X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FTransforms%2FScalar%2FLoopRotation.cpp;h=e98ae953e53240fedce114acec83d6cecf07c200;hb=467116a1c89528963a18306c9b03a0e1e611f2c4;hp=f38944178267d2e6889b939de4d696e90b522fea;hpb=66fe0ded050a826c3c13bcece64f607f667e4845;p=oota-llvm.git diff --git a/lib/Transforms/Scalar/LoopRotation.cpp b/lib/Transforms/Scalar/LoopRotation.cpp index f3894417826..e98ae953e53 100644 --- a/lib/Transforms/Scalar/LoopRotation.cpp +++ b/lib/Transforms/Scalar/LoopRotation.cpp @@ -13,18 +13,21 @@ #define DEBUG_TYPE "loop-rotate" #include "llvm/Transforms/Scalar.h" -#include "llvm/Function.h" +#include "llvm/ADT/Statistic.h" #include "llvm/Analysis/CodeMetrics.h" -#include "llvm/Analysis/DominanceFrontier.h" -#include "llvm/Analysis/LoopPass.h" #include "llvm/Analysis/InstructionSimplify.h" +#include "llvm/Analysis/LoopPass.h" #include "llvm/Analysis/ScalarEvolution.h" -#include "llvm/Transforms/Utils/Local.h" +#include "llvm/Analysis/TargetTransformInfo.h" +#include "llvm/Analysis/ValueTracking.h" +#include "llvm/IR/Function.h" +#include "llvm/IR/IntrinsicInst.h" +#include "llvm/Support/CFG.h" +#include "llvm/Support/Debug.h" #include "llvm/Transforms/Utils/BasicBlockUtils.h" +#include "llvm/Transforms/Utils/Local.h" #include "llvm/Transforms/Utils/SSAUpdater.h" #include "llvm/Transforms/Utils/ValueMapper.h" -#include "llvm/Support/Debug.h" -#include "llvm/ADT/Statistic.h" using namespace llvm; #define MAX_HEADER_SIZE 16 @@ -39,14 +42,9 @@ namespace { initializeLoopRotatePass(*PassRegistry::getPassRegistry()); } - // Rotate Loop L as many times as possible. Return true if - // loop is rotated at least once. - bool runOnLoop(Loop *L, LPPassManager &LPM); - // LCSSA form makes instruction renaming easier. virtual void getAnalysisUsage(AnalysisUsage &AU) const { AU.addPreserved(); - AU.addPreserved(); AU.addRequired(); AU.addPreserved(); AU.addRequiredID(LoopSimplifyID); @@ -54,28 +52,22 @@ namespace { AU.addRequiredID(LCSSAID); AU.addPreservedID(LCSSAID); AU.addPreserved(); + AU.addRequired(); } - // Helper functions - - /// Do actual work + bool runOnLoop(Loop *L, LPPassManager &LPM); + void simplifyLoopLatch(Loop *L); bool rotateLoop(Loop *L); - - /// After loop rotation, loop pre-header has multiple sucessors. - /// Insert one forwarding basic block to ensure that loop pre-header - /// has only one successor. - void preserveCanonicalLoopForm(Loop *L, BasicBlock *OrigHeader, - BasicBlock *OrigPreHeader, - BasicBlock *OrigLatch, BasicBlock *NewHeader, - BasicBlock *Exit); private: LoopInfo *LI; + const TargetTransformInfo *TTI; }; } - + char LoopRotate::ID = 0; INITIALIZE_PASS_BEGIN(LoopRotate, "loop-rotate", "Rotate Loops", false, false) +INITIALIZE_AG_DEPENDENCY(TargetTransformInfo) INITIALIZE_PASS_DEPENDENCY(LoopInfo) INITIALIZE_PASS_DEPENDENCY(LoopSimplify) INITIALIZE_PASS_DEPENDENCY(LCSSA) @@ -87,6 +79,12 @@ Pass *llvm::createLoopRotatePass() { return new LoopRotate(); } /// the loop is rotated at least once. bool LoopRotate::runOnLoop(Loop *L, LPPassManager &LPM) { LI = &getAnalysis(); + TTI = &getAnalysis(); + + // Simplify the loop latch before attempting to rotate the header + // upward. Rotation may not be needed if the loop tail can be folded into the + // loop exit. + simplifyLoopLatch(L); // One loop can be rotated multiple times. bool MadeChange = false; @@ -96,55 +94,220 @@ bool LoopRotate::runOnLoop(Loop *L, LPPassManager &LPM) { return MadeChange; } -/// Rotate loop LP. Return true if the loop is rotated. -bool LoopRotate::rotateLoop(Loop *L) { - BasicBlock *OrigPreHeader = L->getLoopPreheader(); - if (!OrigPreHeader) return false; +/// RewriteUsesOfClonedInstructions - We just cloned the instructions from the +/// old header into the preheader. If there were uses of the values produced by +/// these instruction that were outside of the loop, we have to insert PHI nodes +/// to merge the two values. Do this now. +static void RewriteUsesOfClonedInstructions(BasicBlock *OrigHeader, + BasicBlock *OrigPreheader, + ValueToValueMapTy &ValueMap) { + // Remove PHI node entries that are no longer live. + BasicBlock::iterator I, E = OrigHeader->end(); + for (I = OrigHeader->begin(); PHINode *PN = dyn_cast(I); ++I) + PN->removeIncomingValue(PN->getBasicBlockIndex(OrigPreheader)); - BasicBlock *OrigLatch = L->getLoopLatch(); - if (!OrigLatch) return false; + // Now fix up users of the instructions in OrigHeader, inserting PHI nodes + // as necessary. + SSAUpdater SSA; + for (I = OrigHeader->begin(); I != E; ++I) { + Value *OrigHeaderVal = I; + + // If there are no uses of the value (e.g. because it returns void), there + // is nothing to rewrite. + if (OrigHeaderVal->use_empty()) + continue; + + Value *OrigPreHeaderVal = ValueMap[OrigHeaderVal]; + + // The value now exits in two versions: the initial value in the preheader + // and the loop "next" value in the original header. + SSA.Initialize(OrigHeaderVal->getType(), OrigHeaderVal->getName()); + SSA.AddAvailableValue(OrigHeader, OrigHeaderVal); + SSA.AddAvailableValue(OrigPreheader, OrigPreHeaderVal); + + // Visit each use of the OrigHeader instruction. + for (Value::use_iterator UI = OrigHeaderVal->use_begin(), + UE = OrigHeaderVal->use_end(); UI != UE; ) { + // Grab the use before incrementing the iterator. + Use &U = UI.getUse(); + + // Increment the iterator before removing the use from the list. + ++UI; + + // SSAUpdater can't handle a non-PHI use in the same block as an + // earlier def. We can easily handle those cases manually. + Instruction *UserInst = cast(U.getUser()); + if (!isa(UserInst)) { + BasicBlock *UserBB = UserInst->getParent(); - BasicBlock *OrigHeader = L->getHeader(); + // The original users in the OrigHeader are already using the + // original definitions. + if (UserBB == OrigHeader) + continue; + // Users in the OrigPreHeader need to use the value to which the + // original definitions are mapped. + if (UserBB == OrigPreheader) { + U = OrigPreHeaderVal; + continue; + } + } + + // Anything else can be handled by SSAUpdater. + SSA.RewriteUse(U); + } + } +} + +/// Determine whether the instructions in this range my be safely and cheaply +/// speculated. This is not an important enough situation to develop complex +/// heuristics. We handle a single arithmetic instruction along with any type +/// conversions. +static bool shouldSpeculateInstrs(BasicBlock::iterator Begin, + BasicBlock::iterator End) { + bool seenIncrement = false; + for (BasicBlock::iterator I = Begin; I != End; ++I) { + + if (!isSafeToSpeculativelyExecute(I)) + return false; + + if (isa(I)) + continue; + + switch (I->getOpcode()) { + default: + return false; + case Instruction::GetElementPtr: + // GEPs are cheap if all indices are constant. + if (!cast(I)->hasAllConstantIndices()) + return false; + // fall-thru to increment case + case Instruction::Add: + case Instruction::Sub: + case Instruction::And: + case Instruction::Or: + case Instruction::Xor: + case Instruction::Shl: + case Instruction::LShr: + case Instruction::AShr: + if (seenIncrement) + return false; + seenIncrement = true; + break; + case Instruction::Trunc: + case Instruction::ZExt: + case Instruction::SExt: + // ignore type conversions + break; + } + } + return true; +} + +/// Fold the loop tail into the loop exit by speculating the loop tail +/// instructions. Typically, this is a single post-increment. In the case of a +/// simple 2-block loop, hoisting the increment can be much better than +/// duplicating the entire loop header. In the cast of loops with early exits, +/// rotation will not work anyway, but simplifyLoopLatch will put the loop in +/// canonical form so downstream passes can handle it. +/// +/// I don't believe this invalidates SCEV. +void LoopRotate::simplifyLoopLatch(Loop *L) { + BasicBlock *Latch = L->getLoopLatch(); + if (!Latch || Latch->hasAddressTaken()) + return; + + BranchInst *Jmp = dyn_cast(Latch->getTerminator()); + if (!Jmp || !Jmp->isUnconditional()) + return; + + BasicBlock *LastExit = Latch->getSinglePredecessor(); + if (!LastExit || !L->isLoopExiting(LastExit)) + return; + + BranchInst *BI = dyn_cast(LastExit->getTerminator()); + if (!BI) + return; + + if (!shouldSpeculateInstrs(Latch->begin(), Jmp)) + return; + + DEBUG(dbgs() << "Folding loop latch " << Latch->getName() << " into " + << LastExit->getName() << "\n"); + + // Hoist the instructions from Latch into LastExit. + LastExit->getInstList().splice(BI, Latch->getInstList(), Latch->begin(), Jmp); + + unsigned FallThruPath = BI->getSuccessor(0) == Latch ? 0 : 1; + BasicBlock *Header = Jmp->getSuccessor(0); + assert(Header == L->getHeader() && "expected a backward branch"); + + // Remove Latch from the CFG so that LastExit becomes the new Latch. + BI->setSuccessor(FallThruPath, Header); + Latch->replaceSuccessorsPhiUsesWith(LastExit); + Jmp->eraseFromParent(); + + // Nuke the Latch block. + assert(Latch->empty() && "unable to evacuate Latch"); + LI->removeBlock(Latch); + if (DominatorTree *DT = getAnalysisIfAvailable()) + DT->eraseNode(Latch); + Latch->eraseFromParent(); +} + +/// Rotate loop LP. Return true if the loop is rotated. +bool LoopRotate::rotateLoop(Loop *L) { // If the loop has only one block then there is not much to rotate. if (L->getBlocks().size() == 1) return false; + BasicBlock *OrigHeader = L->getHeader(); + BasicBlock *OrigLatch = L->getLoopLatch(); + + BranchInst *BI = dyn_cast(OrigHeader->getTerminator()); + if (BI == 0 || BI->isUnconditional()) + return false; + // If the loop header is not one of the loop exiting blocks then // either this loop is already rotated or it is not // suitable for loop rotation transformations. if (!L->isLoopExiting(OrigHeader)) return false; - BranchInst *BI = dyn_cast(OrigHeader->getTerminator()); - if (!BI) - return false; - assert(BI->isConditional() && "Branch Instruction is not conditional"); - - // Updating PHInodes in loops with multiple exits adds complexity. - // Keep it simple, and restrict loop rotation to loops with one exit only. - // In future, lift this restriction and support for multiple exits if - // required. - SmallVector ExitBlocks; - L->getExitBlocks(ExitBlocks); - if (ExitBlocks.size() > 1) + // If the loop latch already contains a branch that leaves the loop then the + // loop is already rotated. + if (OrigLatch == 0 || L->isLoopExiting(OrigLatch)) return false; - // Check size of original header and reject loop if it is very big. + // Check size of original header and reject loop if it is very big or we can't + // duplicate blocks inside it. { CodeMetrics Metrics; - Metrics.analyzeBasicBlock(OrigHeader); + Metrics.analyzeBasicBlock(OrigHeader, *TTI); + if (Metrics.notDuplicatable) { + DEBUG(dbgs() << "LoopRotation: NOT rotating - contains non duplicatable" + << " instructions: "; L->dump()); + return false; + } if (Metrics.NumInsts > MAX_HEADER_SIZE) return false; } // Now, this loop is suitable for rotation. + BasicBlock *OrigPreheader = L->getLoopPreheader(); + + // If the loop could not be converted to canonical form, it must have an + // indirectbr in it, just give up. + if (OrigPreheader == 0) + return false; // Anything ScalarEvolution may know about this loop or the PHI nodes // in its header will soon be invalidated. if (ScalarEvolution *SE = getAnalysisIfAvailable()) SE->forgetLoop(L); + DEBUG(dbgs() << "LoopRotation: rotating "; L->dump()); + // Find new Loop header. NewHeader is a Header's one and only successor // that is inside loop. Header's other successor is outside the // loop. Otherwise loop is not suitable for rotation. @@ -153,9 +316,9 @@ bool LoopRotate::rotateLoop(Loop *L) { if (L->contains(Exit)) std::swap(Exit, NewHeader); assert(NewHeader && "Unable to determine new loop header"); - assert(L->contains(NewHeader) && !L->contains(Exit) && + assert(L->contains(NewHeader) && !L->contains(Exit) && "Unable to determine loop header and exit blocks"); - + // This code assumes that the new header has exactly one predecessor. // Remove any single-entry PHI nodes in it. assert(NewHeader->getSinglePredecessor() && @@ -170,14 +333,14 @@ bool LoopRotate::rotateLoop(Loop *L) { // For PHI nodes, the value available in OldPreHeader is just the // incoming value from OldPreHeader. for (; PHINode *PN = dyn_cast(I); ++I) - ValueMap[PN] = PN->getIncomingValue(PN->getBasicBlockIndex(OrigPreHeader)); + ValueMap[PN] = PN->getIncomingValueForBlock(OrigPreheader); // For the rest of the instructions, either hoist to the OrigPreheader if // possible or create a clone in the OldPreHeader if not. - TerminatorInst *LoopEntryBranch = OrigPreHeader->getTerminator(); + TerminatorInst *LoopEntryBranch = OrigPreheader->getTerminator(); while (I != E) { Instruction *Inst = I++; - + // If the instruction's operands are invariant and it doesn't read or write // memory, then it is safe to hoist. Doing this doesn't change the order of // execution in the preheader, but does prevent the instruction from @@ -186,18 +349,19 @@ bool LoopRotate::rotateLoop(Loop *L) { // memory (without proving that the loop doesn't write). if (L->hasLoopInvariantOperands(Inst) && !Inst->mayReadFromMemory() && !Inst->mayWriteToMemory() && - !isa(Inst)) { + !isa(Inst) && !isa(Inst) && + !isa(Inst)) { Inst->moveBefore(LoopEntryBranch); continue; } - + // Otherwise, create a duplicate of the instruction. Instruction *C = Inst->clone(); - + // Eagerly remap the operands of the instruction. RemapInstruction(C, ValueMap, RF_NoModuleLevelChanges|RF_IgnoreMissingEntries); - + // With the operands remapped, see if the instruction constant folds or is // otherwise simplifyable. This commonly occurs because the entry from PHI // nodes allows icmps and other instructions to fold. @@ -222,221 +386,123 @@ bool LoopRotate::rotateLoop(Loop *L) { for (unsigned i = 0, e = TI->getNumSuccessors(); i != e; ++i) for (BasicBlock::iterator BI = TI->getSuccessor(i)->begin(); PHINode *PN = dyn_cast(BI); ++BI) - PN->addIncoming(PN->getIncomingValueForBlock(OrigHeader), OrigPreHeader); + PN->addIncoming(PN->getIncomingValueForBlock(OrigHeader), OrigPreheader); // Now that OrigPreHeader has a clone of OrigHeader's terminator, remove // OrigPreHeader's old terminator (the original branch into the loop), and // remove the corresponding incoming values from the PHI nodes in OrigHeader. LoopEntryBranch->eraseFromParent(); - for (I = OrigHeader->begin(); PHINode *PN = dyn_cast(I); ++I) - PN->removeIncomingValue(PN->getBasicBlockIndex(OrigPreHeader)); - // Now fix up users of the instructions in OrigHeader, inserting PHI nodes - // as necessary. - SSAUpdater SSA; - for (I = OrigHeader->begin(); I != E; ++I) { - Value *OrigHeaderVal = I; - Value *OrigPreHeaderVal = ValueMap[OrigHeaderVal]; - - // If there are no uses of the value (e.g. because it returns void), there - // is nothing to rewrite. - if (OrigHeaderVal->use_empty() && OrigPreHeaderVal->use_empty()) - continue; - - // The value now exits in two versions: the initial value in the preheader - // and the loop "next" value in the original header. - SSA.Initialize(OrigHeaderVal->getType(), OrigHeaderVal->getName()); - SSA.AddAvailableValue(OrigHeader, OrigHeaderVal); - SSA.AddAvailableValue(OrigPreHeader, OrigPreHeaderVal); - - // Visit each use of the OrigHeader instruction. - for (Value::use_iterator UI = OrigHeaderVal->use_begin(), - UE = OrigHeaderVal->use_end(); UI != UE; ) { - // Grab the use before incrementing the iterator. - Use &U = UI.getUse(); - - // Increment the iterator before removing the use from the list. - ++UI; - - // SSAUpdater can't handle a non-PHI use in the same block as an - // earlier def. We can easily handle those cases manually. - Instruction *UserInst = cast(U.getUser()); - if (!isa(UserInst)) { - BasicBlock *UserBB = UserInst->getParent(); + // If there were any uses of instructions in the duplicated block outside the + // loop, update them, inserting PHI nodes as required + RewriteUsesOfClonedInstructions(OrigHeader, OrigPreheader, ValueMap); - // The original users in the OrigHeader are already using the - // original definitions. - if (UserBB == OrigHeader) - continue; + // NewHeader is now the header of the loop. + L->moveToHeader(NewHeader); + assert(L->getHeader() == NewHeader && "Latch block is our new header"); + + + // At this point, we've finished our major CFG changes. As part of cloning + // the loop into the preheader we've simplified instructions and the + // duplicated conditional branch may now be branching on a constant. If it is + // branching on a constant and if that constant means that we enter the loop, + // then we fold away the cond branch to an uncond branch. This simplifies the + // loop in cases important for nested loops, and it also means we don't have + // to split as many edges. + BranchInst *PHBI = cast(OrigPreheader->getTerminator()); + assert(PHBI->isConditional() && "Should be clone of BI condbr!"); + if (!isa(PHBI->getCondition()) || + PHBI->getSuccessor(cast(PHBI->getCondition())->isZero()) + != NewHeader) { + // The conditional branch can't be folded, handle the general case. + // Update DominatorTree to reflect the CFG change we just made. Then split + // edges as necessary to preserve LoopSimplify form. + if (DominatorTree *DT = getAnalysisIfAvailable()) { + // Everything that was dominated by the old loop header is now dominated + // by the original loop preheader. Conceptually the header was merged + // into the preheader, even though we reuse the actual block as a new + // loop latch. + DomTreeNode *OrigHeaderNode = DT->getNode(OrigHeader); + SmallVector HeaderChildren(OrigHeaderNode->begin(), + OrigHeaderNode->end()); + DomTreeNode *OrigPreheaderNode = DT->getNode(OrigPreheader); + for (unsigned I = 0, E = HeaderChildren.size(); I != E; ++I) + DT->changeImmediateDominator(HeaderChildren[I], OrigPreheaderNode); + + assert(DT->getNode(Exit)->getIDom() == OrigPreheaderNode); + assert(DT->getNode(NewHeader)->getIDom() == OrigPreheaderNode); + + // Update OrigHeader to be dominated by the new header block. + DT->changeImmediateDominator(OrigHeader, OrigLatch); + } - // Users in the OrigPreHeader need to use the value to which the - // original definitions are mapped. - if (UserBB == OrigPreHeader) { - U = OrigPreHeaderVal; - continue; + // Right now OrigPreHeader has two successors, NewHeader and ExitBlock, and + // thus is not a preheader anymore. + // Split the edge to form a real preheader. + BasicBlock *NewPH = SplitCriticalEdge(OrigPreheader, NewHeader, this); + NewPH->setName(NewHeader->getName() + ".lr.ph"); + + // Preserve canonical loop form, which means that 'Exit' should have only + // one predecessor. + BasicBlock *ExitSplit = SplitCriticalEdge(L->getLoopLatch(), Exit, this); + ExitSplit->moveBefore(Exit); + } else { + // We can fold the conditional branch in the preheader, this makes things + // simpler. The first step is to remove the extra edge to the Exit block. + Exit->removePredecessor(OrigPreheader, true /*preserve LCSSA*/); + BranchInst *NewBI = BranchInst::Create(NewHeader, PHBI); + NewBI->setDebugLoc(PHBI->getDebugLoc()); + PHBI->eraseFromParent(); + + // With our CFG finalized, update DomTree if it is available. + if (DominatorTree *DT = getAnalysisIfAvailable()) { + // Update OrigHeader to be dominated by the new header block. + DT->changeImmediateDominator(NewHeader, OrigPreheader); + DT->changeImmediateDominator(OrigHeader, OrigLatch); + + // Brute force incremental dominator tree update. Call + // findNearestCommonDominator on all CFG predecessors of each child of the + // original header. + DomTreeNode *OrigHeaderNode = DT->getNode(OrigHeader); + SmallVector HeaderChildren(OrigHeaderNode->begin(), + OrigHeaderNode->end()); + bool Changed; + do { + Changed = false; + for (unsigned I = 0, E = HeaderChildren.size(); I != E; ++I) { + DomTreeNode *Node = HeaderChildren[I]; + BasicBlock *BB = Node->getBlock(); + + pred_iterator PI = pred_begin(BB); + BasicBlock *NearestDom = *PI; + for (pred_iterator PE = pred_end(BB); PI != PE; ++PI) + NearestDom = DT->findNearestCommonDominator(NearestDom, *PI); + + // Remember if this changes the DomTree. + if (Node->getIDom()->getBlock() != NearestDom) { + DT->changeImmediateDominator(BB, NearestDom); + Changed = true; + } } - } - // Anything else can be handled by SSAUpdater. - SSA.RewriteUse(U); + // If the dominator changed, this may have an effect on other + // predecessors, continue until we reach a fixpoint. + } while (Changed); } } - // NewHeader is now the header of the loop. - L->moveToHeader(NewHeader); + assert(L->getLoopPreheader() && "Invalid loop preheader after loop rotation"); + assert(L->getLoopLatch() && "Invalid loop latch after loop rotation"); + + // Now that the CFG and DomTree are in a consistent state again, try to merge + // the OrigHeader block into OrigLatch. This will succeed if they are + // connected by an unconditional branch. This is just a cleanup so the + // emitted code isn't too gross in this common case. + MergeBlockIntoPredecessor(OrigHeader, this); - // Move the original header to the bottom of the loop, where it now more - // naturally belongs. This isn't necessary for correctness, and CodeGen can - // usually reorder blocks on its own to fix things like this up, but it's - // still nice to keep the IR readable. - // - // The original header should have only one predecessor at this point, since - // we checked that the loop had a proper preheader and unique backedge before - // we started. - assert(OrigHeader->getSinglePredecessor() && - "Original loop header has too many predecessors after loop rotation!"); - OrigHeader->moveAfter(OrigHeader->getSinglePredecessor()); - - // Also, since this original header only has one predecessor, zap its - // PHI nodes, which are now trivial. - FoldSingleEntryPHINodes(OrigHeader); - - // TODO: We could just go ahead and merge OrigHeader into its predecessor - // at this point, if we don't mind updating dominator info. - - // Establish a new preheader, update dominators, etc. - preserveCanonicalLoopForm(L, OrigHeader, OrigPreHeader, OrigLatch, - NewHeader, Exit); + DEBUG(dbgs() << "LoopRotation: into "; L->dump()); ++NumRotated; return true; } - -/// After loop rotation, loop pre-header has multiple sucessors. -/// Insert one forwarding basic block to ensure that loop pre-header -/// has only one successor. -void LoopRotate::preserveCanonicalLoopForm(Loop *L, BasicBlock *OrigHeader, - BasicBlock *OrigPreHeader, - BasicBlock *OrigLatch, - BasicBlock *NewHeader, - BasicBlock *Exit) { - - // Right now original pre-header has two successors, new header and - // exit block. Insert new block between original pre-header and - // new header such that loop's new pre-header has only one successor. - BasicBlock *NewPreHeader = - BasicBlock::Create(OrigHeader->getContext(), "bb.nph", - OrigHeader->getParent(), NewHeader); - if (Loop *PL = LI->getLoopFor(OrigPreHeader)) - PL->addBasicBlockToLoop(NewPreHeader, LI->getBase()); - BranchInst::Create(NewHeader, NewPreHeader); - - BranchInst *OrigPH_BI = cast(OrigPreHeader->getTerminator()); - if (OrigPH_BI->getSuccessor(0) == NewHeader) - OrigPH_BI->setSuccessor(0, NewPreHeader); - else { - assert(OrigPH_BI->getSuccessor(1) == NewHeader && - "Unexpected original pre-header terminator"); - OrigPH_BI->setSuccessor(1, NewPreHeader); - } - - PHINode *PN; - for (BasicBlock::iterator I = NewHeader->begin(); - (PN = dyn_cast(I)); ++I) { - int index = PN->getBasicBlockIndex(OrigPreHeader); - assert(index != -1 && "Expected incoming value from Original PreHeader"); - PN->setIncomingBlock(index, NewPreHeader); - assert(PN->getBasicBlockIndex(OrigPreHeader) == -1 && - "Expected only one incoming value from Original PreHeader"); - } - - if (DominatorTree *DT = getAnalysisIfAvailable()) { - DT->addNewBlock(NewPreHeader, OrigPreHeader); - DT->changeImmediateDominator(L->getHeader(), NewPreHeader); - DT->changeImmediateDominator(Exit, OrigPreHeader); - for (Loop::block_iterator BI = L->block_begin(), BE = L->block_end(); - BI != BE; ++BI) { - BasicBlock *B = *BI; - if (L->getHeader() != B) { - DomTreeNode *Node = DT->getNode(B); - if (Node && Node->getBlock() == OrigHeader) - DT->changeImmediateDominator(*BI, L->getHeader()); - } - } - DT->changeImmediateDominator(OrigHeader, OrigLatch); - } - - if (DominanceFrontier *DF = getAnalysisIfAvailable()) { - // New Preheader's dominance frontier is Exit block. - DominanceFrontier::DomSetType NewPHSet; - NewPHSet.insert(Exit); - DF->addBasicBlock(NewPreHeader, NewPHSet); - - // New Header's dominance frontier now includes itself and Exit block - DominanceFrontier::iterator HeadI = DF->find(L->getHeader()); - if (HeadI != DF->end()) { - DominanceFrontier::DomSetType & HeaderSet = HeadI->second; - HeaderSet.clear(); - HeaderSet.insert(L->getHeader()); - HeaderSet.insert(Exit); - } else { - DominanceFrontier::DomSetType HeaderSet; - HeaderSet.insert(L->getHeader()); - HeaderSet.insert(Exit); - DF->addBasicBlock(L->getHeader(), HeaderSet); - } - - // Original header (new Loop Latch)'s dominance frontier is Exit. - DominanceFrontier::iterator LatchI = DF->find(L->getLoopLatch()); - if (LatchI != DF->end()) { - DominanceFrontier::DomSetType &LatchSet = LatchI->second; - LatchSet = LatchI->second; - LatchSet.clear(); - LatchSet.insert(Exit); - } else { - DominanceFrontier::DomSetType LatchSet; - LatchSet.insert(Exit); - DF->addBasicBlock(L->getHeader(), LatchSet); - } - - // If a loop block dominates new loop latch then add to its frontiers - // new header and Exit and remove new latch (which is equal to original - // header). - BasicBlock *NewLatch = L->getLoopLatch(); - - assert(NewLatch == OrigHeader && "NewLatch is inequal to OrigHeader"); - - if (DominatorTree *DT = getAnalysisIfAvailable()) { - for (Loop::block_iterator BI = L->block_begin(), BE = L->block_end(); - BI != BE; ++BI) { - BasicBlock *B = *BI; - if (!DT->dominates(B, NewLatch)) continue; - - DominanceFrontier::iterator BDFI = DF->find(B); - if (BDFI != DF->end()) { - DominanceFrontier::DomSetType &BSet = BDFI->second; - BSet.erase(NewLatch); - BSet.insert(L->getHeader()); - BSet.insert(Exit); - } else { - DominanceFrontier::DomSetType BSet; - BSet.insert(L->getHeader()); - BSet.insert(Exit); - DF->addBasicBlock(B, BSet); - } - } - } - } - - // Preserve canonical loop form, which means Exit block should - // have only one predecessor. - SplitEdge(L->getLoopLatch(), Exit, this); - - assert(NewHeader && L->getHeader() == NewHeader && - "Invalid loop header after loop rotation"); - assert(NewPreHeader && L->getLoopPreheader() == NewPreHeader && - "Invalid loop preheader after loop rotation"); - assert(L->getLoopLatch() && - "Invalid loop latch after loop rotation"); -}