/// If \param [in] BB has more than one predecessor that is a conditional
/// branch, attempt to use parallel and/or for the branch condition. \returns
/// true on success.
///
/// Before:
/// ......
/// %cmp10 = fcmp une float %tmp1, %tmp2
/// br i1 %cmp1, label %if.then, label %lor.rhs
///
/// lor.rhs:
/// ......
/// %cmp11 = fcmp une float %tmp3, %tmp4
/// br i1 %cmp11, label %if.then, label %ifend
///
/// if.end: // the merge block
/// ......
///
/// if.then: // has two predecessors, both of them contains conditional branch.
/// ......
/// br label %if.end;
///
/// After:
/// ......
/// %cmp10 = fcmp une float %tmp1, %tmp2
/// ......
/// %cmp11 = fcmp une float %tmp3, %tmp4
/// %cmp12 = or i1 %cmp10, %cmp11 // parallel-or mode.
/// br i1 %cmp12, label %if.then, label %ifend
///
/// if.end:
/// ......
///
/// if.then:
/// ......
/// br label %if.end;
///
/// Current implementation handles two cases.
/// Case 1: \param BB is on the else-path.
///
/// BB1
/// / |
/// BB2 |
/// / \ |
/// BB3 \ | where, BB1, BB2 contain conditional branches.
/// \ | / BB3 contains unconditional branch.
/// \ | / BB4 corresponds to \param BB which is also the merge.
/// BB => BB4
///
///
/// Corresponding source code:
///
/// if (a == b && c == d)
/// statement; // BB3
///
/// Case 2: \param BB BB is on the then-path.
///
/// BB1
/// / |
/// | BB2
/// \ / | where BB1, BB2 contain conditional branches.
/// BB => BB3 | BB3 contains unconditiona branch and corresponds
/// \ / to \param BB. BB4 is the merge.
/// BB4
///
/// Corresponding source code:
///
/// if (a == b || c == d)
/// statement; // BB3
///
/// In both cases, \param BB is the common successor of conditional branches.
/// In Case 1, \param BB (BB4) has an unconditional branch (BB3) as
/// its predecessor. In Case 2, \param BB (BB3) only has conditional branches
/// as its predecessors.
///
bool FlattenCFGOpt::FlattenParallelAndOr(BasicBlock *BB, IRBuilder<> &Builder,
Pass *P) {
PHINode *PHI = dyn_cast<PHINode>(BB->begin());
if (PHI)
return false; // For simplicity, avoid cases containing PHI nodes.
BasicBlock *LastCondBlock = NULL;
BasicBlock *FirstCondBlock = NULL;
BasicBlock *UnCondBlock = NULL;
int Idx = -1;
// Check predecessors of \param BB.
SmallPtrSet<BasicBlock *, 16> Preds(pred_begin(BB), pred_end(BB));
for (SmallPtrSetIterator<BasicBlock *> PI = Preds.begin(), PE = Preds.end();
PI != PE; ++PI) {
BasicBlock *Pred = *PI;
BranchInst *PBI = dyn_cast<BranchInst>(Pred->getTerminator());
// All predecessors should terminate with a branch.
if (!PBI)
return false;
BasicBlock *PP = Pred->getSinglePredecessor();
if (PBI->isUnconditional()) {
// Case 1: Pred (BB3) is an unconditional block, it should
// have a single predecessor (BB2) that is also a predecessor
// of \param BB (BB4) and should not have address-taken.
// There should exist only one such unconditional
// branch among the predecessors.
if (UnCondBlock || !PP || (Preds.count(PP) == 0) ||
Pred->hasAddressTaken())
return false;
UnCondBlock = Pred;
continue;
}
// Only conditional branches are allowed beyond this point.
assert(PBI->isConditional());
// Condition's unique use should be the branch instruction.
Value *PC = PBI->getCondition();
if (!PC || !PC->hasOneUse())
return false;
if (PP && Preds.count(PP)) {
// These are internal condition blocks to be merged from, e.g.,
// BB2 in both cases.
// Should not be address-taken.
if (Pred->hasAddressTaken())
return false;
// Instructions in the internal condition blocks should be safe
// to hoist up.
for (BasicBlock::iterator BI = Pred->begin(), BE = PBI; BI != BE;) {
Instruction *CI = BI++;
if (isa<PHINode>(CI) || !isSafeToSpeculativelyExecute(CI))
return false;
}
} else {
// This is the condition block to be merged into, e.g. BB1 in
// both cases.
if (FirstCondBlock)
return false;
FirstCondBlock = Pred;
}
// Find whether BB is uniformly on the true (or false) path
// for all of its predecessors.
BasicBlock *PS1 = PBI->getSuccessor(0);
BasicBlock *PS2 = PBI->getSuccessor(1);
BasicBlock *PS = (PS1 == BB) ? PS2 : PS1;
int CIdx = (PS1 == BB) ? 0 : 1;
if (Idx == -1)
Idx = CIdx;
else if (CIdx != Idx)
return false;
// PS is the successor which is not BB. Check successors to identify
// the last conditional branch.
if (Preds.count(PS) == 0) {
// Case 2.
LastCondBlock = Pred;
} else {
// Case 1
BranchInst *BPS = dyn_cast<BranchInst>(PS->getTerminator());
if (BPS && BPS->isUnconditional()) {
// Case 1: PS(BB3) should be an unconditional branch.
LastCondBlock = Pred;
}
}
}
if (!FirstCondBlock || !LastCondBlock || (FirstCondBlock == LastCondBlock))
return false;
TerminatorInst *TBB = LastCondBlock->getTerminator();
BasicBlock *PS1 = TBB->getSuccessor(0);
BasicBlock *PS2 = TBB->getSuccessor(1);
BranchInst *PBI1 = dyn_cast<BranchInst>(PS1->getTerminator());
BranchInst *PBI2 = dyn_cast<BranchInst>(PS2->getTerminator());
// If PS1 does not jump into PS2, but PS2 jumps into PS1,
// attempt branch inversion.
if (!PBI1 || !PBI1->isUnconditional() ||
(PS1->getTerminator()->getSuccessor(0) != PS2)) {
// Check whether PS2 jumps into PS1.
if (!PBI2 || !PBI2->isUnconditional() ||
(PS2->getTerminator()->getSuccessor(0) != PS1))
return false;
// Do branch inversion.
BasicBlock *CurrBlock = LastCondBlock;
bool EverChanged = false;
while (1) {
BranchInst *BI = dyn_cast<BranchInst>(CurrBlock->getTerminator());
CmpInst *CI = dyn_cast<CmpInst>(BI->getCondition());
CmpInst::Predicate Predicate = CI->getPredicate();
// Cannonicalize icmp_ne -> icmp_eq, fcmp_one -> fcmp_oeq
if ((Predicate == CmpInst::ICMP_NE) || (Predicate == CmpInst::FCMP_ONE)) {
CI->setPredicate(ICmpInst::getInversePredicate(Predicate));
BI->swapSuccessors();
EverChanged = true;
}
if (CurrBlock == FirstCondBlock)
break;
CurrBlock = CurrBlock->getSinglePredecessor();
}
return EverChanged;
}
// PS1 must have a conditional branch.
if (!PBI1 || !PBI1->isUnconditional())
return false;
// PS2 should not contain PHI node.
PHI = dyn_cast<PHINode>(PS2->begin());
if (PHI)
return false;
// Do the transformation.
BasicBlock *CB;
BranchInst *PBI = dyn_cast<BranchInst>(FirstCondBlock->getTerminator());
bool Iteration = true;
BasicBlock *SaveInsertBB = Builder.GetInsertBlock();
BasicBlock::iterator SaveInsertPt = Builder.GetInsertPoint();
Value *PC = PBI->getCondition();
do {
CB = PBI->getSuccessor(1 - Idx);
// Delete the conditional branch.
FirstCondBlock->getInstList().pop_back();
FirstCondBlock->getInstList()
.splice(FirstCondBlock->end(), CB->getInstList());
PBI = cast<BranchInst>(FirstCondBlock->getTerminator());
Value *CC = PBI->getCondition();
// Merge conditions.
Builder.SetInsertPoint(PBI);
Value *NC;
if (Idx == 0)
// Case 2, use parallel or.
NC = Builder.CreateOr(PC, CC);
else
// Case 1, use parallel and.
NC = Builder.CreateAnd(PC, CC);
PBI->replaceUsesOfWith(CC, NC);
PC = NC;
if (CB == LastCondBlock)
Iteration = false;
// Remove internal conditional branches.
CB->dropAllReferences();
// make CB unreachable and let downstream to delete the block.
new UnreachableInst(CB->getContext(), CB);
} while (Iteration);
Builder.SetInsertPoint(SaveInsertBB, SaveInsertPt);
DEBUG(dbgs() << "Use parallel and/or in:\n" << *FirstCondBlock);
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 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;
}
