//===-- BranchFolding.cpp - Fold machine code branch instructions ---------===// // // The LLVM Compiler Infrastructure // // This file is distributed under the University of Illinois Open Source // License. See LICENSE.TXT for details. // //===----------------------------------------------------------------------===// // // This pass forwards branches to unconditional branches to make them branch // directly to the target block. This pass often results in dead MBB's, which // it then removes. // // Note that this pass must be run after register allocation, it cannot handle // SSA form. // //===----------------------------------------------------------------------===// #define DEBUG_TYPE "branchfolding" #include "llvm/CodeGen/Passes.h" #include "llvm/CodeGen/MachineModuleInfo.h" #include "llvm/CodeGen/MachineFunctionPass.h" #include "llvm/CodeGen/MachineJumpTableInfo.h" #include "llvm/CodeGen/RegisterScavenging.h" #include "llvm/Target/TargetInstrInfo.h" #include "llvm/Target/TargetMachine.h" #include "llvm/Target/TargetRegisterInfo.h" #include "llvm/Support/CommandLine.h" #include "llvm/Support/Debug.h" #include "llvm/ADT/SmallSet.h" #include "llvm/ADT/Statistic.h" #include "llvm/ADT/STLExtras.h" #include using namespace llvm; STATISTIC(NumDeadBlocks, "Number of dead blocks removed"); STATISTIC(NumBranchOpts, "Number of branches optimized"); STATISTIC(NumTailMerge , "Number of block tails merged"); static cl::opt FlagEnableTailMerge("enable-tail-merge", cl::init(cl::BOU_UNSET), cl::Hidden); // Throttle for huge numbers of predecessors (compile speed problems) static cl::opt TailMergeThreshold("tail-merge-threshold", cl::desc("Max number of predecessors to consider tail merging"), cl::init(100), cl::Hidden); namespace { struct VISIBILITY_HIDDEN BranchFolder : public MachineFunctionPass { static char ID; explicit BranchFolder(bool defaultEnableTailMerge) : MachineFunctionPass(&ID) { switch (FlagEnableTailMerge) { case cl::BOU_UNSET: EnableTailMerge = defaultEnableTailMerge; break; case cl::BOU_TRUE: EnableTailMerge = true; break; case cl::BOU_FALSE: EnableTailMerge = false; break; } } virtual bool runOnMachineFunction(MachineFunction &MF); virtual const char *getPassName() const { return "Control Flow Optimizer"; } const TargetInstrInfo *TII; MachineModuleInfo *MMI; bool MadeChange; private: // Tail Merging. bool EnableTailMerge; bool TailMergeBlocks(MachineFunction &MF); bool TryMergeBlocks(MachineBasicBlock* SuccBB, MachineBasicBlock* PredBB); void ReplaceTailWithBranchTo(MachineBasicBlock::iterator OldInst, MachineBasicBlock *NewDest); MachineBasicBlock *SplitMBBAt(MachineBasicBlock &CurMBB, MachineBasicBlock::iterator BBI1); unsigned ComputeSameTails(unsigned CurHash, unsigned minCommonTailLength); void RemoveBlocksWithHash(unsigned CurHash, MachineBasicBlock* SuccBB, MachineBasicBlock* PredBB); unsigned CreateCommonTailOnlyBlock(MachineBasicBlock *&PredBB, unsigned maxCommonTailLength); typedef std::pair MergePotentialsElt; typedef std::vector::iterator MPIterator; std::vector MergePotentials; typedef std::pair SameTailElt; std::vector SameTails; const TargetRegisterInfo *RegInfo; RegScavenger *RS; // Branch optzn. bool OptimizeBranches(MachineFunction &MF); void OptimizeBlock(MachineBasicBlock *MBB); void RemoveDeadBlock(MachineBasicBlock *MBB); bool OptimizeImpDefsBlock(MachineBasicBlock *MBB); bool CanFallThrough(MachineBasicBlock *CurBB); bool CanFallThrough(MachineBasicBlock *CurBB, bool BranchUnAnalyzable, MachineBasicBlock *TBB, MachineBasicBlock *FBB, const SmallVectorImpl &Cond); }; char BranchFolder::ID = 0; } FunctionPass *llvm::createBranchFoldingPass(bool DefaultEnableTailMerge) { return new BranchFolder(DefaultEnableTailMerge); } /// RemoveDeadBlock - Remove the specified dead machine basic block from the /// function, updating the CFG. void BranchFolder::RemoveDeadBlock(MachineBasicBlock *MBB) { assert(MBB->pred_empty() && "MBB must be dead!"); DOUT << "\nRemoving MBB: " << *MBB; MachineFunction *MF = MBB->getParent(); // drop all successors. while (!MBB->succ_empty()) MBB->removeSuccessor(MBB->succ_end()-1); // If there are any labels in the basic block, unregister them from // MachineModuleInfo. if (MMI && !MBB->empty()) { for (MachineBasicBlock::iterator I = MBB->begin(), E = MBB->end(); I != E; ++I) { if (I->isLabel()) // The label ID # is always operand #0, an immediate. MMI->InvalidateLabel(I->getOperand(0).getImm()); } } // Remove the block. MF->erase(MBB); } /// OptimizeImpDefsBlock - If a basic block is just a bunch of implicit_def /// followed by terminators, and if the implicitly defined registers are not /// used by the terminators, remove those implicit_def's. e.g. /// BB1: /// r0 = implicit_def /// r1 = implicit_def /// br /// This block can be optimized away later if the implicit instructions are /// removed. bool BranchFolder::OptimizeImpDefsBlock(MachineBasicBlock *MBB) { SmallSet ImpDefRegs; MachineBasicBlock::iterator I = MBB->begin(); while (I != MBB->end()) { if (I->getOpcode() != TargetInstrInfo::IMPLICIT_DEF) break; unsigned Reg = I->getOperand(0).getReg(); ImpDefRegs.insert(Reg); for (const unsigned *SubRegs = RegInfo->getSubRegisters(Reg); unsigned SubReg = *SubRegs; ++SubRegs) ImpDefRegs.insert(SubReg); ++I; } if (ImpDefRegs.empty()) return false; MachineBasicBlock::iterator FirstTerm = I; while (I != MBB->end()) { if (!TII->isUnpredicatedTerminator(I)) return false; // See if it uses any of the implicitly defined registers. for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i) { MachineOperand &MO = I->getOperand(i); if (!MO.isRegister() || !MO.isUse()) continue; unsigned Reg = MO.getReg(); if (ImpDefRegs.count(Reg)) return false; } ++I; } I = MBB->begin(); while (I != FirstTerm) { MachineInstr *ImpDefMI = &*I; ++I; MBB->erase(ImpDefMI); } return true; } bool BranchFolder::runOnMachineFunction(MachineFunction &MF) { TII = MF.getTarget().getInstrInfo(); if (!TII) return false; RegInfo = MF.getTarget().getRegisterInfo(); // Fix CFG. The later algorithms expect it to be right. bool EverMadeChange = false; for (MachineFunction::iterator I = MF.begin(), E = MF.end(); I != E; I++) { MachineBasicBlock *MBB = I, *TBB = 0, *FBB = 0; SmallVector Cond; if (!TII->AnalyzeBranch(*MBB, TBB, FBB, Cond)) EverMadeChange |= MBB->CorrectExtraCFGEdges(TBB, FBB, !Cond.empty()); EverMadeChange |= OptimizeImpDefsBlock(MBB); } RS = RegInfo->requiresRegisterScavenging(MF) ? new RegScavenger() : NULL; MMI = getAnalysisToUpdate(); bool MadeChangeThisIteration = true; while (MadeChangeThisIteration) { MadeChangeThisIteration = false; MadeChangeThisIteration |= TailMergeBlocks(MF); MadeChangeThisIteration |= OptimizeBranches(MF); EverMadeChange |= MadeChangeThisIteration; } // See if any jump tables have become mergable or dead as the code generator // did its thing. MachineJumpTableInfo *JTI = MF.getJumpTableInfo(); const std::vector &JTs = JTI->getJumpTables(); if (!JTs.empty()) { // Figure out how these jump tables should be merged. std::vector JTMapping; JTMapping.reserve(JTs.size()); // We always keep the 0th jump table. JTMapping.push_back(0); // Scan the jump tables, seeing if there are any duplicates. Note that this // is N^2, which should be fixed someday. for (unsigned i = 1, e = JTs.size(); i != e; ++i) JTMapping.push_back(JTI->getJumpTableIndex(JTs[i].MBBs)); // If a jump table was merge with another one, walk the function rewriting // references to jump tables to reference the new JT ID's. Keep track of // whether we see a jump table idx, if not, we can delete the JT. BitVector JTIsLive(JTs.size()); for (MachineFunction::iterator BB = MF.begin(), E = MF.end(); BB != E; ++BB) { for (MachineBasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) for (unsigned op = 0, e = I->getNumOperands(); op != e; ++op) { MachineOperand &Op = I->getOperand(op); if (!Op.isJumpTableIndex()) continue; unsigned NewIdx = JTMapping[Op.getIndex()]; Op.setIndex(NewIdx); // Remember that this JT is live. JTIsLive.set(NewIdx); } } // Finally, remove dead jump tables. This happens either because the // indirect jump was unreachable (and thus deleted) or because the jump // table was merged with some other one. for (unsigned i = 0, e = JTIsLive.size(); i != e; ++i) if (!JTIsLive.test(i)) { JTI->RemoveJumpTable(i); EverMadeChange = true; } } delete RS; return EverMadeChange; } //===----------------------------------------------------------------------===// // Tail Merging of Blocks //===----------------------------------------------------------------------===// /// HashMachineInstr - Compute a hash value for MI and its operands. static unsigned HashMachineInstr(const MachineInstr *MI) { unsigned Hash = MI->getOpcode(); for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) { const MachineOperand &Op = MI->getOperand(i); // Merge in bits from the operand if easy. unsigned OperandHash = 0; switch (Op.getType()) { case MachineOperand::MO_Register: OperandHash = Op.getReg(); break; case MachineOperand::MO_Immediate: OperandHash = Op.getImm(); break; case MachineOperand::MO_MachineBasicBlock: OperandHash = Op.getMBB()->getNumber(); break; case MachineOperand::MO_FrameIndex: case MachineOperand::MO_ConstantPoolIndex: case MachineOperand::MO_JumpTableIndex: OperandHash = Op.getIndex(); break; case MachineOperand::MO_GlobalAddress: case MachineOperand::MO_ExternalSymbol: // Global address / external symbol are too hard, don't bother, but do // pull in the offset. OperandHash = Op.getOffset(); break; default: break; } Hash += ((OperandHash << 3) | Op.getType()) << (i&31); } return Hash; } /// HashEndOfMBB - Hash the last few instructions in the MBB. For blocks /// with no successors, we hash two instructions, because cross-jumping /// only saves code when at least two instructions are removed (since a /// branch must be inserted). For blocks with a successor, one of the /// two blocks to be tail-merged will end with a branch already, so /// it gains to cross-jump even for one instruction. static unsigned HashEndOfMBB(const MachineBasicBlock *MBB, unsigned minCommonTailLength) { MachineBasicBlock::const_iterator I = MBB->end(); if (I == MBB->begin()) return 0; // Empty MBB. --I; unsigned Hash = HashMachineInstr(I); if (I == MBB->begin() || minCommonTailLength == 1) return Hash; // Single instr MBB. --I; // Hash in the second-to-last instruction. Hash ^= HashMachineInstr(I) << 2; return Hash; } /// ComputeCommonTailLength - Given two machine basic blocks, compute the number /// of instructions they actually have in common together at their end. Return /// iterators for the first shared instruction in each block. static unsigned ComputeCommonTailLength(MachineBasicBlock *MBB1, MachineBasicBlock *MBB2, MachineBasicBlock::iterator &I1, MachineBasicBlock::iterator &I2) { I1 = MBB1->end(); I2 = MBB2->end(); unsigned TailLen = 0; while (I1 != MBB1->begin() && I2 != MBB2->begin()) { --I1; --I2; if (!I1->isIdenticalTo(I2) || // FIXME: This check is dubious. It's used to get around a problem where // people incorrectly expect inline asm directives to remain in the same // relative order. This is untenable because normal compiler // optimizations (like this one) may reorder and/or merge these // directives. I1->getOpcode() == TargetInstrInfo::INLINEASM) { ++I1; ++I2; break; } ++TailLen; } return TailLen; } /// ReplaceTailWithBranchTo - Delete the instruction OldInst and everything /// after it, replacing it with an unconditional branch to NewDest. This /// returns true if OldInst's block is modified, false if NewDest is modified. void BranchFolder::ReplaceTailWithBranchTo(MachineBasicBlock::iterator OldInst, MachineBasicBlock *NewDest) { MachineBasicBlock *OldBB = OldInst->getParent(); // Remove all the old successors of OldBB from the CFG. while (!OldBB->succ_empty()) OldBB->removeSuccessor(OldBB->succ_begin()); // Remove all the dead instructions from the end of OldBB. OldBB->erase(OldInst, OldBB->end()); // If OldBB isn't immediately before OldBB, insert a branch to it. if (++MachineFunction::iterator(OldBB) != MachineFunction::iterator(NewDest)) TII->InsertBranch(*OldBB, NewDest, 0, SmallVector()); OldBB->addSuccessor(NewDest); ++NumTailMerge; } /// SplitMBBAt - Given a machine basic block and an iterator into it, split the /// MBB so that the part before the iterator falls into the part starting at the /// iterator. This returns the new MBB. MachineBasicBlock *BranchFolder::SplitMBBAt(MachineBasicBlock &CurMBB, MachineBasicBlock::iterator BBI1) { MachineFunction &MF = *CurMBB.getParent(); // Create the fall-through block. MachineFunction::iterator MBBI = &CurMBB; MachineBasicBlock *NewMBB =MF.CreateMachineBasicBlock(CurMBB.getBasicBlock()); CurMBB.getParent()->insert(++MBBI, NewMBB); // Move all the successors of this block to the specified block. NewMBB->transferSuccessors(&CurMBB); // Add an edge from CurMBB to NewMBB for the fall-through. CurMBB.addSuccessor(NewMBB); // Splice the code over. NewMBB->splice(NewMBB->end(), &CurMBB, BBI1, CurMBB.end()); // For targets that use the register scavenger, we must maintain LiveIns. if (RS) { RS->enterBasicBlock(&CurMBB); if (!CurMBB.empty()) RS->forward(prior(CurMBB.end())); BitVector RegsLiveAtExit(RegInfo->getNumRegs()); RS->getRegsUsed(RegsLiveAtExit, false); for (unsigned int i=0, e=RegInfo->getNumRegs(); i!=e; i++) if (RegsLiveAtExit[i]) NewMBB->addLiveIn(i); } return NewMBB; } /// EstimateRuntime - Make a rough estimate for how long it will take to run /// the specified code. static unsigned EstimateRuntime(MachineBasicBlock::iterator I, MachineBasicBlock::iterator E) { unsigned Time = 0; for (; I != E; ++I) { const TargetInstrDesc &TID = I->getDesc(); if (TID.isCall()) Time += 10; else if (TID.isSimpleLoad() || TID.mayStore()) Time += 2; else ++Time; } return Time; } // CurMBB needs to add an unconditional branch to SuccMBB (we removed these // branches temporarily for tail merging). In the case where CurMBB ends // with a conditional branch to the next block, optimize by reversing the // test and conditionally branching to SuccMBB instead. static void FixTail(MachineBasicBlock* CurMBB, MachineBasicBlock *SuccBB, const TargetInstrInfo *TII) { MachineFunction *MF = CurMBB->getParent(); MachineFunction::iterator I = next(MachineFunction::iterator(CurMBB)); MachineBasicBlock *TBB = 0, *FBB = 0; SmallVector Cond; if (I != MF->end() && !TII->AnalyzeBranch(*CurMBB, TBB, FBB, Cond)) { MachineBasicBlock *NextBB = I; if (TBB == NextBB && !Cond.empty() && !FBB) { if (!TII->ReverseBranchCondition(Cond)) { TII->RemoveBranch(*CurMBB); TII->InsertBranch(*CurMBB, SuccBB, NULL, Cond); return; } } } TII->InsertBranch(*CurMBB, SuccBB, NULL, SmallVector()); } static bool MergeCompare(const std::pair &p, const std::pair &q) { if (p.first < q.first) return true; else if (p.first > q.first) return false; else if (p.second->getNumber() < q.second->getNumber()) return true; else if (p.second->getNumber() > q.second->getNumber()) return false; else { // _GLIBCXX_DEBUG checks strict weak ordering, which involves comparing // an object with itself. #ifndef _GLIBCXX_DEBUG assert(0 && "Predecessor appears twice"); #endif return(false); } } /// ComputeSameTails - Look through all the blocks in MergePotentials that have /// hash CurHash (guaranteed to match the last element). Build the vector /// SameTails of all those that have the (same) largest number of instructions /// in common of any pair of these blocks. SameTails entries contain an /// iterator into MergePotentials (from which the MachineBasicBlock can be /// found) and a MachineBasicBlock::iterator into that MBB indicating the /// instruction where the matching code sequence begins. /// Order of elements in SameTails is the reverse of the order in which /// those blocks appear in MergePotentials (where they are not necessarily /// consecutive). unsigned BranchFolder::ComputeSameTails(unsigned CurHash, unsigned minCommonTailLength) { unsigned maxCommonTailLength = 0U; SameTails.clear(); MachineBasicBlock::iterator TrialBBI1, TrialBBI2; MPIterator HighestMPIter = prior(MergePotentials.end()); for (MPIterator CurMPIter = prior(MergePotentials.end()), B = MergePotentials.begin(); CurMPIter!=B && CurMPIter->first==CurHash; --CurMPIter) { for (MPIterator I = prior(CurMPIter); I->first==CurHash ; --I) { unsigned CommonTailLen = ComputeCommonTailLength( CurMPIter->second, I->second, TrialBBI1, TrialBBI2); // If we will have to split a block, there should be at least // minCommonTailLength instructions in common; if not, at worst // we will be replacing a fallthrough into the common tail with a // branch, which at worst breaks even with falling through into // the duplicated common tail, so 1 instruction in common is enough. // We will always pick a block we do not have to split as the common // tail if there is one. // (Empty blocks will get forwarded and need not be considered.) if (CommonTailLen >= minCommonTailLength || (CommonTailLen > 0 && (TrialBBI1==CurMPIter->second->begin() || TrialBBI2==I->second->begin()))) { if (CommonTailLen > maxCommonTailLength) { SameTails.clear(); maxCommonTailLength = CommonTailLen; HighestMPIter = CurMPIter; SameTails.push_back(std::make_pair(CurMPIter, TrialBBI1)); } if (HighestMPIter == CurMPIter && CommonTailLen == maxCommonTailLength) SameTails.push_back(std::make_pair(I, TrialBBI2)); } if (I==B) break; } } return maxCommonTailLength; } /// RemoveBlocksWithHash - Remove all blocks with hash CurHash from /// MergePotentials, restoring branches at ends of blocks as appropriate. void BranchFolder::RemoveBlocksWithHash(unsigned CurHash, MachineBasicBlock* SuccBB, MachineBasicBlock* PredBB) { MPIterator CurMPIter, B; for (CurMPIter = prior(MergePotentials.end()), B = MergePotentials.begin(); CurMPIter->first==CurHash; --CurMPIter) { // Put the unconditional branch back, if we need one. MachineBasicBlock *CurMBB = CurMPIter->second; if (SuccBB && CurMBB != PredBB) FixTail(CurMBB, SuccBB, TII); if (CurMPIter==B) break; } if (CurMPIter->first!=CurHash) CurMPIter++; MergePotentials.erase(CurMPIter, MergePotentials.end()); } /// CreateCommonTailOnlyBlock - None of the blocks to be tail-merged consist /// only of the common tail. Create a block that does by splitting one. unsigned BranchFolder::CreateCommonTailOnlyBlock(MachineBasicBlock *&PredBB, unsigned maxCommonTailLength) { unsigned i, commonTailIndex; unsigned TimeEstimate = ~0U; for (i=0, commonTailIndex=0; isecond==PredBB) { commonTailIndex = i; break; } // Otherwise, make a (fairly bogus) choice based on estimate of // how long it will take the various blocks to execute. unsigned t = EstimateRuntime(SameTails[i].first->second->begin(), SameTails[i].second); if (t<=TimeEstimate) { TimeEstimate = t; commonTailIndex = i; } } MachineBasicBlock::iterator BBI = SameTails[commonTailIndex].second; MachineBasicBlock *MBB = SameTails[commonTailIndex].first->second; DOUT << "\nSplitting " << MBB->getNumber() << ", size " << maxCommonTailLength; MachineBasicBlock *newMBB = SplitMBBAt(*MBB, BBI); SameTails[commonTailIndex].first->second = newMBB; SameTails[commonTailIndex].second = newMBB->begin(); // If we split PredBB, newMBB is the new predecessor. if (PredBB==MBB) PredBB = newMBB; return commonTailIndex; } // See if any of the blocks in MergePotentials (which all have a common single // successor, or all have no successor) can be tail-merged. If there is a // successor, any blocks in MergePotentials that are not tail-merged and // are not immediately before Succ must have an unconditional branch to // Succ added (but the predecessor/successor lists need no adjustment). // The lone predecessor of Succ that falls through into Succ, // if any, is given in PredBB. bool BranchFolder::TryMergeBlocks(MachineBasicBlock *SuccBB, MachineBasicBlock* PredBB) { // It doesn't make sense to save a single instruction since tail merging // will add a jump. // FIXME: Ask the target to provide the threshold? unsigned minCommonTailLength = (SuccBB ? 1 : 2) + 1; MadeChange = false; DOUT << "\nTryMergeBlocks " << MergePotentials.size(); // Sort by hash value so that blocks with identical end sequences sort // together. std::stable_sort(MergePotentials.begin(), MergePotentials.end(),MergeCompare); // Walk through equivalence sets looking for actual exact matches. while (MergePotentials.size() > 1) { unsigned CurHash = prior(MergePotentials.end())->first; // Build SameTails, identifying the set of blocks with this hash code // and with the maximum number of instructions in common. unsigned maxCommonTailLength = ComputeSameTails(CurHash, minCommonTailLength); // If we didn't find any pair that has at least minCommonTailLength // instructions in common, remove all blocks with this hash code and retry. if (SameTails.empty()) { RemoveBlocksWithHash(CurHash, SuccBB, PredBB); continue; } // If one of the blocks is the entire common tail (and not the entry // block, which we can't jump to), we can treat all blocks with this same // tail at once. Use PredBB if that is one of the possibilities, as that // will not introduce any extra branches. MachineBasicBlock *EntryBB = MergePotentials.begin()->second-> getParent()->begin(); unsigned int commonTailIndex, i; for (commonTailIndex=SameTails.size(), i=0; isecond; if (MBB->begin() == SameTails[i].second && MBB != EntryBB) { commonTailIndex = i; if (MBB==PredBB) break; } } if (commonTailIndex==SameTails.size()) { // None of the blocks consist entirely of the common tail. // Split a block so that one does. commonTailIndex = CreateCommonTailOnlyBlock(PredBB, maxCommonTailLength); } MachineBasicBlock *MBB = SameTails[commonTailIndex].first->second; // MBB is common tail. Adjust all other BB's to jump to this one. // Traversal must be forwards so erases work. DOUT << "\nUsing common tail " << MBB->getNumber() << " for "; for (unsigned int i=0; isecond->getNumber() << ","; // Hack the end off BB i, making it jump to BB commonTailIndex instead. ReplaceTailWithBranchTo(SameTails[i].second, MBB); // BB i is no longer a predecessor of SuccBB; remove it from the worklist. MergePotentials.erase(SameTails[i].first); } DOUT << "\n"; // We leave commonTailIndex in the worklist in case there are other blocks // that match it with a smaller number of instructions. MadeChange = true; } return MadeChange; } bool BranchFolder::TailMergeBlocks(MachineFunction &MF) { if (!EnableTailMerge) return false; MadeChange = false; // First find blocks with no successors. MergePotentials.clear(); for (MachineFunction::iterator I = MF.begin(), E = MF.end(); I != E; ++I) { if (I->succ_empty()) MergePotentials.push_back(std::make_pair(HashEndOfMBB(I, 2U), I)); } // See if we can do any tail merging on those. if (MergePotentials.size() < TailMergeThreshold && MergePotentials.size() >= 2) MadeChange |= TryMergeBlocks(NULL, NULL); // Look at blocks (IBB) with multiple predecessors (PBB). // We change each predecessor to a canonical form, by // (1) temporarily removing any unconditional branch from the predecessor // to IBB, and // (2) alter conditional branches so they branch to the other block // not IBB; this may require adding back an unconditional branch to IBB // later, where there wasn't one coming in. E.g. // Bcc IBB // fallthrough to QBB // here becomes // Bncc QBB // with a conceptual B to IBB after that, which never actually exists. // With those changes, we see whether the predecessors' tails match, // and merge them if so. We change things out of canonical form and // back to the way they were later in the process. (OptimizeBranches // would undo some of this, but we can't use it, because we'd get into // a compile-time infinite loop repeatedly doing and undoing the same // transformations.) for (MachineFunction::iterator I = MF.begin(), E = MF.end(); I != E; ++I) { if (I->pred_size() >= 2 && I->pred_size() < TailMergeThreshold) { MachineBasicBlock *IBB = I; MachineBasicBlock *PredBB = prior(I); MergePotentials.clear(); for (MachineBasicBlock::pred_iterator P = I->pred_begin(), E2 = I->pred_end(); P != E2; ++P) { MachineBasicBlock* PBB = *P; // Skip blocks that loop to themselves, can't tail merge these. if (PBB==IBB) continue; MachineBasicBlock *TBB = 0, *FBB = 0; SmallVector Cond; if (!TII->AnalyzeBranch(*PBB, TBB, FBB, Cond)) { // Failing case: IBB is the target of a cbr, and // we cannot reverse the branch. SmallVector NewCond(Cond); if (!Cond.empty() && TBB==IBB) { if (TII->ReverseBranchCondition(NewCond)) continue; // This is the QBB case described above if (!FBB) FBB = next(MachineFunction::iterator(PBB)); } // Failing case: the only way IBB can be reached from PBB is via // exception handling. Happens for landing pads. Would be nice // to have a bit in the edge so we didn't have to do all this. if (IBB->isLandingPad()) { MachineFunction::iterator IP = PBB; IP++; MachineBasicBlock* PredNextBB = NULL; if (IP!=MF.end()) PredNextBB = IP; if (TBB==NULL) { if (IBB!=PredNextBB) // fallthrough continue; } else if (FBB) { if (TBB!=IBB && FBB!=IBB) // cbr then ubr continue; } else if (Cond.empty()) { if (TBB!=IBB) // ubr continue; } else { if (TBB!=IBB && IBB!=PredNextBB) // cbr continue; } } // Remove the unconditional branch at the end, if any. if (TBB && (Cond.empty() || FBB)) { TII->RemoveBranch(*PBB); if (!Cond.empty()) // reinsert conditional branch only, for now TII->InsertBranch(*PBB, (TBB==IBB) ? FBB : TBB, 0, NewCond); } MergePotentials.push_back(std::make_pair(HashEndOfMBB(PBB, 1U), *P)); } } if (MergePotentials.size() >= 2) MadeChange |= TryMergeBlocks(I, PredBB); // Reinsert an unconditional branch if needed. // The 1 below can occur as a result of removing blocks in TryMergeBlocks. PredBB = prior(I); // this may have been changed in TryMergeBlocks if (MergePotentials.size()==1 && MergePotentials.begin()->second != PredBB) FixTail(MergePotentials.begin()->second, I, TII); } } return MadeChange; } //===----------------------------------------------------------------------===// // Branch Optimization //===----------------------------------------------------------------------===// bool BranchFolder::OptimizeBranches(MachineFunction &MF) { MadeChange = false; // Make sure blocks are numbered in order MF.RenumberBlocks(); for (MachineFunction::iterator I = ++MF.begin(), E = MF.end(); I != E; ) { MachineBasicBlock *MBB = I++; OptimizeBlock(MBB); // If it is dead, remove it. if (MBB->pred_empty()) { RemoveDeadBlock(MBB); MadeChange = true; ++NumDeadBlocks; } } return MadeChange; } /// CanFallThrough - Return true if the specified block (with the specified /// branch condition) can implicitly transfer control to the block after it by /// falling off the end of it. This should return false if it can reach the /// block after it, but it uses an explicit branch to do so (e.g. a table jump). /// /// True is a conservative answer. /// bool BranchFolder::CanFallThrough(MachineBasicBlock *CurBB, bool BranchUnAnalyzable, MachineBasicBlock *TBB, MachineBasicBlock *FBB, const SmallVectorImpl &Cond) { MachineFunction::iterator Fallthrough = CurBB; ++Fallthrough; // If FallthroughBlock is off the end of the function, it can't fall through. if (Fallthrough == CurBB->getParent()->end()) return false; // If FallthroughBlock isn't a successor of CurBB, no fallthrough is possible. if (!CurBB->isSuccessor(Fallthrough)) return false; // If we couldn't analyze the branch, assume it could fall through. if (BranchUnAnalyzable) return true; // If there is no branch, control always falls through. if (TBB == 0) return true; // If there is some explicit branch to the fallthrough block, it can obviously // reach, even though the branch should get folded to fall through implicitly. if (MachineFunction::iterator(TBB) == Fallthrough || MachineFunction::iterator(FBB) == Fallthrough) return true; // If it's an unconditional branch to some block not the fall through, it // doesn't fall through. if (Cond.empty()) return false; // Otherwise, if it is conditional and has no explicit false block, it falls // through. return FBB == 0; } /// CanFallThrough - Return true if the specified can implicitly transfer /// control to the block after it by falling off the end of it. This should /// return false if it can reach the block after it, but it uses an explicit /// branch to do so (e.g. a table jump). /// /// True is a conservative answer. /// bool BranchFolder::CanFallThrough(MachineBasicBlock *CurBB) { MachineBasicBlock *TBB = 0, *FBB = 0; SmallVector Cond; bool CurUnAnalyzable = TII->AnalyzeBranch(*CurBB, TBB, FBB, Cond); return CanFallThrough(CurBB, CurUnAnalyzable, TBB, FBB, Cond); } /// IsBetterFallthrough - Return true if it would be clearly better to /// fall-through to MBB1 than to fall through into MBB2. This has to return /// a strict ordering, returning true for both (MBB1,MBB2) and (MBB2,MBB1) will /// result in infinite loops. static bool IsBetterFallthrough(MachineBasicBlock *MBB1, MachineBasicBlock *MBB2) { // Right now, we use a simple heuristic. If MBB2 ends with a call, and // MBB1 doesn't, we prefer to fall through into MBB1. This allows us to // optimize branches that branch to either a return block or an assert block // into a fallthrough to the return. if (MBB1->empty() || MBB2->empty()) return false; // If there is a clear successor ordering we make sure that one block // will fall through to the next if (MBB1->isSuccessor(MBB2)) return true; if (MBB2->isSuccessor(MBB1)) return false; MachineInstr *MBB1I = --MBB1->end(); MachineInstr *MBB2I = --MBB2->end(); return MBB2I->getDesc().isCall() && !MBB1I->getDesc().isCall(); } /// OptimizeBlock - Analyze and optimize control flow related to the specified /// block. This is never called on the entry block. void BranchFolder::OptimizeBlock(MachineBasicBlock *MBB) { MachineFunction::iterator FallThrough = MBB; ++FallThrough; // If this block is empty, make everyone use its fall-through, not the block // explicitly. Landing pads should not do this since the landing-pad table // points to this block. if (MBB->empty() && !MBB->isLandingPad()) { // Dead block? Leave for cleanup later. if (MBB->pred_empty()) return; if (FallThrough == MBB->getParent()->end()) { // TODO: Simplify preds to not branch here if possible! } else { // Rewrite all predecessors of the old block to go to the fallthrough // instead. while (!MBB->pred_empty()) { MachineBasicBlock *Pred = *(MBB->pred_end()-1); Pred->ReplaceUsesOfBlockWith(MBB, FallThrough); } // If MBB was the target of a jump table, update jump tables to go to the // fallthrough instead. MBB->getParent()->getJumpTableInfo()-> ReplaceMBBInJumpTables(MBB, FallThrough); MadeChange = true; } return; } // Check to see if we can simplify the terminator of the block before this // one. MachineBasicBlock &PrevBB = *prior(MachineFunction::iterator(MBB)); MachineBasicBlock *PriorTBB = 0, *PriorFBB = 0; SmallVector PriorCond; bool PriorUnAnalyzable = TII->AnalyzeBranch(PrevBB, PriorTBB, PriorFBB, PriorCond); if (!PriorUnAnalyzable) { // If the CFG for the prior block has extra edges, remove them. MadeChange |= PrevBB.CorrectExtraCFGEdges(PriorTBB, PriorFBB, !PriorCond.empty()); // If the previous branch is conditional and both conditions go to the same // destination, remove the branch, replacing it with an unconditional one or // a fall-through. if (PriorTBB && PriorTBB == PriorFBB) { TII->RemoveBranch(PrevBB); PriorCond.clear(); if (PriorTBB != MBB) TII->InsertBranch(PrevBB, PriorTBB, 0, PriorCond); MadeChange = true; ++NumBranchOpts; return OptimizeBlock(MBB); } // If the previous branch *only* branches to *this* block (conditional or // not) remove the branch. if (PriorTBB == MBB && PriorFBB == 0) { TII->RemoveBranch(PrevBB); MadeChange = true; ++NumBranchOpts; return OptimizeBlock(MBB); } // If the prior block branches somewhere else on the condition and here if // the condition is false, remove the uncond second branch. if (PriorFBB == MBB) { TII->RemoveBranch(PrevBB); TII->InsertBranch(PrevBB, PriorTBB, 0, PriorCond); MadeChange = true; ++NumBranchOpts; return OptimizeBlock(MBB); } // If the prior block branches here on true and somewhere else on false, and // if the branch condition is reversible, reverse the branch to create a // fall-through. if (PriorTBB == MBB) { SmallVector NewPriorCond(PriorCond); if (!TII->ReverseBranchCondition(NewPriorCond)) { TII->RemoveBranch(PrevBB); TII->InsertBranch(PrevBB, PriorFBB, 0, NewPriorCond); MadeChange = true; ++NumBranchOpts; return OptimizeBlock(MBB); } } // If this block doesn't fall through (e.g. it ends with an uncond branch or // has no successors) and if the pred falls through into this block, and if // it would otherwise fall through into the block after this, move this // block to the end of the function. // // We consider it more likely that execution will stay in the function (e.g. // due to loops) than it is to exit it. This asserts in loops etc, moving // the assert condition out of the loop body. if (!PriorCond.empty() && PriorFBB == 0 && MachineFunction::iterator(PriorTBB) == FallThrough && !CanFallThrough(MBB)) { bool DoTransform = true; // We have to be careful that the succs of PredBB aren't both no-successor // blocks. If neither have successors and if PredBB is the second from // last block in the function, we'd just keep swapping the two blocks for // last. Only do the swap if one is clearly better to fall through than // the other. if (FallThrough == --MBB->getParent()->end() && !IsBetterFallthrough(PriorTBB, MBB)) DoTransform = false; // We don't want to do this transformation if we have control flow like: // br cond BB2 // BB1: // .. // jmp BBX // BB2: // .. // ret // // In this case, we could actually be moving the return block *into* a // loop! if (DoTransform && !MBB->succ_empty() && (!CanFallThrough(PriorTBB) || PriorTBB->empty())) DoTransform = false; if (DoTransform) { // Reverse the branch so we will fall through on the previous true cond. SmallVector NewPriorCond(PriorCond); if (!TII->ReverseBranchCondition(NewPriorCond)) { DOUT << "\nMoving MBB: " << *MBB; DOUT << "To make fallthrough to: " << *PriorTBB << "\n"; TII->RemoveBranch(PrevBB); TII->InsertBranch(PrevBB, MBB, 0, NewPriorCond); // Move this block to the end of the function. MBB->moveAfter(--MBB->getParent()->end()); MadeChange = true; ++NumBranchOpts; return; } } } } // Analyze the branch in the current block. MachineBasicBlock *CurTBB = 0, *CurFBB = 0; SmallVector CurCond; bool CurUnAnalyzable = TII->AnalyzeBranch(*MBB, CurTBB, CurFBB, CurCond); if (!CurUnAnalyzable) { // If the CFG for the prior block has extra edges, remove them. MadeChange |= MBB->CorrectExtraCFGEdges(CurTBB, CurFBB, !CurCond.empty()); // If this is a two-way branch, and the FBB branches to this block, reverse // the condition so the single-basic-block loop is faster. Instead of: // Loop: xxx; jcc Out; jmp Loop // we want: // Loop: xxx; jncc Loop; jmp Out if (CurTBB && CurFBB && CurFBB == MBB && CurTBB != MBB) { SmallVector NewCond(CurCond); if (!TII->ReverseBranchCondition(NewCond)) { TII->RemoveBranch(*MBB); TII->InsertBranch(*MBB, CurFBB, CurTBB, NewCond); MadeChange = true; ++NumBranchOpts; return OptimizeBlock(MBB); } } // If this branch is the only thing in its block, see if we can forward // other blocks across it. if (CurTBB && CurCond.empty() && CurFBB == 0 && MBB->begin()->getDesc().isBranch() && CurTBB != MBB) { // This block may contain just an unconditional branch. Because there can // be 'non-branch terminators' in the block, try removing the branch and // then seeing if the block is empty. TII->RemoveBranch(*MBB); // If this block is just an unconditional branch to CurTBB, we can // usually completely eliminate the block. The only case we cannot // completely eliminate the block is when the block before this one // falls through into MBB and we can't understand the prior block's branch // condition. if (MBB->empty()) { bool PredHasNoFallThrough = TII->BlockHasNoFallThrough(PrevBB); if (PredHasNoFallThrough || !PriorUnAnalyzable || !PrevBB.isSuccessor(MBB)) { // If the prior block falls through into us, turn it into an // explicit branch to us to make updates simpler. if (!PredHasNoFallThrough && PrevBB.isSuccessor(MBB) && PriorTBB != MBB && PriorFBB != MBB) { if (PriorTBB == 0) { assert(PriorCond.empty() && PriorFBB == 0 && "Bad branch analysis"); PriorTBB = MBB; } else { assert(PriorFBB == 0 && "Machine CFG out of date!"); PriorFBB = MBB; } TII->RemoveBranch(PrevBB); TII->InsertBranch(PrevBB, PriorTBB, PriorFBB, PriorCond); } // Iterate through all the predecessors, revectoring each in-turn. size_t PI = 0; bool DidChange = false; bool HasBranchToSelf = false; while(PI != MBB->pred_size()) { MachineBasicBlock *PMBB = *(MBB->pred_begin() + PI); if (PMBB == MBB) { // If this block has an uncond branch to itself, leave it. ++PI; HasBranchToSelf = true; } else { DidChange = true; PMBB->ReplaceUsesOfBlockWith(MBB, CurTBB); } } // Change any jumptables to go to the new MBB. MBB->getParent()->getJumpTableInfo()-> ReplaceMBBInJumpTables(MBB, CurTBB); if (DidChange) { ++NumBranchOpts; MadeChange = true; if (!HasBranchToSelf) return; } } } // Add the branch back if the block is more than just an uncond branch. TII->InsertBranch(*MBB, CurTBB, 0, CurCond); } } // If the prior block doesn't fall through into this block, and if this // block doesn't fall through into some other block, see if we can find a // place to move this block where a fall-through will happen. if (!CanFallThrough(&PrevBB, PriorUnAnalyzable, PriorTBB, PriorFBB, PriorCond)) { // Now we know that there was no fall-through into this block, check to // see if it has a fall-through into its successor. bool CurFallsThru = CanFallThrough(MBB, CurUnAnalyzable, CurTBB, CurFBB, CurCond); if (!MBB->isLandingPad()) { // Check all the predecessors of this block. If one of them has no fall // throughs, move this block right after it. for (MachineBasicBlock::pred_iterator PI = MBB->pred_begin(), E = MBB->pred_end(); PI != E; ++PI) { // Analyze the branch at the end of the pred. MachineBasicBlock *PredBB = *PI; MachineFunction::iterator PredFallthrough = PredBB; ++PredFallthrough; if (PredBB != MBB && !CanFallThrough(PredBB) && (!CurFallsThru || !CurTBB || !CurFBB) && (!CurFallsThru || MBB->getNumber() >= PredBB->getNumber())) { // If the current block doesn't fall through, just move it. // If the current block can fall through and does not end with a // conditional branch, we need to append an unconditional jump to // the (current) next block. To avoid a possible compile-time // infinite loop, move blocks only backward in this case. // Also, if there are already 2 branches here, we cannot add a third; // this means we have the case // Bcc next // B elsewhere // next: if (CurFallsThru) { MachineBasicBlock *NextBB = next(MachineFunction::iterator(MBB)); CurCond.clear(); TII->InsertBranch(*MBB, NextBB, 0, CurCond); } MBB->moveAfter(PredBB); MadeChange = true; return OptimizeBlock(MBB); } } } if (!CurFallsThru) { // Check all successors to see if we can move this block before it. for (MachineBasicBlock::succ_iterator SI = MBB->succ_begin(), E = MBB->succ_end(); SI != E; ++SI) { // Analyze the branch at the end of the block before the succ. MachineBasicBlock *SuccBB = *SI; MachineFunction::iterator SuccPrev = SuccBB; --SuccPrev; std::vector SuccPrevCond; // If this block doesn't already fall-through to that successor, and if // the succ doesn't already have a block that can fall through into it, // and if the successor isn't an EH destination, we can arrange for the // fallthrough to happen. if (SuccBB != MBB && !CanFallThrough(SuccPrev) && !SuccBB->isLandingPad()) { MBB->moveBefore(SuccBB); MadeChange = true; return OptimizeBlock(MBB); } } // Okay, there is no really great place to put this block. If, however, // the block before this one would be a fall-through if this block were // removed, move this block to the end of the function. if (FallThrough != MBB->getParent()->end() && PrevBB.isSuccessor(FallThrough)) { MBB->moveAfter(--MBB->getParent()->end()); MadeChange = true; return; } } } }