkTermInst = CR->getParent()->getTerminator();
if (BranchInst * checkBranchInst = dyn_cast<BranchInst>(checkTermInst)) {
if (checkBranchInst->isUnconditional()) {
BasicBlock * defTarget = checkBranchInst->getSuccessor(0);
BranchInst * modifiedBranchInst = BranchInst::Create(clonedHeader, defTarget, CR, checkBranchInst);
checkBranchInst->replaceAllUsesWith(modifiedBranchInst);
checkBranchInst->eraseFromParent();
} else {
SPM_DEBUG( dbgs() << "FASan: Unexpected conditional branch (preheader should branch unconditional, other array checks will introduce conditional branches) " << * checkTermInst << "\n" );
abort();
}
} else {
SPM_DEBUG( dbgs() << "FASan: unsupported terminator type " << * checkTermInst << "\n" );
abort();
}
}
#if 0
IRB.SetInsertPoint(&(*CI.Final->begin()));
IRB.CreateCall(ReuseFnDestroy_, Args);
#endif
SPM_DEBUG(dbgs() << "RangedAddressSanitizer: call instruction: " << *CR
<< "\n");
}
// inline calls
#ifdef FASAN_INLINE_RUNTIME
for (CallInst * call : ToInline) {
assert(call);
InlineFunctionInfo IFI;
InlineFunction(call, IFI, false);
}
#endif
SPM_DEBUG( F.dump() );
return true;
}
bool PPCCTRLoops::convertToCTRLoop(Loop *L) {
bool MadeChange = false;
Triple TT = Triple(L->getHeader()->getParent()->getParent()->
getTargetTriple());
if (!TT.isArch32Bit() && !TT.isArch64Bit())
return MadeChange; // Unknown arch. type.
// Process nested loops first.
for (Loop::iterator I = L->begin(), E = L->end(); I != E; ++I) {
MadeChange |= convertToCTRLoop(*I);
}
// If a nested loop has been converted, then we can't convert this loop.
if (MadeChange)
return MadeChange;
#ifndef NDEBUG
// Stop trying after reaching the limit (if any).
int Limit = CTRLoopLimit;
if (Limit >= 0) {
if (Counter >= CTRLoopLimit)
return false;
Counter++;
}
#endif
// We don't want to spill/restore the counter register, and so we don't
// want to use the counter register if the loop contains calls.
for (Loop::block_iterator I = L->block_begin(), IE = L->block_end();
I != IE; ++I)
if (mightUseCTR(TT, *I))
return MadeChange;
SmallVector<BasicBlock*, 4> ExitingBlocks;
L->getExitingBlocks(ExitingBlocks);
BasicBlock *CountedExitBlock = 0;
const SCEV *ExitCount = 0;
BranchInst *CountedExitBranch = 0;
for (SmallVectorImpl<BasicBlock *>::iterator I = ExitingBlocks.begin(),
IE = ExitingBlocks.end(); I != IE; ++I) {
const SCEV *EC = SE->getExitCount(L, *I);
DEBUG(dbgs() << "Exit Count for " << *L << " from block " <<
(*I)->getName() << ": " << *EC << "\n");
if (isa<SCEVCouldNotCompute>(EC))
continue;
if (const SCEVConstant *ConstEC = dyn_cast<SCEVConstant>(EC)) {
if (ConstEC->getValue()->isZero())
continue;
} else if (!SE->isLoopInvariant(EC, L))
continue;
if (SE->getTypeSizeInBits(EC->getType()) > (TT.isArch64Bit() ? 64 : 32))
continue;
// We now have a loop-invariant count of loop iterations (which is not the
// constant zero) for which we know that this loop will not exit via this
// exisiting block.
// We need to make sure that this block will run on every loop iteration.
// For this to be true, we must dominate all blocks with backedges. Such
// blocks are in-loop predecessors to the header block.
bool NotAlways = false;
for (pred_iterator PI = pred_begin(L->getHeader()),
PIE = pred_end(L->getHeader()); PI != PIE; ++PI) {
if (!L->contains(*PI))
continue;
if (!DT->dominates(*I, *PI)) {
NotAlways = true;
break;
}
}
if (NotAlways)
continue;
// Make sure this blocks ends with a conditional branch.
Instruction *TI = (*I)->getTerminator();
if (!TI)
continue;
if (BranchInst *BI = dyn_cast<BranchInst>(TI)) {
if (!BI->isConditional())
continue;
CountedExitBranch = BI;
} else
continue;
// Note that this block may not be the loop latch block, even if the loop
// has a latch block.
CountedExitBlock = *I;
ExitCount = EC;
break;
}
if (!CountedExitBlock)
return MadeChange;
BasicBlock *Preheader = L->getLoopPreheader();
// If we don't have a preheader, then insert one. If we already have a
// preheader, then we can use it (except if the preheader contains a use of
// the CTR register because some such uses might be reordered by the
// selection DAG after the mtctr instruction).
if (!Preheader || mightUseCTR(TT, Preheader))
Preheader = InsertPreheaderForLoop(L, this);
if (!Preheader)
return MadeChange;
DEBUG(dbgs() << "Preheader for exit count: " << Preheader->getName() << "\n");
// Insert the count into the preheader and replace the condition used by the
// selected branch.
MadeChange = true;
SCEVExpander SCEVE(*SE, "loopcnt");
LLVMContext &C = SE->getContext();
Type *CountType = TT.isArch64Bit() ? Type::getInt64Ty(C) :
Type::getInt32Ty(C);
if (!ExitCount->getType()->isPointerTy() &&
ExitCount->getType() != CountType)
ExitCount = SE->getZeroExtendExpr(ExitCount, CountType);
ExitCount = SE->getAddExpr(ExitCount,
SE->getConstant(CountType, 1));
Value *ECValue = SCEVE.expandCodeFor(ExitCount, CountType,
Preheader->getTerminator());
IRBuilder<> CountBuilder(Preheader->getTerminator());
Module *M = Preheader->getParent()->getParent();
Value *MTCTRFunc = Intrinsic::getDeclaration(M, Intrinsic::ppc_mtctr,
CountType);
CountBuilder.CreateCall(MTCTRFunc, ECValue);
IRBuilder<> CondBuilder(CountedExitBranch);
Value *DecFunc =
Intrinsic::getDeclaration(M, Intrinsic::ppc_is_decremented_ctr_nonzero);
Value *NewCond = CondBuilder.CreateCall(DecFunc);
Value *OldCond = CountedExitBranch->getCondition();
CountedExitBranch->setCondition(NewCond);
// The false branch must exit the loop.
if (!L->contains(CountedExitBranch->getSuccessor(0)))
CountedExitBranch->swapSuccessors();
// The old condition may be dead now, and may have even created a dead PHI
// (the original induction variable).
RecursivelyDeleteTriviallyDeadInstructions(OldCond);
DeleteDeadPHIs(CountedExitBlock);
++NumCTRLoops;
return MadeChange;
}
bool llvm::UnrollRuntimeLoopRemainder(Loop *L, unsigned Count,
bool AllowExpensiveTripCount,
bool UseEpilogRemainder,
bool UnrollRemainder,
LoopInfo *LI, ScalarEvolution *SE,
DominatorTree *DT, AssumptionCache *AC,
OptimizationRemarkEmitter *ORE,
bool PreserveLCSSA) {
DEBUG(dbgs() << "Trying runtime unrolling on Loop: \n");
DEBUG(L->dump());
DEBUG(UseEpilogRemainder ? dbgs() << "Using epilog remainder.\n" :
dbgs() << "Using prolog remainder.\n");
// Make sure the loop is in canonical form.
if (!L->isLoopSimplifyForm()) {
DEBUG(dbgs() << "Not in simplify form!\n");
return false;
}
// Guaranteed by LoopSimplifyForm.
BasicBlock *Latch = L->getLoopLatch();
BasicBlock *Header = L->getHeader();
BranchInst *LatchBR = cast<BranchInst>(Latch->getTerminator());
unsigned ExitIndex = LatchBR->getSuccessor(0) == Header ? 1 : 0;
BasicBlock *LatchExit = LatchBR->getSuccessor(ExitIndex);
// Cloning the loop basic blocks (`CloneLoopBlocks`) requires that one of the
// targets of the Latch be an exit block out of the loop. This needs
// to be guaranteed by the callers of UnrollRuntimeLoopRemainder.
assert(!L->contains(LatchExit) &&
"one of the loop latch successors should be the exit block!");
// These are exit blocks other than the target of the latch exiting block.
SmallVector<BasicBlock *, 4> OtherExits;
bool isMultiExitUnrollingEnabled =
canSafelyUnrollMultiExitLoop(L, OtherExits, LatchExit, PreserveLCSSA,
UseEpilogRemainder) &&
canProfitablyUnrollMultiExitLoop(L, OtherExits, LatchExit, PreserveLCSSA,
UseEpilogRemainder);
// Support only single exit and exiting block unless multi-exit loop unrolling is enabled.
if (!isMultiExitUnrollingEnabled &&
(!L->getExitingBlock() || OtherExits.size())) {
DEBUG(
dbgs()
<< "Multiple exit/exiting blocks in loop and multi-exit unrolling not "
"enabled!\n");
return false;
}
// Use Scalar Evolution to compute the trip count. This allows more loops to
// be unrolled than relying on induction var simplification.
if (!SE)
return false;
// Only unroll loops with a computable trip count, and the trip count needs
// to be an int value (allowing a pointer type is a TODO item).
// We calculate the backedge count by using getExitCount on the Latch block,
// which is proven to be the only exiting block in this loop. This is same as
// calculating getBackedgeTakenCount on the loop (which computes SCEV for all
// exiting blocks).
const SCEV *BECountSC = SE->getExitCount(L, Latch);
if (isa<SCEVCouldNotCompute>(BECountSC) ||
!BECountSC->getType()->isIntegerTy()) {
DEBUG(dbgs() << "Could not compute exit block SCEV\n");
return false;
}
unsigned BEWidth = cast<IntegerType>(BECountSC->getType())->getBitWidth();
// Add 1 since the backedge count doesn't include the first loop iteration.
const SCEV *TripCountSC =
SE->getAddExpr(BECountSC, SE->getConstant(BECountSC->getType(), 1));
if (isa<SCEVCouldNotCompute>(TripCountSC)) {
DEBUG(dbgs() << "Could not compute trip count SCEV.\n");
return false;
}
BasicBlock *PreHeader = L->getLoopPreheader();
BranchInst *PreHeaderBR = cast<BranchInst>(PreHeader->getTerminator());
const DataLayout &DL = Header->getModule()->getDataLayout();
SCEVExpander Expander(*SE, DL, "loop-unroll");
if (!AllowExpensiveTripCount &&
Expander.isHighCostExpansion(TripCountSC, L, PreHeaderBR)) {
DEBUG(dbgs() << "High cost for expanding trip count scev!\n");
return false;
}
// This constraint lets us deal with an overflowing trip count easily; see the
// comment on ModVal below.
if (Log2_32(Count) > BEWidth) {
DEBUG(dbgs()
<< "Count failed constraint on overflow trip count calculation.\n");
return false;
}
// Loop structure is the following:
//
// PreHeader
// Header
// ...
// Latch
// LatchExit
BasicBlock *NewPreHeader;
BasicBlock *NewExit = nullptr;
BasicBlock *PrologExit = nullptr;
BasicBlock *EpilogPreHeader = nullptr;
BasicBlock *PrologPreHeader = nullptr;
if (UseEpilogRemainder) {
// If epilog remainder
// Split PreHeader to insert a branch around loop for unrolling.
NewPreHeader = SplitBlock(PreHeader, PreHeader->getTerminator(), DT, LI);
NewPreHeader->setName(PreHeader->getName() + ".new");
// Split LatchExit to create phi nodes from branch above.
SmallVector<BasicBlock*, 4> Preds(predecessors(LatchExit));
NewExit = SplitBlockPredecessors(LatchExit, Preds, ".unr-lcssa",
DT, LI, PreserveLCSSA);
// Split NewExit to insert epilog remainder loop.
EpilogPreHeader = SplitBlock(NewExit, NewExit->getTerminator(), DT, LI);
EpilogPreHeader->setName(Header->getName() + ".epil.preheader");
} else {
// If prolog remainder
// Split the original preheader twice to insert prolog remainder loop
PrologPreHeader = SplitEdge(PreHeader, Header, DT, LI);
PrologPreHeader->setName(Header->getName() + ".prol.preheader");
PrologExit = SplitBlock(PrologPreHeader, PrologPreHeader->getTerminator(),
DT, LI);
PrologExit->setName(Header->getName() + ".prol.loopexit");
// Split PrologExit to get NewPreHeader.
NewPreHeader = SplitBlock(PrologExit, PrologExit->getTerminator(), DT, LI);
NewPreHeader->setName(PreHeader->getName() + ".new");
}
// Loop structure should be the following:
// Epilog Prolog
//
// PreHeader PreHeader
// *NewPreHeader *PrologPreHeader
// Header *PrologExit
// ... *NewPreHeader
// Latch Header
// *NewExit ...
// *EpilogPreHeader Latch
// LatchExit LatchExit
// Calculate conditions for branch around loop for unrolling
// in epilog case and around prolog remainder loop in prolog case.
// Compute the number of extra iterations required, which is:
// extra iterations = run-time trip count % loop unroll factor
PreHeaderBR = cast<BranchInst>(PreHeader->getTerminator());
Value *TripCount = Expander.expandCodeFor(TripCountSC, TripCountSC->getType(),
PreHeaderBR);
Value *BECount = Expander.expandCodeFor(BECountSC, BECountSC->getType(),
PreHeaderBR);
IRBuilder<> B(PreHeaderBR);
Value *ModVal;
// Calculate ModVal = (BECount + 1) % Count.
// Note that TripCount is BECount + 1.
if (isPowerOf2_32(Count)) {
// When Count is power of 2 we don't BECount for epilog case, however we'll
// need it for a branch around unrolling loop for prolog case.
ModVal = B.CreateAnd(TripCount, Count - 1, "xtraiter");
// 1. There are no iterations to be run in the prolog/epilog loop.
// OR
// 2. The addition computing TripCount overflowed.
//
// If (2) is true, we know that TripCount really is (1 << BEWidth) and so
// the number of iterations that remain to be run in the original loop is a
// multiple Count == (1 << Log2(Count)) because Log2(Count) <= BEWidth (we
// explicitly check this above).
} else {
// As (BECount + 1) can potentially unsigned overflow we count
// (BECount % Count) + 1 which is overflow safe as BECount % Count < Count.
Value *ModValTmp = B.CreateURem(BECount,
ConstantInt::get(BECount->getType(),
Count));
Value *ModValAdd = B.CreateAdd(ModValTmp,
ConstantInt::get(ModValTmp->getType(), 1));
// At that point (BECount % Count) + 1 could be equal to Count.
// To handle this case we need to take mod by Count one more time.
ModVal = B.CreateURem(ModValAdd,
ConstantInt::get(BECount->getType(), Count),
"xtraiter");
}
Value *BranchVal =
UseEpilogRemainder ? B.CreateICmpULT(BECount,
ConstantInt::get(BECount->getType(),
Count - 1)) :
B.CreateIsNotNull(ModVal, "lcmp.mod");
BasicBlock *RemainderLoop = UseEpilogRemainder ? NewExit : PrologPreHeader;
BasicBlock *UnrollingLoop = UseEpilogRemainder ? NewPreHeader : PrologExit;
// Branch to either remainder (extra iterations) loop or unrolling loop.
B.CreateCondBr(BranchVal, RemainderLoop, UnrollingLoop);
PreHeaderBR->eraseFromParent();
if (DT) {
if (UseEpilogRemainder)
DT->changeImmediateDominator(NewExit, PreHeader);
else
DT->changeImmediateDominator(PrologExit, PreHeader);
}
Function *F = Header->getParent();
// Get an ordered list of blocks in the loop to help with the ordering of the
// cloned blocks in the prolog/epilog code
LoopBlocksDFS LoopBlocks(L);
LoopBlocks.perform(LI);
//
// For each extra loop iteration, create a copy of the loop's basic blocks
// and generate a condition that branches to the copy depending on the
// number of 'left over' iterations.
//
std::vector<BasicBlock *> NewBlocks;
ValueToValueMapTy VMap;
// For unroll factor 2 remainder loop will have 1 iterations.
// Do not create 1 iteration loop.
bool CreateRemainderLoop = (Count != 2);
// Clone all the basic blocks in the loop. If Count is 2, we don't clone
// the loop, otherwise we create a cloned loop to execute the extra
// iterations. This function adds the appropriate CFG connections.
BasicBlock *InsertBot = UseEpilogRemainder ? LatchExit : PrologExit;
BasicBlock *InsertTop = UseEpilogRemainder ? EpilogPreHeader : PrologPreHeader;
Loop *remainderLoop = CloneLoopBlocks(
L, ModVal, CreateRemainderLoop, UseEpilogRemainder, UnrollRemainder,
InsertTop, InsertBot,
NewPreHeader, NewBlocks, LoopBlocks, VMap, DT, LI);
// Insert the cloned blocks into the function.
F->getBasicBlockList().splice(InsertBot->getIterator(),
F->getBasicBlockList(),
NewBlocks[0]->getIterator(),
F->end());
// Now the loop blocks are cloned and the other exiting blocks from the
// remainder are connected to the original Loop's exit blocks. The remaining
// work is to update the phi nodes in the original loop, and take in the
// values from the cloned region. Also update the dominator info for
// OtherExits and their immediate successors, since we have new edges into
// OtherExits.
SmallSet<BasicBlock*, 8> ImmediateSuccessorsOfExitBlocks;
for (auto *BB : OtherExits) {
for (auto &II : *BB) {
// Given we preserve LCSSA form, we know that the values used outside the
// loop will be used through these phi nodes at the exit blocks that are
// transformed below.
if (!isa<PHINode>(II))
break;
PHINode *Phi = cast<PHINode>(&II);
unsigned oldNumOperands = Phi->getNumIncomingValues();
// Add the incoming values from the remainder code to the end of the phi
// node.
for (unsigned i =0; i < oldNumOperands; i++){
Value *newVal = VMap[Phi->getIncomingValue(i)];
// newVal can be a constant or derived from values outside the loop, and
// hence need not have a VMap value.
if (!newVal)
newVal = Phi->getIncomingValue(i);
Phi->addIncoming(newVal,
cast<BasicBlock>(VMap[Phi->getIncomingBlock(i)]));
}
}
#if defined(EXPENSIVE_CHECKS) && !defined(NDEBUG)
for (BasicBlock *SuccBB : successors(BB)) {
assert(!(any_of(OtherExits,
[SuccBB](BasicBlock *EB) { return EB == SuccBB; }) ||
SuccBB == LatchExit) &&
"Breaks the definition of dedicated exits!");
}
#endif
// Update the dominator info because the immediate dominator is no longer the
// header of the original Loop. BB has edges both from L and remainder code.
// Since the preheader determines which loop is run (L or directly jump to
// the remainder code), we set the immediate dominator as the preheader.
if (DT) {
DT->changeImmediateDominator(BB, PreHeader);
// Also update the IDom for immediate successors of BB. If the current
// IDom is the header, update the IDom to be the preheader because that is
// the nearest common dominator of all predecessors of SuccBB. We need to
// check for IDom being the header because successors of exit blocks can
// have edges from outside the loop, and we should not incorrectly update
// the IDom in that case.
for (BasicBlock *SuccBB: successors(BB))
if (ImmediateSuccessorsOfExitBlocks.insert(SuccBB).second) {
if (DT->getNode(SuccBB)->getIDom()->getBlock() == Header) {
assert(!SuccBB->getSinglePredecessor() &&
"BB should be the IDom then!");
DT->changeImmediateDominator(SuccBB, PreHeader);
}
}
}
}
// Loop structure should be the following:
// Epilog Prolog
//
// PreHeader PreHeader
// NewPreHeader PrologPreHeader
// Header PrologHeader
// ... ...
// Latch PrologLatch
// NewExit PrologExit
// EpilogPreHeader NewPreHeader
// EpilogHeader Header
// ... ...
// EpilogLatch Latch
// LatchExit LatchExit
// Rewrite the cloned instruction operands to use the values created when the
// clone is created.
for (BasicBlock *BB : NewBlocks) {
for (Instruction &I : *BB) {
RemapInstruction(&I, VMap,
RF_NoModuleLevelChanges | RF_IgnoreMissingLocals);
}
}
if (UseEpilogRemainder) {
// Connect the epilog code to the original loop and update the
// PHI functions.
ConnectEpilog(L, ModVal, NewExit, LatchExit, PreHeader,
EpilogPreHeader, NewPreHeader, VMap, DT, LI,
PreserveLCSSA);
// Update counter in loop for unrolling.
// I should be multiply of Count.
IRBuilder<> B2(NewPreHeader->getTerminator());
Value *TestVal = B2.CreateSub(TripCount, ModVal, "unroll_iter");
BranchInst *LatchBR = cast<BranchInst>(Latch->getTerminator());
B2.SetInsertPoint(LatchBR);
PHINode *NewIdx = PHINode::Create(TestVal->getType(), 2, "niter",
Header->getFirstNonPHI());
Value *IdxSub =
B2.CreateSub(NewIdx, ConstantInt::get(NewIdx->getType(), 1),
NewIdx->getName() + ".nsub");
Value *IdxCmp;
if (LatchBR->getSuccessor(0) == Header)
IdxCmp = B2.CreateIsNotNull(IdxSub, NewIdx->getName() + ".ncmp");
else
IdxCmp = B2.CreateIsNull(IdxSub, NewIdx->getName() + ".ncmp");
NewIdx->addIncoming(TestVal, NewPreHeader);
NewIdx->addIncoming(IdxSub, Latch);
LatchBR->setCondition(IdxCmp);
} else {
// Connect the prolog code to the original loop and update the
// PHI functions.
ConnectProlog(L, BECount, Count, PrologExit, LatchExit, PreHeader,
NewPreHeader, VMap, DT, LI, PreserveLCSSA);
}
// If this loop is nested, then the loop unroller changes the code in the
// parent loop, so the Scalar Evolution pass needs to be run again.
if (Loop *ParentLoop = L->getParentLoop())
SE->forgetLoop(ParentLoop);
// Canonicalize to LoopSimplifyForm both original and remainder loops. We
// cannot rely on the LoopUnrollPass to do this because it only does
// canonicalization for parent/subloops and not the sibling loops.
if (OtherExits.size() > 0) {
// Generate dedicated exit blocks for the original loop, to preserve
// LoopSimplifyForm.
formDedicatedExitBlocks(L, DT, LI, PreserveLCSSA);
// Generate dedicated exit blocks for the remainder loop if one exists, to
// preserve LoopSimplifyForm.
if (remainderLoop)
formDedicatedExitBlocks(remainderLoop, DT, LI, PreserveLCSSA);
}
if (remainderLoop && UnrollRemainder) {
DEBUG(dbgs() << "Unrolling remainder loop\n");
UnrollLoop(remainderLoop, /*Count*/Count - 1, /*TripCount*/Count - 1,
/*Force*/false, /*AllowRuntime*/false,
/*AllowExpensiveTripCount*/false, /*PreserveCondBr*/true,
/*PreserveOnlyFirst*/false, /*TripMultiple*/1,
/*PeelCount*/0, /*UnrollRemainder*/false, LI, SE, DT, AC, ORE,
PreserveLCSSA);
}
NumRuntimeUnrolled++;
return true;
}
bool llvm::UnrollRuntimeLoopRemainder(Loop *L, unsigned Count,
bool AllowExpensiveTripCount,
bool UseEpilogRemainder,
LoopInfo *LI, ScalarEvolution *SE,
DominatorTree *DT, bool PreserveLCSSA) {
// for now, only unroll loops that contain a single exit
if (!L->getExitingBlock())
return false;
// Make sure the loop is in canonical form, and there is a single
// exit block only.
if (!L->isLoopSimplifyForm())
return false;
BasicBlock *Exit = L->getUniqueExitBlock(); // successor out of loop
if (!Exit)
return false;
// Use Scalar Evolution to compute the trip count. This allows more loops to
// be unrolled than relying on induction var simplification.
if (!SE)
return false;
// Only unroll loops with a computable trip count, and the trip count needs
// to be an int value (allowing a pointer type is a TODO item).
const SCEV *BECountSC = SE->getBackedgeTakenCount(L);
if (isa<SCEVCouldNotCompute>(BECountSC) ||
!BECountSC->getType()->isIntegerTy())
return false;
unsigned BEWidth = cast<IntegerType>(BECountSC->getType())->getBitWidth();
// Add 1 since the backedge count doesn't include the first loop iteration.
const SCEV *TripCountSC =
SE->getAddExpr(BECountSC, SE->getConstant(BECountSC->getType(), 1));
if (isa<SCEVCouldNotCompute>(TripCountSC))
return false;
BasicBlock *Header = L->getHeader();
BasicBlock *PreHeader = L->getLoopPreheader();
BranchInst *PreHeaderBR = cast<BranchInst>(PreHeader->getTerminator());
const DataLayout &DL = Header->getModule()->getDataLayout();
SCEVExpander Expander(*SE, DL, "loop-unroll");
if (!AllowExpensiveTripCount &&
Expander.isHighCostExpansion(TripCountSC, L, PreHeaderBR))
return false;
// This constraint lets us deal with an overflowing trip count easily; see the
// comment on ModVal below.
if (Log2_32(Count) > BEWidth)
return false;
BasicBlock *Latch = L->getLoopLatch();
// Loop structure is the following:
//
// PreHeader
// Header
// ...
// Latch
// Exit
BasicBlock *NewPreHeader;
BasicBlock *NewExit = nullptr;
BasicBlock *PrologExit = nullptr;
BasicBlock *EpilogPreHeader = nullptr;
BasicBlock *PrologPreHeader = nullptr;
if (UseEpilogRemainder) {
// If epilog remainder
// Split PreHeader to insert a branch around loop for unrolling.
NewPreHeader = SplitBlock(PreHeader, PreHeader->getTerminator(), DT, LI);
NewPreHeader->setName(PreHeader->getName() + ".new");
// Split Exit to create phi nodes from branch above.
SmallVector<BasicBlock*, 4> Preds(predecessors(Exit));
NewExit = SplitBlockPredecessors(Exit, Preds, ".unr-lcssa",
DT, LI, PreserveLCSSA);
// Split NewExit to insert epilog remainder loop.
EpilogPreHeader = SplitBlock(NewExit, NewExit->getTerminator(), DT, LI);
EpilogPreHeader->setName(Header->getName() + ".epil.preheader");
} else {
// If prolog remainder
// Split the original preheader twice to insert prolog remainder loop
PrologPreHeader = SplitEdge(PreHeader, Header, DT, LI);
PrologPreHeader->setName(Header->getName() + ".prol.preheader");
PrologExit = SplitBlock(PrologPreHeader, PrologPreHeader->getTerminator(),
DT, LI);
PrologExit->setName(Header->getName() + ".prol.loopexit");
// Split PrologExit to get NewPreHeader.
NewPreHeader = SplitBlock(PrologExit, PrologExit->getTerminator(), DT, LI);
NewPreHeader->setName(PreHeader->getName() + ".new");
}
// Loop structure should be the following:
// Epilog Prolog
//
// PreHeader PreHeader
// *NewPreHeader *PrologPreHeader
// Header *PrologExit
// ... *NewPreHeader
// Latch Header
// *NewExit ...
// *EpilogPreHeader Latch
// Exit Exit
// Calculate conditions for branch around loop for unrolling
// in epilog case and around prolog remainder loop in prolog case.
// Compute the number of extra iterations required, which is:
// extra iterations = run-time trip count % loop unroll factor
PreHeaderBR = cast<BranchInst>(PreHeader->getTerminator());
Value *TripCount = Expander.expandCodeFor(TripCountSC, TripCountSC->getType(),
PreHeaderBR);
Value *BECount = Expander.expandCodeFor(BECountSC, BECountSC->getType(),
PreHeaderBR);
IRBuilder<> B(PreHeaderBR);
Value *ModVal;
// Calculate ModVal = (BECount + 1) % Count.
// Note that TripCount is BECount + 1.
if (isPowerOf2_32(Count)) {
// When Count is power of 2 we don't BECount for epilog case, however we'll
// need it for a branch around unrolling loop for prolog case.
ModVal = B.CreateAnd(TripCount, Count - 1, "xtraiter");
// 1. There are no iterations to be run in the prolog/epilog loop.
// OR
// 2. The addition computing TripCount overflowed.
//
// If (2) is true, we know that TripCount really is (1 << BEWidth) and so
// the number of iterations that remain to be run in the original loop is a
// multiple Count == (1 << Log2(Count)) because Log2(Count) <= BEWidth (we
// explicitly check this above).
} else {
// As (BECount + 1) can potentially unsigned overflow we count
// (BECount % Count) + 1 which is overflow safe as BECount % Count < Count.
Value *ModValTmp = B.CreateURem(BECount,
ConstantInt::get(BECount->getType(),
Count));
Value *ModValAdd = B.CreateAdd(ModValTmp,
ConstantInt::get(ModValTmp->getType(), 1));
// At that point (BECount % Count) + 1 could be equal to Count.
// To handle this case we need to take mod by Count one more time.
ModVal = B.CreateURem(ModValAdd,
ConstantInt::get(BECount->getType(), Count),
"xtraiter");
}
Value *BranchVal =
UseEpilogRemainder ? B.CreateICmpULT(BECount,
ConstantInt::get(BECount->getType(),
Count - 1)) :
B.CreateIsNotNull(ModVal, "lcmp.mod");
BasicBlock *RemainderLoop = UseEpilogRemainder ? NewExit : PrologPreHeader;
BasicBlock *UnrollingLoop = UseEpilogRemainder ? NewPreHeader : PrologExit;
// Branch to either remainder (extra iterations) loop or unrolling loop.
B.CreateCondBr(BranchVal, RemainderLoop, UnrollingLoop);
PreHeaderBR->eraseFromParent();
Function *F = Header->getParent();
// Get an ordered list of blocks in the loop to help with the ordering of the
// cloned blocks in the prolog/epilog code
LoopBlocksDFS LoopBlocks(L);
LoopBlocks.perform(LI);
//
// For each extra loop iteration, create a copy of the loop's basic blocks
// and generate a condition that branches to the copy depending on the
// number of 'left over' iterations.
//
std::vector<BasicBlock *> NewBlocks;
ValueToValueMapTy VMap;
// For unroll factor 2 remainder loop will have 1 iterations.
// Do not create 1 iteration loop.
bool CreateRemainderLoop = (Count != 2);
// Clone all the basic blocks in the loop. If Count is 2, we don't clone
// the loop, otherwise we create a cloned loop to execute the extra
// iterations. This function adds the appropriate CFG connections.
BasicBlock *InsertBot = UseEpilogRemainder ? Exit : PrologExit;
BasicBlock *InsertTop = UseEpilogRemainder ? EpilogPreHeader : PrologPreHeader;
CloneLoopBlocks(L, ModVal, CreateRemainderLoop, UseEpilogRemainder, InsertTop,
InsertBot, NewPreHeader, NewBlocks, LoopBlocks, VMap, LI);
// Insert the cloned blocks into the function.
F->getBasicBlockList().splice(InsertBot->getIterator(),
F->getBasicBlockList(),
NewBlocks[0]->getIterator(),
F->end());
// Loop structure should be the following:
// Epilog Prolog
//
// PreHeader PreHeader
// NewPreHeader PrologPreHeader
// Header PrologHeader
// ... ...
// Latch PrologLatch
// NewExit PrologExit
// EpilogPreHeader NewPreHeader
// EpilogHeader Header
// ... ...
// EpilogLatch Latch
// Exit Exit
// Rewrite the cloned instruction operands to use the values created when the
// clone is created.
for (BasicBlock *BB : NewBlocks) {
for (Instruction &I : *BB) {
RemapInstruction(&I, VMap,
RF_NoModuleLevelChanges | RF_IgnoreMissingLocals);
}
}
if (UseEpilogRemainder) {
// Connect the epilog code to the original loop and update the
// PHI functions.
ConnectEpilog(L, ModVal, NewExit, Exit, PreHeader,
EpilogPreHeader, NewPreHeader, VMap, DT, LI,
PreserveLCSSA);
// Update counter in loop for unrolling.
// I should be multiply of Count.
IRBuilder<> B2(NewPreHeader->getTerminator());
Value *TestVal = B2.CreateSub(TripCount, ModVal, "unroll_iter");
BranchInst *LatchBR = cast<BranchInst>(Latch->getTerminator());
B2.SetInsertPoint(LatchBR);
PHINode *NewIdx = PHINode::Create(TestVal->getType(), 2, "niter",
Header->getFirstNonPHI());
Value *IdxSub =
B2.CreateSub(NewIdx, ConstantInt::get(NewIdx->getType(), 1),
NewIdx->getName() + ".nsub");
Value *IdxCmp;
if (LatchBR->getSuccessor(0) == Header)
IdxCmp = B2.CreateIsNotNull(IdxSub, NewIdx->getName() + ".ncmp");
else
IdxCmp = B2.CreateIsNull(IdxSub, NewIdx->getName() + ".ncmp");
NewIdx->addIncoming(TestVal, NewPreHeader);
NewIdx->addIncoming(IdxSub, Latch);
LatchBR->setCondition(IdxCmp);
} else {
// Connect the prolog code to the original loop and update the
// PHI functions.
ConnectProlog(L, BECount, Count, PrologExit, PreHeader, NewPreHeader,
VMap, DT, LI, PreserveLCSSA);
}
// If this loop is nested, then the loop unroller changes the code in the
// parent loop, so the Scalar Evolution pass needs to be run again.
if (Loop *ParentLoop = L->getParentLoop())
SE->forgetLoop(ParentLoop);
NumRuntimeUnrolled++;
return true;
}
bool PPCCTRLoops::convertToCTRLoop(Loop *L) {
bool MadeChange = false;
// Do not convert small short loops to CTR loop.
unsigned ConstTripCount = SE->getSmallConstantTripCount(L);
if (ConstTripCount && ConstTripCount < SmallCTRLoopThreshold) {
SmallPtrSet<const Value *, 32> EphValues;
auto AC = getAnalysis<AssumptionCacheTracker>().getAssumptionCache(
*L->getHeader()->getParent());
CodeMetrics::collectEphemeralValues(L, &AC, EphValues);
CodeMetrics Metrics;
for (BasicBlock *BB : L->blocks())
Metrics.analyzeBasicBlock(BB, *TTI, EphValues);
// 6 is an approximate latency for the mtctr instruction.
if (Metrics.NumInsts <= (6 * SchedModel.getIssueWidth()))
return false;
}
// Process nested loops first.
for (Loop::iterator I = L->begin(), E = L->end(); I != E; ++I) {
MadeChange |= convertToCTRLoop(*I);
LLVM_DEBUG(dbgs() << "Nested loop converted\n");
}
// If a nested loop has been converted, then we can't convert this loop.
if (MadeChange)
return MadeChange;
// Bail out if the loop has irreducible control flow.
LoopBlocksRPO RPOT(L);
RPOT.perform(LI);
if (containsIrreducibleCFG<const BasicBlock *>(RPOT, *LI))
return false;
#ifndef NDEBUG
// Stop trying after reaching the limit (if any).
int Limit = CTRLoopLimit;
if (Limit >= 0) {
if (Counter >= CTRLoopLimit)
return false;
Counter++;
}
#endif
// We don't want to spill/restore the counter register, and so we don't
// want to use the counter register if the loop contains calls.
for (Loop::block_iterator I = L->block_begin(), IE = L->block_end();
I != IE; ++I)
if (mightUseCTR(*I))
return MadeChange;
SmallVector<BasicBlock*, 4> ExitingBlocks;
L->getExitingBlocks(ExitingBlocks);
// If there is an exit edge known to be frequently taken,
// we should not transform this loop.
for (auto &BB : ExitingBlocks) {
Instruction *TI = BB->getTerminator();
if (!TI) continue;
if (BranchInst *BI = dyn_cast<BranchInst>(TI)) {
uint64_t TrueWeight = 0, FalseWeight = 0;
if (!BI->isConditional() ||
!BI->extractProfMetadata(TrueWeight, FalseWeight))
continue;
// If the exit path is more frequent than the loop path,
// we return here without further analysis for this loop.
bool TrueIsExit = !L->contains(BI->getSuccessor(0));
if (( TrueIsExit && FalseWeight < TrueWeight) ||
(!TrueIsExit && FalseWeight > TrueWeight))
return MadeChange;
}
}
BasicBlock *CountedExitBlock = nullptr;
const SCEV *ExitCount = nullptr;
BranchInst *CountedExitBranch = nullptr;
for (SmallVectorImpl<BasicBlock *>::iterator I = ExitingBlocks.begin(),
IE = ExitingBlocks.end(); I != IE; ++I) {
const SCEV *EC = SE->getExitCount(L, *I);
LLVM_DEBUG(dbgs() << "Exit Count for " << *L << " from block "
<< (*I)->getName() << ": " << *EC << "\n");
if (isa<SCEVCouldNotCompute>(EC))
continue;
if (const SCEVConstant *ConstEC = dyn_cast<SCEVConstant>(EC)) {
if (ConstEC->getValue()->isZero())
continue;
} else if (!SE->isLoopInvariant(EC, L))
continue;
if (SE->getTypeSizeInBits(EC->getType()) > (TM->isPPC64() ? 64 : 32))
continue;
// If this exiting block is contained in a nested loop, it is not eligible
// for insertion of the branch-and-decrement since the inner loop would
// end up messing up the value in the CTR.
if (LI->getLoopFor(*I) != L)
continue;
// We now have a loop-invariant count of loop iterations (which is not the
// constant zero) for which we know that this loop will not exit via this
// existing block.
// We need to make sure that this block will run on every loop iteration.
// For this to be true, we must dominate all blocks with backedges. Such
// blocks are in-loop predecessors to the header block.
bool NotAlways = false;
for (pred_iterator PI = pred_begin(L->getHeader()),
PIE = pred_end(L->getHeader()); PI != PIE; ++PI) {
if (!L->contains(*PI))
continue;
if (!DT->dominates(*I, *PI)) {
NotAlways = true;
break;
}
}
if (NotAlways)
continue;
// Make sure this blocks ends with a conditional branch.
Instruction *TI = (*I)->getTerminator();
if (!TI)
continue;
if (BranchInst *BI = dyn_cast<BranchInst>(TI)) {
if (!BI->isConditional())
continue;
CountedExitBranch = BI;
} else
continue;
// Note that this block may not be the loop latch block, even if the loop
// has a latch block.
CountedExitBlock = *I;
ExitCount = EC;
break;
}
if (!CountedExitBlock)
return MadeChange;
BasicBlock *Preheader = L->getLoopPreheader();
// If we don't have a preheader, then insert one. If we already have a
// preheader, then we can use it (except if the preheader contains a use of
// the CTR register because some such uses might be reordered by the
// selection DAG after the mtctr instruction).
if (!Preheader || mightUseCTR(Preheader))
Preheader = InsertPreheaderForLoop(L, DT, LI, PreserveLCSSA);
if (!Preheader)
return MadeChange;
LLVM_DEBUG(dbgs() << "Preheader for exit count: " << Preheader->getName()
<< "\n");
// Insert the count into the preheader and replace the condition used by the
// selected branch.
MadeChange = true;
SCEVExpander SCEVE(*SE, *DL, "loopcnt");
LLVMContext &C = SE->getContext();
Type *CountType = TM->isPPC64() ? Type::getInt64Ty(C) : Type::getInt32Ty(C);
if (!ExitCount->getType()->isPointerTy() &&
ExitCount->getType() != CountType)
ExitCount = SE->getZeroExtendExpr(ExitCount, CountType);
ExitCount = SE->getAddExpr(ExitCount, SE->getOne(CountType));
Value *ECValue =
SCEVE.expandCodeFor(ExitCount, CountType, Preheader->getTerminator());
IRBuilder<> CountBuilder(Preheader->getTerminator());
Module *M = Preheader->getParent()->getParent();
Function *MTCTRFunc =
Intrinsic::getDeclaration(M, Intrinsic::ppc_mtctr, CountType);
CountBuilder.CreateCall(MTCTRFunc, ECValue);
IRBuilder<> CondBuilder(CountedExitBranch);
Function *DecFunc =
Intrinsic::getDeclaration(M, Intrinsic::ppc_is_decremented_ctr_nonzero);
Value *NewCond = CondBuilder.CreateCall(DecFunc, {});
Value *OldCond = CountedExitBranch->getCondition();
CountedExitBranch->setCondition(NewCond);
// The false branch must exit the loop.
if (!L->contains(CountedExitBranch->getSuccessor(0)))
CountedExitBranch->swapSuccessors();
// The old condition may be dead now, and may have even created a dead PHI
// (the original induction variable).
RecursivelyDeleteTriviallyDeadInstructions(OldCond);
// Run through the basic blocks of the loop and see if any of them have dead
// PHIs that can be removed.
for (auto I : L->blocks())
DeleteDeadPHIs(I);
++NumCTRLoops;
return MadeChange;
}
