X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FTransforms%2FUtils%2FLoopSimplify.cpp;h=f7787dafd5bff575dfed653a75385a8134ff1d99;hb=805a83c0419aa453b78b2a062f46dc7f72137f79;hp=e25ff90b47798584280a6264df2f17b2a1da53dd;hpb=4ee451de366474b9c228b4e5fa573795a715216d;p=oota-llvm.git diff --git a/lib/Transforms/Utils/LoopSimplify.cpp b/lib/Transforms/Utils/LoopSimplify.cpp index e25ff90b477..f7787dafd5b 100644 --- a/lib/Transforms/Utils/LoopSimplify.cpp +++ b/lib/Transforms/Utils/LoopSimplify.cpp @@ -23,6 +23,11 @@ // // This pass also guarantees that loops will have exactly one backedge. // +// Indirectbr instructions introduce several complications. If the loop +// contains or is entered by an indirectbr instruction, it may not be possible +// to transform the loop and make these guarantees. Client code should check +// that these conditions are true before relying on them. +// // Note that the simplifycfg pass will clean up blocks which are split out but // end up being unnecessary, so usage of this pass should not pessimize // generated code. @@ -32,430 +37,188 @@ // //===----------------------------------------------------------------------===// -#define DEBUG_TYPE "loopsimplify" #include "llvm/Transforms/Scalar.h" -#include "llvm/Constants.h" -#include "llvm/Instructions.h" -#include "llvm/Function.h" -#include "llvm/Type.h" -#include "llvm/Analysis/AliasAnalysis.h" -#include "llvm/Analysis/Dominators.h" -#include "llvm/Analysis/LoopInfo.h" -#include "llvm/Support/CFG.h" -#include "llvm/Support/Compiler.h" +#include "llvm/ADT/DepthFirstIterator.h" #include "llvm/ADT/SetOperations.h" #include "llvm/ADT/SetVector.h" +#include "llvm/ADT/SmallVector.h" #include "llvm/ADT/Statistic.h" -#include "llvm/ADT/DepthFirstIterator.h" +#include "llvm/Analysis/AliasAnalysis.h" +#include "llvm/Analysis/DependenceAnalysis.h" +#include "llvm/Analysis/InstructionSimplify.h" +#include "llvm/Analysis/LoopInfo.h" +#include "llvm/Analysis/ScalarEvolution.h" +#include "llvm/IR/CFG.h" +#include "llvm/IR/Constants.h" +#include "llvm/IR/Dominators.h" +#include "llvm/IR/Function.h" +#include "llvm/IR/Instructions.h" +#include "llvm/IR/IntrinsicInst.h" +#include "llvm/IR/LLVMContext.h" +#include "llvm/IR/Type.h" +#include "llvm/Support/Debug.h" +#include "llvm/Transforms/Utils/BasicBlockUtils.h" +#include "llvm/Transforms/Utils/Local.h" +#include "llvm/Transforms/Utils/LoopUtils.h" using namespace llvm; +#define DEBUG_TYPE "loop-simplify" + STATISTIC(NumInserted, "Number of pre-header or exit blocks inserted"); STATISTIC(NumNested , "Number of nested loops split out"); -namespace { - struct VISIBILITY_HIDDEN LoopSimplify : public FunctionPass { - static char ID; // Pass identification, replacement for typeid - LoopSimplify() : FunctionPass((intptr_t)&ID) {} - - // AA - If we have an alias analysis object to update, this is it, otherwise - // this is null. - AliasAnalysis *AA; - LoopInfo *LI; - DominatorTree *DT; - virtual bool runOnFunction(Function &F); - - virtual void getAnalysisUsage(AnalysisUsage &AU) const { - // We need loop information to identify the loops... - AU.addRequired(); - AU.addRequired(); - - AU.addPreserved(); - AU.addPreserved(); - AU.addPreserved(); - AU.addPreservedID(BreakCriticalEdgesID); // No critical edges added. - } - - /// verifyAnalysis() - Verify loop nest. - void verifyAnalysis() const { -#ifndef NDEBUG - LoopInfo *NLI = &getAnalysis(); - for (LoopInfo::iterator I = NLI->begin(), E = NLI->end(); I != E; ++I) - (*I)->verifyLoop(); -#endif - } - - private: - bool ProcessLoop(Loop *L); - BasicBlock *SplitBlockPredecessors(BasicBlock *BB, const char *Suffix, - const std::vector &Preds); - BasicBlock *RewriteLoopExitBlock(Loop *L, BasicBlock *Exit); - void InsertPreheaderForLoop(Loop *L); - Loop *SeparateNestedLoop(Loop *L); - void InsertUniqueBackedgeBlock(Loop *L); - void PlaceSplitBlockCarefully(BasicBlock *NewBB, - std::vector &SplitPreds, - Loop *L); - }; - - char LoopSimplify::ID = 0; - RegisterPass - X("loopsimplify", "Canonicalize natural loops", true); -} - -// Publically exposed interface to pass... -const PassInfo *llvm::LoopSimplifyID = X.getPassInfo(); -FunctionPass *llvm::createLoopSimplifyPass() { return new LoopSimplify(); } - -/// runOnFunction - Run down all loops in the CFG (recursively, but we could do -/// it in any convenient order) inserting preheaders... -/// -bool LoopSimplify::runOnFunction(Function &F) { - bool Changed = false; - LI = &getAnalysis(); - AA = getAnalysisToUpdate(); - DT = &getAnalysis(); - - // Check to see that no blocks (other than the header) in loops have - // predecessors that are not in loops. This is not valid for natural loops, - // but can occur if the blocks are unreachable. Since they are unreachable we - // can just shamelessly destroy their terminators to make them not branch into - // the loop! - for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB) { - // This case can only occur for unreachable blocks. Blocks that are - // unreachable can't be in loops, so filter those blocks out. - if (LI->getLoopFor(BB)) continue; - - bool BlockUnreachable = false; - TerminatorInst *TI = BB->getTerminator(); - - // Check to see if any successors of this block are non-loop-header loops - // that are not the header. - for (unsigned i = 0, e = TI->getNumSuccessors(); i != e; ++i) { - // If this successor is not in a loop, BB is clearly ok. - Loop *L = LI->getLoopFor(TI->getSuccessor(i)); - if (!L) continue; - - // If the succ is the loop header, and if L is a top-level loop, then this - // is an entrance into a loop through the header, which is also ok. - if (L->getHeader() == TI->getSuccessor(i) && L->getParentLoop() == 0) - continue; - - // Otherwise, this is an entrance into a loop from some place invalid. - // Either the loop structure is invalid and this is not a natural loop (in - // which case the compiler is buggy somewhere else) or BB is unreachable. - BlockUnreachable = true; - break; - } - - // If this block is ok, check the next one. - if (!BlockUnreachable) continue; - - // Otherwise, this block is dead. To clean up the CFG and to allow later - // loop transformations to ignore this case, we delete the edges into the - // loop by replacing the terminator. - - // Remove PHI entries from the successors. - for (unsigned i = 0, e = TI->getNumSuccessors(); i != e; ++i) - TI->getSuccessor(i)->removePredecessor(BB); - - // Add a new unreachable instruction before the old terminator. - new UnreachableInst(TI); - - // Delete the dead terminator. - if (AA) AA->deleteValue(TI); - if (!TI->use_empty()) - TI->replaceAllUsesWith(UndefValue::get(TI->getType())); - TI->eraseFromParent(); - Changed |= true; - } - - for (LoopInfo::iterator I = LI->begin(), E = LI->end(); I != E; ++I) - Changed |= ProcessLoop(*I); - - return Changed; -} - -/// ProcessLoop - Walk the loop structure in depth first order, ensuring that -/// all loops have preheaders. -/// -bool LoopSimplify::ProcessLoop(Loop *L) { - bool Changed = false; -ReprocessLoop: - - // Canonicalize inner loops before outer loops. Inner loop canonicalization - // can provide work for the outer loop to canonicalize. - for (Loop::iterator I = L->begin(), E = L->end(); I != E; ++I) - Changed |= ProcessLoop(*I); - - assert(L->getBlocks()[0] == L->getHeader() && - "Header isn't first block in loop?"); - - // Does the loop already have a preheader? If so, don't insert one. - if (L->getLoopPreheader() == 0) { - InsertPreheaderForLoop(L); - NumInserted++; - Changed = true; - } - - // Next, check to make sure that all exit nodes of the loop only have - // predecessors that are inside of the loop. This check guarantees that the - // loop preheader/header will dominate the exit blocks. If the exit block has - // predecessors from outside of the loop, split the edge now. - SmallVector ExitBlocks; - L->getExitBlocks(ExitBlocks); - - SetVector ExitBlockSet(ExitBlocks.begin(), ExitBlocks.end()); - for (SetVector::iterator I = ExitBlockSet.begin(), - E = ExitBlockSet.end(); I != E; ++I) { - BasicBlock *ExitBlock = *I; - for (pred_iterator PI = pred_begin(ExitBlock), PE = pred_end(ExitBlock); - PI != PE; ++PI) - // Must be exactly this loop: no subloops, parent loops, or non-loop preds - // allowed. - if (!L->contains(*PI)) { - RewriteLoopExitBlock(L, ExitBlock); - NumInserted++; - Changed = true; - break; - } - } - - // If the header has more than two predecessors at this point (from the - // preheader and from multiple backedges), we must adjust the loop. - unsigned NumBackedges = L->getNumBackEdges(); - if (NumBackedges != 1) { - // If this is really a nested loop, rip it out into a child loop. Don't do - // this for loops with a giant number of backedges, just factor them into a - // common backedge instead. - if (NumBackedges < 8) { - if (Loop *NL = SeparateNestedLoop(L)) { - ++NumNested; - // This is a big restructuring change, reprocess the whole loop. - ProcessLoop(NL); - Changed = true; - // GCC doesn't tail recursion eliminate this. - goto ReprocessLoop; - } - } - - // If we either couldn't, or didn't want to, identify nesting of the loops, - // insert a new block that all backedges target, then make it jump to the - // loop header. - InsertUniqueBackedgeBlock(L); - NumInserted++; - Changed = true; +// If the block isn't already, move the new block to right after some 'outside +// block' block. This prevents the preheader from being placed inside the loop +// body, e.g. when the loop hasn't been rotated. +static void placeSplitBlockCarefully(BasicBlock *NewBB, + SmallVectorImpl &SplitPreds, + Loop *L) { + // Check to see if NewBB is already well placed. + Function::iterator BBI = NewBB; --BBI; + for (unsigned i = 0, e = SplitPreds.size(); i != e; ++i) { + if (&*BBI == SplitPreds[i]) + return; } - // Scan over the PHI nodes in the loop header. Since they now have only two - // incoming values (the loop is canonicalized), we may have simplified the PHI - // down to 'X = phi [X, Y]', which should be replaced with 'Y'. - PHINode *PN; - for (BasicBlock::iterator I = L->getHeader()->begin(); - (PN = dyn_cast(I++)); ) - if (Value *V = PN->hasConstantValue()) { - PN->replaceAllUsesWith(V); - PN->eraseFromParent(); - } - - return Changed; -} - -/// SplitBlockPredecessors - Split the specified block into two blocks. We want -/// to move the predecessors specified in the Preds list to point to the new -/// block, leaving the remaining predecessors pointing to BB. This method -/// updates the SSA PHINode's, but no other analyses. -/// -BasicBlock *LoopSimplify::SplitBlockPredecessors(BasicBlock *BB, - const char *Suffix, - const std::vector &Preds) { - - // Create new basic block, insert right before the original block... - BasicBlock *NewBB = new BasicBlock(BB->getName()+Suffix, BB->getParent(), BB); - - // The preheader first gets an unconditional branch to the loop header... - BranchInst *BI = new BranchInst(BB, NewBB); - - // For every PHI node in the block, insert a PHI node into NewBB where the - // incoming values from the out of loop edges are moved to NewBB. We have two - // possible cases here. If the loop is dead, we just insert dummy entries - // into the PHI nodes for the new edge. If the loop is not dead, we move the - // incoming edges in BB into new PHI nodes in NewBB. - // - if (!Preds.empty()) { // Is the loop not obviously dead? - // Check to see if the values being merged into the new block need PHI - // nodes. If so, insert them. - for (BasicBlock::iterator I = BB->begin(); isa(I); ) { - PHINode *PN = cast(I); - ++I; - - // Check to see if all of the values coming in are the same. If so, we - // don't need to create a new PHI node. - Value *InVal = PN->getIncomingValueForBlock(Preds[0]); - for (unsigned i = 1, e = Preds.size(); i != e; ++i) - if (InVal != PN->getIncomingValueForBlock(Preds[i])) { - InVal = 0; - break; - } - - // If the values coming into the block are not the same, we need a PHI. - if (InVal == 0) { - // Create the new PHI node, insert it into NewBB at the end of the block - PHINode *NewPHI = new PHINode(PN->getType(), PN->getName()+".ph", BI); - if (AA) AA->copyValue(PN, NewPHI); - - // Move all of the edges from blocks outside the loop to the new PHI - for (unsigned i = 0, e = Preds.size(); i != e; ++i) { - Value *V = PN->removeIncomingValue(Preds[i], false); - NewPHI->addIncoming(V, Preds[i]); - } - InVal = NewPHI; - } else { - // Remove all of the edges coming into the PHI nodes from outside of the - // block. - for (unsigned i = 0, e = Preds.size(); i != e; ++i) - PN->removeIncomingValue(Preds[i], false); - } - - // Add an incoming value to the PHI node in the loop for the preheader - // edge. - PN->addIncoming(InVal, NewBB); - - // Can we eliminate this phi node now? - if (Value *V = PN->hasConstantValue(true)) { - Instruction *I = dyn_cast(V); - // If I is in NewBB, the Dominator call will fail, because NewBB isn't - // registered in DominatorTree yet. Handle this case explicitly. - if (!I || (I->getParent() != NewBB && - getAnalysis().dominates(I, PN))) { - PN->replaceAllUsesWith(V); - if (AA) AA->deleteValue(PN); - BB->getInstList().erase(PN); - } - } - } - - // Now that the PHI nodes are updated, actually move the edges from - // Preds to point to NewBB instead of BB. - // - for (unsigned i = 0, e = Preds.size(); i != e; ++i) { - TerminatorInst *TI = Preds[i]->getTerminator(); - for (unsigned s = 0, e = TI->getNumSuccessors(); s != e; ++s) - if (TI->getSuccessor(s) == BB) - TI->setSuccessor(s, NewBB); - } + // If it isn't already after an outside block, move it after one. This is + // always good as it makes the uncond branch from the outside block into a + // fall-through. - } else { // Otherwise the loop is dead... - for (BasicBlock::iterator I = BB->begin(); isa(I); ++I) { - PHINode *PN = cast(I); - // Insert dummy values as the incoming value... - PN->addIncoming(Constant::getNullValue(PN->getType()), NewBB); + // Figure out *which* outside block to put this after. Prefer an outside + // block that neighbors a BB actually in the loop. + BasicBlock *FoundBB = nullptr; + for (unsigned i = 0, e = SplitPreds.size(); i != e; ++i) { + Function::iterator BBI = SplitPreds[i]; + if (++BBI != NewBB->getParent()->end() && + L->contains(BBI)) { + FoundBB = SplitPreds[i]; + break; } } - return NewBB; + // If our heuristic for a *good* bb to place this after doesn't find + // anything, just pick something. It's likely better than leaving it within + // the loop. + if (!FoundBB) + FoundBB = SplitPreds[0]; + NewBB->moveAfter(FoundBB); } /// InsertPreheaderForLoop - Once we discover that a loop doesn't have a /// preheader, this method is called to insert one. This method has two phases: /// preheader insertion and analysis updating. /// -void LoopSimplify::InsertPreheaderForLoop(Loop *L) { +BasicBlock *llvm::InsertPreheaderForLoop(Loop *L, Pass *PP) { BasicBlock *Header = L->getHeader(); // Compute the set of predecessors of the loop that are not in the loop. - std::vector OutsideBlocks; + SmallVector OutsideBlocks; for (pred_iterator PI = pred_begin(Header), PE = pred_end(Header); - PI != PE; ++PI) - if (!L->contains(*PI)) // Coming in from outside the loop? - OutsideBlocks.push_back(*PI); // Keep track of it... + PI != PE; ++PI) { + BasicBlock *P = *PI; + if (!L->contains(P)) { // Coming in from outside the loop? + // If the loop is branched to from an indirect branch, we won't + // be able to fully transform the loop, because it prohibits + // edge splitting. + if (isa(P->getTerminator())) return nullptr; + + // Keep track of it. + OutsideBlocks.push_back(P); + } + } // Split out the loop pre-header. - BasicBlock *NewBB = - SplitBlockPredecessors(Header, ".preheader", OutsideBlocks); - - - //===--------------------------------------------------------------------===// - // Update analysis results now that we have performed the transformation - // - - // We know that we have loop information to update... update it now. - if (Loop *Parent = L->getParentLoop()) - Parent->addBasicBlockToLoop(NewBB, LI->getBase()); + BasicBlock *PreheaderBB; + if (!Header->isLandingPad()) { + PreheaderBB = SplitBlockPredecessors(Header, OutsideBlocks, ".preheader", + PP); + } else { + SmallVector NewBBs; + SplitLandingPadPredecessors(Header, OutsideBlocks, ".preheader", + ".split-lp", PP, NewBBs); + PreheaderBB = NewBBs[0]; + } - DT->splitBlock(NewBB); - if (DominanceFrontier *DF = getAnalysisToUpdate()) - DF->splitBlock(NewBB); + PreheaderBB->getTerminator()->setDebugLoc( + Header->getFirstNonPHI()->getDebugLoc()); + DEBUG(dbgs() << "LoopSimplify: Creating pre-header " + << PreheaderBB->getName() << "\n"); // Make sure that NewBB is put someplace intelligent, which doesn't mess up // code layout too horribly. - PlaceSplitBlockCarefully(NewBB, OutsideBlocks, L); + placeSplitBlockCarefully(PreheaderBB, OutsideBlocks, L); + + return PreheaderBB; } -/// RewriteLoopExitBlock - Ensure that the loop preheader dominates all exit -/// blocks. This method is used to split exit blocks that have predecessors -/// outside of the loop. -BasicBlock *LoopSimplify::RewriteLoopExitBlock(Loop *L, BasicBlock *Exit) { - std::vector LoopBlocks; - for (pred_iterator I = pred_begin(Exit), E = pred_end(Exit); I != E; ++I) - if (L->contains(*I)) - LoopBlocks.push_back(*I); +/// \brief Ensure that the loop preheader dominates all exit blocks. +/// +/// This method is used to split exit blocks that have predecessors outside of +/// the loop. +static BasicBlock *rewriteLoopExitBlock(Loop *L, BasicBlock *Exit, Pass *PP) { + SmallVector LoopBlocks; + for (pred_iterator I = pred_begin(Exit), E = pred_end(Exit); I != E; ++I) { + BasicBlock *P = *I; + if (L->contains(P)) { + // Don't do this if the loop is exited via an indirect branch. + if (isa(P->getTerminator())) return nullptr; + + LoopBlocks.push_back(P); + } + } assert(!LoopBlocks.empty() && "No edges coming in from outside the loop?"); - BasicBlock *NewBB = SplitBlockPredecessors(Exit, ".loopexit", LoopBlocks); - - // Update Loop Information - we know that the new block will be in whichever - // loop the Exit block is in. Note that it may not be in that immediate loop, - // if the successor is some other loop header. In that case, we continue - // walking up the loop tree to find a loop that contains both the successor - // block and the predecessor block. - Loop *SuccLoop = LI->getLoopFor(Exit); - while (SuccLoop && !SuccLoop->contains(L->getHeader())) - SuccLoop = SuccLoop->getParentLoop(); - if (SuccLoop) - SuccLoop->addBasicBlockToLoop(NewBB, LI->getBase()); - - // Update Dominator Information - DT->splitBlock(NewBB); - if (DominanceFrontier *DF = getAnalysisToUpdate()) - DF->splitBlock(NewBB); - - return NewBB; + BasicBlock *NewExitBB = nullptr; + + if (Exit->isLandingPad()) { + SmallVector NewBBs; + SplitLandingPadPredecessors(Exit, ArrayRef(&LoopBlocks[0], + LoopBlocks.size()), + ".loopexit", ".nonloopexit", + PP, NewBBs); + NewExitBB = NewBBs[0]; + } else { + NewExitBB = SplitBlockPredecessors(Exit, LoopBlocks, ".loopexit", PP); + } + + DEBUG(dbgs() << "LoopSimplify: Creating dedicated exit block " + << NewExitBB->getName() << "\n"); + return NewExitBB; } -/// AddBlockAndPredsToSet - Add the specified block, and all of its -/// predecessors, to the specified set, if it's not already in there. Stop -/// predecessor traversal when we reach StopBlock. -static void AddBlockAndPredsToSet(BasicBlock *InputBB, BasicBlock *StopBlock, +/// Add the specified block, and all of its predecessors, to the specified set, +/// if it's not already in there. Stop predecessor traversal when we reach +/// StopBlock. +static void addBlockAndPredsToSet(BasicBlock *InputBB, BasicBlock *StopBlock, std::set &Blocks) { - std::vector WorkList; - WorkList.push_back(InputBB); + SmallVector Worklist; + Worklist.push_back(InputBB); do { - BasicBlock *BB = WorkList.back(); WorkList.pop_back(); + BasicBlock *BB = Worklist.pop_back_val(); if (Blocks.insert(BB).second && BB != StopBlock) // If BB is not already processed and it is not a stop block then // insert its predecessor in the work list for (pred_iterator I = pred_begin(BB), E = pred_end(BB); I != E; ++I) { BasicBlock *WBB = *I; - WorkList.push_back(WBB); + Worklist.push_back(WBB); } - } while(!WorkList.empty()); + } while (!Worklist.empty()); } -/// FindPHIToPartitionLoops - The first part of loop-nestification is to find a -/// PHI node that tells us how to partition the loops. -static PHINode *FindPHIToPartitionLoops(Loop *L, DominatorTree *DT, - AliasAnalysis *AA) { +/// \brief The first part of loop-nestification is to find a PHI node that tells +/// us how to partition the loops. +static PHINode *findPHIToPartitionLoops(Loop *L, AliasAnalysis *AA, + DominatorTree *DT) { for (BasicBlock::iterator I = L->getHeader()->begin(); isa(I); ) { PHINode *PN = cast(I); ++I; - if (Value *V = PN->hasConstantValue()) - if (!isa(V) || DT->dominates(cast(V), PN)) { - // This is a degenerate PHI already, don't modify it! - PN->replaceAllUsesWith(V); - if (AA) AA->deleteValue(PN); - PN->eraseFromParent(); - continue; - } + if (Value *V = SimplifyInstruction(PN, nullptr, nullptr, DT)) { + // This is a degenerate PHI already, don't modify it! + PN->replaceAllUsesWith(V); + if (AA) AA->deleteValue(PN); + PN->eraseFromParent(); + continue; + } // Scan this PHI node looking for a use of the PHI node by itself. for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) @@ -464,49 +227,13 @@ static PHINode *FindPHIToPartitionLoops(Loop *L, DominatorTree *DT, // We found something tasty to remove. return PN; } - return 0; -} - -// PlaceSplitBlockCarefully - If the block isn't already, move the new block to -// right after some 'outside block' block. This prevents the preheader from -// being placed inside the loop body, e.g. when the loop hasn't been rotated. -void LoopSimplify::PlaceSplitBlockCarefully(BasicBlock *NewBB, - std::vector&SplitPreds, - Loop *L) { - // Check to see if NewBB is already well placed. - Function::iterator BBI = NewBB; --BBI; - for (unsigned i = 0, e = SplitPreds.size(); i != e; ++i) { - if (&*BBI == SplitPreds[i]) - return; - } - - // If it isn't already after an outside block, move it after one. This is - // always good as it makes the uncond branch from the outside block into a - // fall-through. - - // Figure out *which* outside block to put this after. Prefer an outside - // block that neighbors a BB actually in the loop. - BasicBlock *FoundBB = 0; - for (unsigned i = 0, e = SplitPreds.size(); i != e; ++i) { - Function::iterator BBI = SplitPreds[i]; - if (++BBI != NewBB->getParent()->end() && - L->contains(BBI)) { - FoundBB = SplitPreds[i]; - break; - } - } - - // If our heuristic for a *good* bb to place this after doesn't find - // anything, just pick something. It's likely better than leaving it within - // the loop. - if (!FoundBB) - FoundBB = SplitPreds[0]; - NewBB->moveAfter(FoundBB); + return nullptr; } - -/// SeparateNestedLoop - If this loop has multiple backedges, try to pull one of -/// them out into a nested loop. This is important for code that looks like +/// \brief If this loop has multiple backedges, try to pull one of them out into +/// a nested loop. +/// +/// This is important for code that looks like /// this: /// /// Loop: @@ -522,31 +249,49 @@ void LoopSimplify::PlaceSplitBlockCarefully(BasicBlock *NewBB, /// If we are able to separate out a loop, return the new outer loop that was /// created. /// -Loop *LoopSimplify::SeparateNestedLoop(Loop *L) { - PHINode *PN = FindPHIToPartitionLoops(L, DT, AA); - if (PN == 0) return 0; // No known way to partition. +static Loop *separateNestedLoop(Loop *L, BasicBlock *Preheader, + AliasAnalysis *AA, DominatorTree *DT, + LoopInfo *LI, ScalarEvolution *SE, Pass *PP) { + // Don't try to separate loops without a preheader. + if (!Preheader) + return nullptr; + + // The header is not a landing pad; preheader insertion should ensure this. + assert(!L->getHeader()->isLandingPad() && + "Can't insert backedge to landing pad"); + + PHINode *PN = findPHIToPartitionLoops(L, AA, DT); + if (!PN) return nullptr; // No known way to partition. // Pull out all predecessors that have varying values in the loop. This // handles the case when a PHI node has multiple instances of itself as // arguments. - std::vector OuterLoopPreds; - for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) + SmallVector OuterLoopPreds; + for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) { if (PN->getIncomingValue(i) != PN || - !L->contains(PN->getIncomingBlock(i))) + !L->contains(PN->getIncomingBlock(i))) { + // We can't split indirectbr edges. + if (isa(PN->getIncomingBlock(i)->getTerminator())) + return nullptr; OuterLoopPreds.push_back(PN->getIncomingBlock(i)); + } + } + DEBUG(dbgs() << "LoopSimplify: Splitting out a new outer loop\n"); - BasicBlock *Header = L->getHeader(); - BasicBlock *NewBB = SplitBlockPredecessors(Header, ".outer", OuterLoopPreds); + // If ScalarEvolution is around and knows anything about values in + // this loop, tell it to forget them, because we're about to + // substantially change it. + if (SE) + SE->forgetLoop(L); - // Update dominator information - DT->splitBlock(NewBB); - if (DominanceFrontier *DF = getAnalysisToUpdate()) - DF->splitBlock(NewBB); + BasicBlock *Header = L->getHeader(); + BasicBlock *NewBB = + SplitBlockPredecessors(Header, OuterLoopPreds, ".outer", PP); // Make sure that NewBB is put someplace intelligent, which doesn't mess up // code layout too horribly. - PlaceSplitBlockCarefully(NewBB, OuterLoopPreds, L); - + placeSplitBlockCarefully(NewBB, OuterLoopPreds, L); + // Create the new outer loop. Loop *NewOuter = new Loop(); @@ -556,23 +301,25 @@ Loop *LoopSimplify::SeparateNestedLoop(Loop *L) { else LI->changeTopLevelLoop(L, NewOuter); - // This block is going to be our new header block: add it to this loop and all - // parent loops. - NewOuter->addBasicBlockToLoop(NewBB, LI->getBase()); - // L is now a subloop of our outer loop. NewOuter->addChildLoop(L); - for (unsigned i = 0, e = L->getBlocks().size(); i != e; ++i) - NewOuter->addBlockEntry(L->getBlocks()[i]); + for (Loop::block_iterator I = L->block_begin(), E = L->block_end(); + I != E; ++I) + NewOuter->addBlockEntry(*I); + + // Now reset the header in L, which had been moved by + // SplitBlockPredecessors for the outer loop. + L->moveToHeader(Header); // Determine which blocks should stay in L and which should be moved out to // the Outer loop now. std::set BlocksInL; - for (pred_iterator PI = pred_begin(Header), E = pred_end(Header); PI!=E; ++PI) - if (DT->dominates(Header, *PI)) - AddBlockAndPredsToSet(*PI, Header, BlocksInL); - + for (pred_iterator PI=pred_begin(Header), E = pred_end(Header); PI!=E; ++PI) { + BasicBlock *P = *PI; + if (DT->dominates(Header, P)) + addBlockAndPredsToSet(P, Header, BlocksInL); + } // Scan all of the loop children of L, moving them to OuterLoop if they are // not part of the inner loop. @@ -599,29 +346,47 @@ Loop *LoopSimplify::SeparateNestedLoop(Loop *L) { return NewOuter; } - - -/// InsertUniqueBackedgeBlock - This method is called when the specified loop -/// has more than one backedge in it. If this occurs, revector all of these -/// backedges to target a new basic block and have that block branch to the loop -/// header. This ensures that loops have exactly one backedge. +/// \brief This method is called when the specified loop has more than one +/// backedge in it. /// -void LoopSimplify::InsertUniqueBackedgeBlock(Loop *L) { +/// If this occurs, revector all of these backedges to target a new basic block +/// and have that block branch to the loop header. This ensures that loops +/// have exactly one backedge. +static BasicBlock *insertUniqueBackedgeBlock(Loop *L, BasicBlock *Preheader, + AliasAnalysis *AA, + DominatorTree *DT, LoopInfo *LI) { assert(L->getNumBackEdges() > 1 && "Must have > 1 backedge!"); // Get information about the loop - BasicBlock *Preheader = L->getLoopPreheader(); BasicBlock *Header = L->getHeader(); Function *F = Header->getParent(); + // Unique backedge insertion currently depends on having a preheader. + if (!Preheader) + return nullptr; + + // The header is not a landing pad; preheader insertion should ensure this. + assert(!Header->isLandingPad() && "Can't insert backedge to landing pad"); + // Figure out which basic blocks contain back-edges to the loop header. std::vector BackedgeBlocks; - for (pred_iterator I = pred_begin(Header), E = pred_end(Header); I != E; ++I) - if (*I != Preheader) BackedgeBlocks.push_back(*I); + for (pred_iterator I = pred_begin(Header), E = pred_end(Header); I != E; ++I){ + BasicBlock *P = *I; + + // Indirectbr edges cannot be split, so we must fail if we find one. + if (isa(P->getTerminator())) + return nullptr; + + if (P != Preheader) BackedgeBlocks.push_back(P); + } // Create and insert the new backedge block... - BasicBlock *BEBlock = new BasicBlock(Header->getName()+".backedge", F); - BranchInst *BETerminator = new BranchInst(Header, BEBlock); + BasicBlock *BEBlock = BasicBlock::Create(Header->getContext(), + Header->getName()+".backedge", F); + BranchInst *BETerminator = BranchInst::Create(Header, BEBlock); + + DEBUG(dbgs() << "LoopSimplify: Inserting unique backedge block " + << BEBlock->getName() << "\n"); // Move the new backedge block to right after the last backedge block. Function::iterator InsertPos = BackedgeBlocks.back(); ++InsertPos; @@ -631,16 +396,15 @@ void LoopSimplify::InsertUniqueBackedgeBlock(Loop *L) { // the backedge block which correspond to any PHI nodes in the header block. for (BasicBlock::iterator I = Header->begin(); isa(I); ++I) { PHINode *PN = cast(I); - PHINode *NewPN = new PHINode(PN->getType(), PN->getName()+".be", - BETerminator); - NewPN->reserveOperandSpace(BackedgeBlocks.size()); + PHINode *NewPN = PHINode::Create(PN->getType(), BackedgeBlocks.size(), + PN->getName()+".be", BETerminator); if (AA) AA->copyValue(PN, NewPN); // Loop over the PHI node, moving all entries except the one for the // preheader over to the new PHI node. unsigned PreheaderIdx = ~0U; bool HasUniqueIncomingValue = true; - Value *UniqueValue = 0; + Value *UniqueValue = nullptr; for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) { BasicBlock *IBB = PN->getIncomingBlock(i); Value *IV = PN->getIncomingValue(i); @@ -649,7 +413,7 @@ void LoopSimplify::InsertUniqueBackedgeBlock(Loop *L) { } else { NewPN->addIncoming(IV, IBB); if (HasUniqueIncomingValue) { - if (UniqueValue == 0) + if (!UniqueValue) UniqueValue = IV; else if (UniqueValue != IV) HasUniqueIncomingValue = false; @@ -697,6 +461,393 @@ void LoopSimplify::InsertUniqueBackedgeBlock(Loop *L) { // Update dominator information DT->splitBlock(BEBlock); - if (DominanceFrontier *DF = getAnalysisToUpdate()) - DF->splitBlock(BEBlock); + + return BEBlock; +} + +/// \brief Simplify one loop and queue further loops for simplification. +/// +/// FIXME: Currently this accepts both lots of analyses that it uses and a raw +/// Pass pointer. The Pass pointer is used by numerous utilities to update +/// specific analyses. Rather than a pass it would be much cleaner and more +/// explicit if they accepted the analysis directly and then updated it. +static bool simplifyOneLoop(Loop *L, SmallVectorImpl &Worklist, + AliasAnalysis *AA, DominatorTree *DT, LoopInfo *LI, + ScalarEvolution *SE, Pass *PP) { + bool Changed = false; +ReprocessLoop: + + // Check to see that no blocks (other than the header) in this loop have + // predecessors that are not in the loop. This is not valid for natural + // loops, but can occur if the blocks are unreachable. Since they are + // unreachable we can just shamelessly delete those CFG edges! + for (Loop::block_iterator BB = L->block_begin(), E = L->block_end(); + BB != E; ++BB) { + if (*BB == L->getHeader()) continue; + + SmallPtrSet BadPreds; + for (pred_iterator PI = pred_begin(*BB), + PE = pred_end(*BB); PI != PE; ++PI) { + BasicBlock *P = *PI; + if (!L->contains(P)) + BadPreds.insert(P); + } + + // Delete each unique out-of-loop (and thus dead) predecessor. + for (SmallPtrSet::iterator I = BadPreds.begin(), + E = BadPreds.end(); I != E; ++I) { + + DEBUG(dbgs() << "LoopSimplify: Deleting edge from dead predecessor " + << (*I)->getName() << "\n"); + + // Inform each successor of each dead pred. + for (succ_iterator SI = succ_begin(*I), SE = succ_end(*I); SI != SE; ++SI) + (*SI)->removePredecessor(*I); + // Zap the dead pred's terminator and replace it with unreachable. + TerminatorInst *TI = (*I)->getTerminator(); + TI->replaceAllUsesWith(UndefValue::get(TI->getType())); + (*I)->getTerminator()->eraseFromParent(); + new UnreachableInst((*I)->getContext(), *I); + Changed = true; + } + } + + // If there are exiting blocks with branches on undef, resolve the undef in + // the direction which will exit the loop. This will help simplify loop + // trip count computations. + SmallVector ExitingBlocks; + L->getExitingBlocks(ExitingBlocks); + for (SmallVectorImpl::iterator I = ExitingBlocks.begin(), + E = ExitingBlocks.end(); I != E; ++I) + if (BranchInst *BI = dyn_cast((*I)->getTerminator())) + if (BI->isConditional()) { + if (UndefValue *Cond = dyn_cast(BI->getCondition())) { + + DEBUG(dbgs() << "LoopSimplify: Resolving \"br i1 undef\" to exit in " + << (*I)->getName() << "\n"); + + BI->setCondition(ConstantInt::get(Cond->getType(), + !L->contains(BI->getSuccessor(0)))); + + // This may make the loop analyzable, force SCEV recomputation. + if (SE) + SE->forgetLoop(L); + + Changed = true; + } + } + + // Does the loop already have a preheader? If so, don't insert one. + BasicBlock *Preheader = L->getLoopPreheader(); + if (!Preheader) { + Preheader = InsertPreheaderForLoop(L, PP); + if (Preheader) { + ++NumInserted; + Changed = true; + } + } + + // Next, check to make sure that all exit nodes of the loop only have + // predecessors that are inside of the loop. This check guarantees that the + // loop preheader/header will dominate the exit blocks. If the exit block has + // predecessors from outside of the loop, split the edge now. + SmallVector ExitBlocks; + L->getExitBlocks(ExitBlocks); + + SmallSetVector ExitBlockSet(ExitBlocks.begin(), + ExitBlocks.end()); + for (SmallSetVector::iterator I = ExitBlockSet.begin(), + E = ExitBlockSet.end(); I != E; ++I) { + BasicBlock *ExitBlock = *I; + for (pred_iterator PI = pred_begin(ExitBlock), PE = pred_end(ExitBlock); + PI != PE; ++PI) + // Must be exactly this loop: no subloops, parent loops, or non-loop preds + // allowed. + if (!L->contains(*PI)) { + if (rewriteLoopExitBlock(L, ExitBlock, PP)) { + ++NumInserted; + Changed = true; + } + break; + } + } + + // If the header has more than two predecessors at this point (from the + // preheader and from multiple backedges), we must adjust the loop. + BasicBlock *LoopLatch = L->getLoopLatch(); + if (!LoopLatch) { + // If this is really a nested loop, rip it out into a child loop. Don't do + // this for loops with a giant number of backedges, just factor them into a + // common backedge instead. + if (L->getNumBackEdges() < 8) { + if (Loop *OuterL = separateNestedLoop(L, Preheader, AA, DT, LI, SE, PP)) { + ++NumNested; + // Enqueue the outer loop as it should be processed next in our + // depth-first nest walk. + Worklist.push_back(OuterL); + + // This is a big restructuring change, reprocess the whole loop. + Changed = true; + // GCC doesn't tail recursion eliminate this. + // FIXME: It isn't clear we can't rely on LLVM to TRE this. + goto ReprocessLoop; + } + } + + // If we either couldn't, or didn't want to, identify nesting of the loops, + // insert a new block that all backedges target, then make it jump to the + // loop header. + LoopLatch = insertUniqueBackedgeBlock(L, Preheader, AA, DT, LI); + if (LoopLatch) { + ++NumInserted; + Changed = true; + } + } + + // Scan over the PHI nodes in the loop header. Since they now have only two + // incoming values (the loop is canonicalized), we may have simplified the PHI + // down to 'X = phi [X, Y]', which should be replaced with 'Y'. + PHINode *PN; + for (BasicBlock::iterator I = L->getHeader()->begin(); + (PN = dyn_cast(I++)); ) + if (Value *V = SimplifyInstruction(PN, nullptr, nullptr, DT)) { + if (AA) AA->deleteValue(PN); + if (SE) SE->forgetValue(PN); + PN->replaceAllUsesWith(V); + PN->eraseFromParent(); + } + + // If this loop has multiple exits and the exits all go to the same + // block, attempt to merge the exits. This helps several passes, such + // as LoopRotation, which do not support loops with multiple exits. + // SimplifyCFG also does this (and this code uses the same utility + // function), however this code is loop-aware, where SimplifyCFG is + // not. That gives it the advantage of being able to hoist + // loop-invariant instructions out of the way to open up more + // opportunities, and the disadvantage of having the responsibility + // to preserve dominator information. + bool UniqueExit = true; + if (!ExitBlocks.empty()) + for (unsigned i = 1, e = ExitBlocks.size(); i != e; ++i) + if (ExitBlocks[i] != ExitBlocks[0]) { + UniqueExit = false; + break; + } + if (UniqueExit) { + for (unsigned i = 0, e = ExitingBlocks.size(); i != e; ++i) { + BasicBlock *ExitingBlock = ExitingBlocks[i]; + if (!ExitingBlock->getSinglePredecessor()) continue; + BranchInst *BI = dyn_cast(ExitingBlock->getTerminator()); + if (!BI || !BI->isConditional()) continue; + CmpInst *CI = dyn_cast(BI->getCondition()); + if (!CI || CI->getParent() != ExitingBlock) continue; + + // Attempt to hoist out all instructions except for the + // comparison and the branch. + bool AllInvariant = true; + bool AnyInvariant = false; + for (BasicBlock::iterator I = ExitingBlock->begin(); &*I != BI; ) { + Instruction *Inst = I++; + // Skip debug info intrinsics. + if (isa(Inst)) + continue; + if (Inst == CI) + continue; + if (!L->makeLoopInvariant(Inst, AnyInvariant, + Preheader ? Preheader->getTerminator() + : nullptr)) { + AllInvariant = false; + break; + } + } + if (AnyInvariant) { + Changed = true; + // The loop disposition of all SCEV expressions that depend on any + // hoisted values have also changed. + if (SE) + SE->forgetLoopDispositions(L); + } + if (!AllInvariant) continue; + + // The block has now been cleared of all instructions except for + // a comparison and a conditional branch. SimplifyCFG may be able + // to fold it now. + if (!FoldBranchToCommonDest(BI)) continue; + + // Success. The block is now dead, so remove it from the loop, + // update the dominator tree and delete it. + DEBUG(dbgs() << "LoopSimplify: Eliminating exiting block " + << ExitingBlock->getName() << "\n"); + + // Notify ScalarEvolution before deleting this block. Currently assume the + // parent loop doesn't change (spliting edges doesn't count). If blocks, + // CFG edges, or other values in the parent loop change, then we need call + // to forgetLoop() for the parent instead. + if (SE) + SE->forgetLoop(L); + + assert(pred_begin(ExitingBlock) == pred_end(ExitingBlock)); + Changed = true; + LI->removeBlock(ExitingBlock); + + DomTreeNode *Node = DT->getNode(ExitingBlock); + const std::vector *> &Children = + Node->getChildren(); + while (!Children.empty()) { + DomTreeNode *Child = Children.front(); + DT->changeImmediateDominator(Child, Node->getIDom()); + } + DT->eraseNode(ExitingBlock); + + BI->getSuccessor(0)->removePredecessor(ExitingBlock); + BI->getSuccessor(1)->removePredecessor(ExitingBlock); + ExitingBlock->eraseFromParent(); + } + } + + return Changed; +} + +bool llvm::simplifyLoop(Loop *L, DominatorTree *DT, LoopInfo *LI, Pass *PP, + AliasAnalysis *AA, ScalarEvolution *SE) { + bool Changed = false; + + // Worklist maintains our depth-first queue of loops in this nest to process. + SmallVector Worklist; + Worklist.push_back(L); + + // Walk the worklist from front to back, pushing newly found sub loops onto + // the back. This will let us process loops from back to front in depth-first + // order. We can use this simple process because loops form a tree. + for (unsigned Idx = 0; Idx != Worklist.size(); ++Idx) { + Loop *L2 = Worklist[Idx]; + for (Loop::iterator I = L2->begin(), E = L2->end(); I != E; ++I) + Worklist.push_back(*I); + } + + while (!Worklist.empty()) + Changed |= simplifyOneLoop(Worklist.pop_back_val(), Worklist, AA, DT, LI, SE, PP); + + return Changed; +} + +namespace { + struct LoopSimplify : public FunctionPass { + static char ID; // Pass identification, replacement for typeid + LoopSimplify() : FunctionPass(ID) { + initializeLoopSimplifyPass(*PassRegistry::getPassRegistry()); + } + + // AA - If we have an alias analysis object to update, this is it, otherwise + // this is null. + AliasAnalysis *AA; + DominatorTree *DT; + LoopInfo *LI; + ScalarEvolution *SE; + + bool runOnFunction(Function &F) override; + + void getAnalysisUsage(AnalysisUsage &AU) const override { + // We need loop information to identify the loops... + AU.addRequired(); + AU.addPreserved(); + + AU.addRequired(); + AU.addPreserved(); + + AU.addPreserved(); + AU.addPreserved(); + AU.addPreserved(); + AU.addPreservedID(BreakCriticalEdgesID); // No critical edges added. + } + + /// verifyAnalysis() - Verify LoopSimplifyForm's guarantees. + void verifyAnalysis() const override; + }; +} + +char LoopSimplify::ID = 0; +INITIALIZE_PASS_BEGIN(LoopSimplify, "loop-simplify", + "Canonicalize natural loops", true, false) +INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass) +INITIALIZE_PASS_DEPENDENCY(LoopInfo) +INITIALIZE_PASS_END(LoopSimplify, "loop-simplify", + "Canonicalize natural loops", true, false) + +// Publicly exposed interface to pass... +char &llvm::LoopSimplifyID = LoopSimplify::ID; +Pass *llvm::createLoopSimplifyPass() { return new LoopSimplify(); } + +/// runOnFunction - Run down all loops in the CFG (recursively, but we could do +/// it in any convenient order) inserting preheaders... +/// +bool LoopSimplify::runOnFunction(Function &F) { + bool Changed = false; + AA = getAnalysisIfAvailable(); + LI = &getAnalysis(); + DT = &getAnalysis().getDomTree(); + SE = getAnalysisIfAvailable(); + + // Simplify each loop nest in the function. + for (LoopInfo::iterator I = LI->begin(), E = LI->end(); I != E; ++I) + Changed |= simplifyLoop(*I, DT, LI, this, AA, SE); + + return Changed; +} + +// FIXME: Restore this code when we re-enable verification in verifyAnalysis +// below. +#if 0 +static void verifyLoop(Loop *L) { + // Verify subloops. + for (Loop::iterator I = L->begin(), E = L->end(); I != E; ++I) + verifyLoop(*I); + + // It used to be possible to just assert L->isLoopSimplifyForm(), however + // with the introduction of indirectbr, there are now cases where it's + // not possible to transform a loop as necessary. We can at least check + // that there is an indirectbr near any time there's trouble. + + // Indirectbr can interfere with preheader and unique backedge insertion. + if (!L->getLoopPreheader() || !L->getLoopLatch()) { + bool HasIndBrPred = false; + for (pred_iterator PI = pred_begin(L->getHeader()), + PE = pred_end(L->getHeader()); PI != PE; ++PI) + if (isa((*PI)->getTerminator())) { + HasIndBrPred = true; + break; + } + assert(HasIndBrPred && + "LoopSimplify has no excuse for missing loop header info!"); + (void)HasIndBrPred; + } + + // Indirectbr can interfere with exit block canonicalization. + if (!L->hasDedicatedExits()) { + bool HasIndBrExiting = false; + SmallVector ExitingBlocks; + L->getExitingBlocks(ExitingBlocks); + for (unsigned i = 0, e = ExitingBlocks.size(); i != e; ++i) { + if (isa((ExitingBlocks[i])->getTerminator())) { + HasIndBrExiting = true; + break; + } + } + + assert(HasIndBrExiting && + "LoopSimplify has no excuse for missing exit block info!"); + (void)HasIndBrExiting; + } +} +#endif + +void LoopSimplify::verifyAnalysis() const { + // FIXME: This routine is being called mid-way through the loop pass manager + // as loop passes destroy this analysis. That's actually fine, but we have no + // way of expressing that here. Once all of the passes that destroy this are + // hoisted out of the loop pass manager we can add back verification here. +#if 0 + for (LoopInfo::iterator I = LI->begin(), E = LI->end(); I != E; ++I) + verifyLoop(*I); +#endif }