/// RewriteSingleStoreAlloca - If there is only a single store to this value, /// replace any loads of it that are directly dominated by the definition with /// the value stored. void PromoteMem2Reg::RewriteSingleStoreAlloca(AllocaInst *AI, AllocaInfo &Info, LargeBlockInfo &LBI) { StoreInst *OnlyStore = Info.OnlyStore; bool StoringGlobalVal = !isa<Instruction>(OnlyStore->getOperand(0)); BasicBlock *StoreBB = OnlyStore->getParent(); int StoreIndex = -1; // Clear out UsingBlocks. We will reconstruct it here if needed. Info.UsingBlocks.clear(); for (Value::use_iterator UI = AI->use_begin(), E = AI->use_end(); UI != E; ) { Instruction *UserInst = cast<Instruction>(*UI++); if (!isa<LoadInst>(UserInst)) { assert(UserInst == OnlyStore && "Should only have load/stores"); continue; } LoadInst *LI = cast<LoadInst>(UserInst); // Okay, if we have a load from the alloca, we want to replace it with the // only value stored to the alloca. We can do this if the value is // dominated by the store. If not, we use the rest of the mem2reg machinery // to insert the phi nodes as needed. if (!StoringGlobalVal) { // Non-instructions are always dominated. if (LI->getParent() == StoreBB) { // If we have a use that is in the same block as the store, compare the // indices of the two instructions to see which one came first. If the // load came before the store, we can't handle it. if (StoreIndex == -1) StoreIndex = LBI.getInstructionIndex(OnlyStore); if (unsigned(StoreIndex) > LBI.getInstructionIndex(LI)) { // Can't handle this load, bail out. Info.UsingBlocks.push_back(StoreBB); continue; } } else if (LI->getParent() != StoreBB && !dominates(StoreBB, LI->getParent())) { // If the load and store are in different blocks, use BB dominance to // check their relationships. If the store doesn't dom the use, bail // out. Info.UsingBlocks.push_back(LI->getParent()); continue; } } // Otherwise, we *can* safely rewrite this load. Value *ReplVal = OnlyStore->getOperand(0); // If the replacement value is the load, this must occur in unreachable // code. if (ReplVal == LI) ReplVal = UndefValue::get(LI->getType()); LI->replaceAllUsesWith(ReplVal); if (AST && LI->getType()->isPointerTy()) AST->deleteValue(LI); LI->eraseFromParent(); LBI.deleteValue(LI); } }
bool Scalarizer::visitStoreInst(StoreInst &SI) { if (!ScalarizeLoadStore) return false; if (!SI.isSimple()) return false; VectorLayout Layout; Value *FullValue = SI.getValueOperand(); if (!getVectorLayout(FullValue->getType(), SI.getAlignment(), Layout)) return false; unsigned NumElems = Layout.VecTy->getNumElements(); IRBuilder<> Builder(SI.getParent(), &SI); Scatterer Ptr = scatter(&SI, SI.getPointerOperand()); Scatterer Val = scatter(&SI, FullValue); ValueVector Stores; Stores.resize(NumElems); for (unsigned I = 0; I < NumElems; ++I) { unsigned Align = Layout.getElemAlign(I); Stores[I] = Builder.CreateAlignedStore(Val[I], Ptr[I], Align); } transferMetadata(&SI, Stores); return true; }
void DSGraphStats::visitStore(StoreInst &SI) { if (isNodeForValueUntyped(SI.getOperand(1), 0,SI.getParent()->getParent())) { NumUntypedMemAccesses++; } else { NumTypedMemAccesses++; } }
/// SimplifyStoreAtEndOfBlock - Turn things like: /// if () { *P = v1; } else { *P = v2 } /// into a phi node with a store in the successor. /// /// Simplify things like: /// *P = v1; if () { *P = v2; } /// into a phi node with a store in the successor. /// bool InstCombiner::SimplifyStoreAtEndOfBlock(StoreInst &SI) { BasicBlock *StoreBB = SI.getParent(); // Check to see if the successor block has exactly two incoming edges. If // so, see if the other predecessor contains a store to the same location. // if so, insert a PHI node (if needed) and move the stores down. BasicBlock *DestBB = StoreBB->getTerminator()->getSuccessor(0); // Determine whether Dest has exactly two predecessors and, if so, compute // the other predecessor. pred_iterator PI = pred_begin(DestBB); BasicBlock *P = *PI; BasicBlock *OtherBB = nullptr; if (P != StoreBB) OtherBB = P; if (++PI == pred_end(DestBB)) return false; P = *PI; if (P != StoreBB) { if (OtherBB) return false; OtherBB = P; } if (++PI != pred_end(DestBB)) return false; // Bail out if all the relevant blocks aren't distinct (this can happen, // for example, if SI is in an infinite loop) if (StoreBB == DestBB || OtherBB == DestBB) return false; // Verify that the other block ends in a branch and is not otherwise empty. BasicBlock::iterator BBI(OtherBB->getTerminator()); BranchInst *OtherBr = dyn_cast<BranchInst>(BBI); if (!OtherBr || BBI == OtherBB->begin()) return false; // If the other block ends in an unconditional branch, check for the 'if then // else' case. there is an instruction before the branch. StoreInst *OtherStore = nullptr; if (OtherBr->isUnconditional()) { --BBI; // Skip over debugging info. while (isa<DbgInfoIntrinsic>(BBI) || (isa<BitCastInst>(BBI) && BBI->getType()->isPointerTy())) { if (BBI==OtherBB->begin()) return false; --BBI; } // If this isn't a store, isn't a store to the same location, or is not the // right kind of store, bail out. OtherStore = dyn_cast<StoreInst>(BBI); if (!OtherStore || OtherStore->getOperand(1) != SI.getOperand(1) || !SI.isSameOperationAs(OtherStore)) return false; } else { // Otherwise, the other block ended with a conditional branch. If one of the // destinations is StoreBB, then we have the if/then case. if (OtherBr->getSuccessor(0) != StoreBB && OtherBr->getSuccessor(1) != StoreBB) return false; // Okay, we know that OtherBr now goes to Dest and StoreBB, so this is an // if/then triangle. See if there is a store to the same ptr as SI that // lives in OtherBB. for (;; --BBI) { // Check to see if we find the matching store. if ((OtherStore = dyn_cast<StoreInst>(BBI))) { if (OtherStore->getOperand(1) != SI.getOperand(1) || !SI.isSameOperationAs(OtherStore)) return false; break; } // If we find something that may be using or overwriting the stored // value, or if we run out of instructions, we can't do the xform. if (BBI->mayReadFromMemory() || BBI->mayWriteToMemory() || BBI == OtherBB->begin()) return false; } // In order to eliminate the store in OtherBr, we have to // make sure nothing reads or overwrites the stored value in // StoreBB. for (BasicBlock::iterator I = StoreBB->begin(); &*I != &SI; ++I) { // FIXME: This should really be AA driven. if (I->mayReadFromMemory() || I->mayWriteToMemory()) return false; } } // Insert a PHI node now if we need it. Value *MergedVal = OtherStore->getOperand(0); if (MergedVal != SI.getOperand(0)) { PHINode *PN = PHINode::Create(MergedVal->getType(), 2, "storemerge"); PN->addIncoming(SI.getOperand(0), SI.getParent()); PN->addIncoming(OtherStore->getOperand(0), OtherBB); MergedVal = InsertNewInstBefore(PN, DestBB->front()); } // Advance to a place where it is safe to insert the new store and // insert it. BBI = DestBB->getFirstInsertionPt(); StoreInst *NewSI = new StoreInst(MergedVal, SI.getOperand(1), SI.isVolatile(), SI.getAlignment(), SI.getOrdering(), SI.getSynchScope()); InsertNewInstBefore(NewSI, *BBI); NewSI->setDebugLoc(OtherStore->getDebugLoc()); // If the two stores had AA tags, merge them. AAMDNodes AATags; SI.getAAMetadata(AATags); if (AATags) { OtherStore->getAAMetadata(AATags, /* Merge = */ true); NewSI->setAAMetadata(AATags); } // Nuke the old stores. EraseInstFromFunction(SI); EraseInstFromFunction(*OtherStore); return true; }
/// \brief Rewrite as many loads as possible given a single store. /// /// When there is only a single store, we can use the domtree to trivially /// replace all of the dominated loads with the stored value. Do so, and return /// true if this has successfully promoted the alloca entirely. If this returns /// false there were some loads which were not dominated by the single store /// and thus must be phi-ed with undef. We fall back to the standard alloca /// promotion algorithm in that case. static bool rewriteSingleStoreAlloca(AllocaInst *AI, AllocaInfo &Info, LargeBlockInfo &LBI, DominatorTree &DT, AliasSetTracker *AST) { StoreInst *OnlyStore = Info.OnlyStore; bool StoringGlobalVal = !isa<Instruction>(OnlyStore->getOperand(0)); BasicBlock *StoreBB = OnlyStore->getParent(); int StoreIndex = -1; // Clear out UsingBlocks. We will reconstruct it here if needed. Info.UsingBlocks.clear(); for (Value::use_iterator UI = AI->use_begin(), E = AI->use_end(); UI != E;) { Instruction *UserInst = cast<Instruction>(*UI++); if (!isa<LoadInst>(UserInst)) { assert(UserInst == OnlyStore && "Should only have load/stores"); continue; } LoadInst *LI = cast<LoadInst>(UserInst); // Okay, if we have a load from the alloca, we want to replace it with the // only value stored to the alloca. We can do this if the value is // dominated by the store. If not, we use the rest of the mem2reg machinery // to insert the phi nodes as needed. if (!StoringGlobalVal) { // Non-instructions are always dominated. if (LI->getParent() == StoreBB) { // If we have a use that is in the same block as the store, compare the // indices of the two instructions to see which one came first. If the // load came before the store, we can't handle it. if (StoreIndex == -1) StoreIndex = LBI.getInstructionIndex(OnlyStore); if (unsigned(StoreIndex) > LBI.getInstructionIndex(LI)) { // Can't handle this load, bail out. Info.UsingBlocks.push_back(StoreBB); continue; } } else if (LI->getParent() != StoreBB && !DT.dominates(StoreBB, LI->getParent())) { // If the load and store are in different blocks, use BB dominance to // check their relationships. If the store doesn't dom the use, bail // out. Info.UsingBlocks.push_back(LI->getParent()); continue; } } // Otherwise, we *can* safely rewrite this load. Value *ReplVal = OnlyStore->getOperand(0); // If the replacement value is the load, this must occur in unreachable // code. if (ReplVal == LI) ReplVal = UndefValue::get(LI->getType()); LI->replaceAllUsesWith(ReplVal); if (AST && LI->getType()->isPointerTy()) AST->deleteValue(LI); LI->eraseFromParent(); LBI.deleteValue(LI); } // Finally, after the scan, check to see if the store is all that is left. if (!Info.UsingBlocks.empty()) return false; // If not, we'll have to fall back for the remainder. // Record debuginfo for the store and remove the declaration's // debuginfo. if (DbgDeclareInst *DDI = Info.DbgDeclare) { DIBuilder DIB(*AI->getParent()->getParent()->getParent()); ConvertDebugDeclareToDebugValue(DDI, Info.OnlyStore, DIB); DDI->eraseFromParent(); } // Remove the (now dead) store and alloca. Info.OnlyStore->eraseFromParent(); LBI.deleteValue(Info.OnlyStore); if (AST) AST->deleteValue(AI); AI->eraseFromParent(); LBI.deleteValue(AI); return true; }
Value* LoopTripCount::insertTripCount(Loop* L, Instruction* InsertPos) { // inspired from Loop::getCanonicalInductionVariable BasicBlock *H = L->getHeader(); BasicBlock* LoopPred = L->getLoopPredecessor(); BasicBlock* startBB = NULL;//which basicblock stores start value int OneStep = 0;// the extra add or plus step for calc Assert(LoopPred, "Require Loop has a Pred"); DEBUG(errs()<<"loop depth:"<<L->getLoopDepth()<<"\n"); /** whats difference on use of predecessor and preheader??*/ //RET_ON_FAIL(self->getLoopLatch()&&self->getLoopPreheader()); //assert(self->getLoopLatch() && self->getLoopPreheader() && "need loop simplify form" ); ret_null_fail(L->getLoopLatch(), "need loop simplify form"); BasicBlock* TE = NULL;//True Exit SmallVector<BasicBlock*,4> Exits; L->getExitingBlocks(Exits); if(Exits.size()==1) TE = Exits.front(); else{ if(std::find(Exits.begin(),Exits.end(),L->getLoopLatch())!=Exits.end()) TE = L->getLoopLatch(); else{ SmallVector<llvm::Loop::Edge,4> ExitEdges; L->getExitEdges(ExitEdges); //stl 用法,先把所有满足条件的元素(出口的结束符是不可到达)移动到数组的末尾,再统一删除 ExitEdges.erase(std::remove_if(ExitEdges.begin(), ExitEdges.end(), [](llvm::Loop::Edge& I){ return isa<UnreachableInst>(I.second->getTerminator()); }), ExitEdges.end()); if(ExitEdges.size()==1) TE = const_cast<BasicBlock*>(ExitEdges.front().first); } } //process true exit ret_null_fail(TE, "need have a true exit"); Instruction* IndOrNext = NULL; Value* END = NULL; //终止块的终止指令:分情况讨论branchinst,switchinst; //跳转指令br bool a1,a2;condition<-->bool if(isa<BranchInst>(TE->getTerminator())){ const BranchInst* EBR = cast<BranchInst>(TE->getTerminator()); Assert(EBR->isConditional(), "end branch is not conditional"); ICmpInst* EC = dyn_cast<ICmpInst>(EBR->getCondition()); if(EC->getPredicate() == EC->ICMP_SGT){ Assert(!L->contains(EBR->getSuccessor(0)), *EBR<<":abnormal exit with great than");//终止块的终止指令---->跳出执行循环外的指令 OneStep += 1; } else if(EC->getPredicate() == EC->ICMP_EQ) Assert(!L->contains(EBR->getSuccessor(0)), *EBR<<":abnormal exit with great than"); else if(EC->getPredicate() == EC->ICMP_SLT) { ret_null_fail(!L->contains(EBR->getSuccessor(1)), *EBR<<":abnormal exit with less than"); } else { ret_null_fail(0, *EC<<" unknow combination of end condition"); } IndOrNext = dyn_cast<Instruction>(castoff(EC->getOperand(0)));//去掉类型转化 END = EC->getOperand(1); DEBUG(errs()<<"end value:"<<*EC<<"\n"); }else if(isa<SwitchInst>(TE->getTerminator())){ SwitchInst* ESW = const_cast<SwitchInst*>(cast<SwitchInst>(TE->getTerminator())); IndOrNext = dyn_cast<Instruction>(castoff(ESW->getCondition())); for(auto I = ESW->case_begin(),E = ESW->case_end();I!=E;++I){ if(!L->contains(I.getCaseSuccessor())){ ret_null_fail(!END,""); assert(!END && "shouldn't have two ends"); END = I.getCaseValue(); } } DEBUG(errs()<<"end value:"<<*ESW<<"\n"); }else{ assert(0 && "unknow terminator type"); } ret_null_fail(L->isLoopInvariant(END), "end value should be loop invariant");//至此得END值 Value* start = NULL; Value* ind = NULL; Instruction* next = NULL; bool addfirst = false;//add before icmp ed DISABLE(errs()<<*IndOrNext<<"\n"); if(isa<LoadInst>(IndOrNext)){ //memory depend analysis Value* PSi = IndOrNext->getOperand(0);//point type Step.i int SICount[2] = {0};//store in predecessor count,store in loop body count for( auto I = PSi->use_begin(),E = PSi->use_end();I!=E;++I){ DISABLE(errs()<<**I<<"\n"); StoreInst* SI = dyn_cast<StoreInst>(*I); if(!SI || SI->getOperand(1) != PSi) continue; if(!start&&L->isLoopInvariant(SI->getOperand(0))) { if(SI->getParent() != LoopPred) if(std::find(pred_begin(LoopPred),pred_end(LoopPred),SI->getParent()) == pred_end(LoopPred)) continue; start = SI->getOperand(0); startBB = SI->getParent(); ++SICount[0]; } Instruction* SI0 = dyn_cast<Instruction>(SI->getOperand(0)); if(L->contains(SI) && SI0 && SI0->getOpcode() == Instruction::Add){ next = SI0; ++SICount[1]; } } Assert(SICount[0]==1 && SICount[1]==1, ""); ind = IndOrNext; }else{ if(isa<PHINode>(IndOrNext)){ PHINode* PHI = cast<PHINode>(IndOrNext); ind = IndOrNext; if(castoff(PHI->getIncomingValue(0)) == castoff(PHI->getIncomingValue(1)) && PHI->getParent() != H) ind = castoff(PHI->getIncomingValue(0)); addfirst = false; }else if(IndOrNext->getOpcode() == Instruction::Add){ next = IndOrNext; addfirst = true; }else{ Assert(0 ,"unknow how to analysis"); } for(auto I = H->begin();isa<PHINode>(I);++I){ PHINode* P = cast<PHINode>(I); if(ind && P == ind){ //start = P->getIncomingValueForBlock(L->getLoopPredecessor()); start = tryFindStart(P, L, startBB); next = dyn_cast<Instruction>(P->getIncomingValueForBlock(L->getLoopLatch())); }else if(next && P->getIncomingValueForBlock(L->getLoopLatch()) == next){ //start = P->getIncomingValueForBlock(L->getLoopPredecessor()); start = tryFindStart(P, L, startBB); ind = P; } } } Assert(start ,"couldn't find a start value"); //process complex loops later //DEBUG(if(L->getLoopDepth()>1 || !L->getSubLoops().empty()) return NULL); DEBUG(errs()<<"start value:"<<*start<<"\n"); DEBUG(errs()<<"ind value:"<<*ind<<"\n"); DEBUG(errs()<<"next value:"<<*next<<"\n"); //process non add later unsigned next_phi_idx = 0; ConstantInt* Step = NULL,*PrevStep = NULL;/*only used if next is phi node*/ ret_null_fail(next, ""); PHINode* next_phi = dyn_cast<PHINode>(next); do{ if(next_phi) { next = dyn_cast<Instruction>(next_phi->getIncomingValue(next_phi_idx)); ret_null_fail(next, ""); DEBUG(errs()<<"next phi "<<next_phi_idx<<":"<<*next<<"\n"); if(Step&&PrevStep){ Assert(Step->getSExtValue() == PrevStep->getSExtValue(),""); } PrevStep = Step; } Assert(next->getOpcode() == Instruction::Add , "why induction increment is not Add"); Assert(next->getOperand(0) == ind ,"why induction increment is not add it self"); Step = dyn_cast<ConstantInt>(next->getOperand(1)); Assert(Step,""); }while(next_phi && ++next_phi_idx<next_phi->getNumIncomingValues()); //RET_ON_FAIL(Step->equalsInt(1)); //assert(VERBOSE(Step->equalsInt(1),Step) && "why induction increment number is not 1"); Value* RES = NULL; //if there are no predecessor, we can insert code into start value basicblock IRBuilder<> Builder(InsertPos); Assert(start->getType()->isIntegerTy() && END->getType()->isIntegerTy() , " why increment is not integer type"); if(start->getType() != END->getType()){ start = Builder.CreateCast(CastInst::getCastOpcode(start, false, END->getType(), false),start,END->getType()); } if(Step->getType() != END->getType()){ //Because Step is a Constant, so it casted is constant Step = dyn_cast<ConstantInt>(Builder.CreateCast(CastInst::getCastOpcode(Step, false, END->getType(), false),Step,END->getType())); AssertRuntime(Step); } if(Step->isMinusOne()) RES = Builder.CreateSub(start,END); else//Step Couldn't be zero RES = Builder.CreateSub(END, start); if(addfirst) OneStep -= 1; if(Step->isMinusOne()) OneStep*=-1; assert(OneStep<=1 && OneStep>=-1); RES = (OneStep==1)?Builder.CreateAdd(RES,Step):(OneStep==-1)?Builder.CreateSub(RES, Step):RES; if(!Step->isMinusOne()&&!Step->isOne()) RES = Builder.CreateSDiv(RES, Step); RES->setName(H->getName()+".tc"); return RES; }
void PromoteMem2Reg::run() { Function &F = *DT.getRoot()->getParent(); if (AST) PointerAllocaValues.resize(Allocas.size()); AllocaDbgDeclares.resize(Allocas.size()); AllocaInfo Info; LargeBlockInfo LBI; for (unsigned AllocaNum = 0; AllocaNum != Allocas.size(); ++AllocaNum) { AllocaInst *AI = Allocas[AllocaNum]; assert(isAllocaPromotable(AI) && "Cannot promote non-promotable alloca!"); assert(AI->getParent()->getParent() == &F && "All allocas should be in the same function, which is same as DF!"); removeLifetimeIntrinsicUsers(AI); if (AI->use_empty()) { // If there are no uses of the alloca, just delete it now. if (AST) AST->deleteValue(AI); AI->eraseFromParent(); // Remove the alloca from the Allocas list, since it has been processed RemoveFromAllocasList(AllocaNum); ++NumDeadAlloca; continue; } // Calculate the set of read and write-locations for each alloca. This is // analogous to finding the 'uses' and 'definitions' of each variable. Info.AnalyzeAlloca(AI); // If there is only a single store to this value, replace any loads of // it that are directly dominated by the definition with the value stored. if (Info.DefiningBlocks.size() == 1) { RewriteSingleStoreAlloca(AI, Info, LBI); // Finally, after the scan, check to see if the store is all that is left. if (Info.UsingBlocks.empty()) { // Record debuginfo for the store and remove the declaration's debuginfo. if (DbgDeclareInst *DDI = Info.DbgDeclare) { if (!DIB) DIB = new DIBuilder(*DDI->getParent()->getParent()->getParent()); ConvertDebugDeclareToDebugValue(DDI, Info.OnlyStore, *DIB); DDI->eraseFromParent(); } // Remove the (now dead) store and alloca. Info.OnlyStore->eraseFromParent(); LBI.deleteValue(Info.OnlyStore); if (AST) AST->deleteValue(AI); AI->eraseFromParent(); LBI.deleteValue(AI); // The alloca has been processed, move on. RemoveFromAllocasList(AllocaNum); ++NumSingleStore; continue; } } // If the alloca is only read and written in one basic block, just perform a // linear sweep over the block to eliminate it. if (Info.OnlyUsedInOneBlock) { PromoteSingleBlockAlloca(AI, Info, LBI); // Finally, after the scan, check to see if the stores are all that is // left. if (Info.UsingBlocks.empty()) { // Remove the (now dead) stores and alloca. while (!AI->use_empty()) { StoreInst *SI = cast<StoreInst>(AI->use_back()); // Record debuginfo for the store before removing it. if (DbgDeclareInst *DDI = Info.DbgDeclare) { if (!DIB) DIB = new DIBuilder(*SI->getParent()->getParent()->getParent()); ConvertDebugDeclareToDebugValue(DDI, SI, *DIB); } SI->eraseFromParent(); LBI.deleteValue(SI); } if (AST) AST->deleteValue(AI); AI->eraseFromParent(); LBI.deleteValue(AI); // The alloca has been processed, move on. RemoveFromAllocasList(AllocaNum); // The alloca's debuginfo can be removed as well. if (DbgDeclareInst *DDI = Info.DbgDeclare) DDI->eraseFromParent(); ++NumLocalPromoted; continue; } } // If we haven't computed dominator tree levels, do so now. if (DomLevels.empty()) { SmallVector<DomTreeNode*, 32> Worklist; DomTreeNode *Root = DT.getRootNode(); DomLevels[Root] = 0; Worklist.push_back(Root); while (!Worklist.empty()) { DomTreeNode *Node = Worklist.pop_back_val(); unsigned ChildLevel = DomLevels[Node] + 1; for (DomTreeNode::iterator CI = Node->begin(), CE = Node->end(); CI != CE; ++CI) { DomLevels[*CI] = ChildLevel; Worklist.push_back(*CI); } } } // If we haven't computed a numbering for the BB's in the function, do so // now. if (BBNumbers.empty()) { unsigned ID = 0; for (Function::iterator I = F.begin(), E = F.end(); I != E; ++I) BBNumbers[I] = ID++; } // If we have an AST to keep updated, remember some pointer value that is // stored into the alloca. if (AST) PointerAllocaValues[AllocaNum] = Info.AllocaPointerVal; // Remember the dbg.declare intrinsic describing this alloca, if any. if (Info.DbgDeclare) AllocaDbgDeclares[AllocaNum] = Info.DbgDeclare; // Keep the reverse mapping of the 'Allocas' array for the rename pass. AllocaLookup[Allocas[AllocaNum]] = AllocaNum; // At this point, we're committed to promoting the alloca using IDF's, and // the standard SSA construction algorithm. Determine which blocks need PHI // nodes and see if we can optimize out some work by avoiding insertion of // dead phi nodes. DetermineInsertionPoint(AI, AllocaNum, Info); } if (Allocas.empty()) return; // All of the allocas must have been trivial! LBI.clear(); // Set the incoming values for the basic block to be null values for all of // the alloca's. We do this in case there is a load of a value that has not // been stored yet. In this case, it will get this null value. // RenamePassData::ValVector Values(Allocas.size()); for (unsigned i = 0, e = Allocas.size(); i != e; ++i) Values[i] = UndefValue::get(Allocas[i]->getAllocatedType()); // Walks all basic blocks in the function performing the SSA rename algorithm // and inserting the phi nodes we marked as necessary // std::vector<RenamePassData> RenamePassWorkList; RenamePassWorkList.push_back(RenamePassData(F.begin(), 0, Values)); do { RenamePassData RPD; RPD.swap(RenamePassWorkList.back()); RenamePassWorkList.pop_back(); // RenamePass may add new worklist entries. RenamePass(RPD.BB, RPD.Pred, RPD.Values, RenamePassWorkList); } while (!RenamePassWorkList.empty()); // The renamer uses the Visited set to avoid infinite loops. Clear it now. Visited.clear(); // Remove the allocas themselves from the function. for (unsigned i = 0, e = Allocas.size(); i != e; ++i) { Instruction *A = Allocas[i]; // If there are any uses of the alloca instructions left, they must be in // unreachable basic blocks that were not processed by walking the dominator // tree. Just delete the users now. if (!A->use_empty()) A->replaceAllUsesWith(UndefValue::get(A->getType())); if (AST) AST->deleteValue(A); A->eraseFromParent(); } // Remove alloca's dbg.declare instrinsics from the function. for (unsigned i = 0, e = AllocaDbgDeclares.size(); i != e; ++i) if (DbgDeclareInst *DDI = AllocaDbgDeclares[i]) DDI->eraseFromParent(); // Loop over all of the PHI nodes and see if there are any that we can get // rid of because they merge all of the same incoming values. This can // happen due to undef values coming into the PHI nodes. This process is // iterative, because eliminating one PHI node can cause others to be removed. bool EliminatedAPHI = true; while (EliminatedAPHI) { EliminatedAPHI = false; for (DenseMap<std::pair<BasicBlock*, unsigned>, PHINode*>::iterator I = NewPhiNodes.begin(), E = NewPhiNodes.end(); I != E;) { PHINode *PN = I->second; // If this PHI node merges one value and/or undefs, get the value. if (Value *V = SimplifyInstruction(PN, 0, 0, &DT)) { if (AST && PN->getType()->isPointerTy()) AST->deleteValue(PN); PN->replaceAllUsesWith(V); PN->eraseFromParent(); NewPhiNodes.erase(I++); EliminatedAPHI = true; continue; } ++I; } } // At this point, the renamer has added entries to PHI nodes for all reachable // code. Unfortunately, there may be unreachable blocks which the renamer // hasn't traversed. If this is the case, the PHI nodes may not // have incoming values for all predecessors. Loop over all PHI nodes we have // created, inserting undef values if they are missing any incoming values. // for (DenseMap<std::pair<BasicBlock*, unsigned>, PHINode*>::iterator I = NewPhiNodes.begin(), E = NewPhiNodes.end(); I != E; ++I) { // We want to do this once per basic block. As such, only process a block // when we find the PHI that is the first entry in the block. PHINode *SomePHI = I->second; BasicBlock *BB = SomePHI->getParent(); if (&BB->front() != SomePHI) continue; // Only do work here if there the PHI nodes are missing incoming values. We // know that all PHI nodes that were inserted in a block will have the same // number of incoming values, so we can just check any of them. if (SomePHI->getNumIncomingValues() == getNumPreds(BB)) continue; // Get the preds for BB. SmallVector<BasicBlock*, 16> Preds(pred_begin(BB), pred_end(BB)); // Ok, now we know that all of the PHI nodes are missing entries for some // basic blocks. Start by sorting the incoming predecessors for efficient // access. std::sort(Preds.begin(), Preds.end()); // Now we loop through all BB's which have entries in SomePHI and remove // them from the Preds list. for (unsigned i = 0, e = SomePHI->getNumIncomingValues(); i != e; ++i) { // Do a log(n) search of the Preds list for the entry we want. SmallVector<BasicBlock*, 16>::iterator EntIt = std::lower_bound(Preds.begin(), Preds.end(), SomePHI->getIncomingBlock(i)); assert(EntIt != Preds.end() && *EntIt == SomePHI->getIncomingBlock(i)&& "PHI node has entry for a block which is not a predecessor!"); // Remove the entry Preds.erase(EntIt); } // At this point, the blocks left in the preds list must have dummy // entries inserted into every PHI nodes for the block. Update all the phi // nodes in this block that we are inserting (there could be phis before // mem2reg runs). unsigned NumBadPreds = SomePHI->getNumIncomingValues(); BasicBlock::iterator BBI = BB->begin(); while ((SomePHI = dyn_cast<PHINode>(BBI++)) && SomePHI->getNumIncomingValues() == NumBadPreds) { Value *UndefVal = UndefValue::get(SomePHI->getType()); for (unsigned pred = 0, e = Preds.size(); pred != e; ++pred) SomePHI->addIncoming(UndefVal, Preds[pred]); } } NewPhiNodes.clear(); }
/// Rewrite as many loads as possible given a single store. /// /// When there is only a single store, we can use the domtree to trivially /// replace all of the dominated loads with the stored value. Do so, and return /// true if this has successfully promoted the alloca entirely. If this returns /// false there were some loads which were not dominated by the single store /// and thus must be phi-ed with undef. We fall back to the standard alloca /// promotion algorithm in that case. static bool rewriteSingleStoreAlloca(AllocaInst *AI, AllocaInfo &Info, LargeBlockInfo &LBI, const DataLayout &DL, DominatorTree &DT, AssumptionCache *AC) { StoreInst *OnlyStore = Info.OnlyStore; bool StoringGlobalVal = !isa<Instruction>(OnlyStore->getOperand(0)); BasicBlock *StoreBB = OnlyStore->getParent(); int StoreIndex = -1; // Clear out UsingBlocks. We will reconstruct it here if needed. Info.UsingBlocks.clear(); for (auto UI = AI->user_begin(), E = AI->user_end(); UI != E;) { Instruction *UserInst = cast<Instruction>(*UI++); if (!isa<LoadInst>(UserInst)) { assert(UserInst == OnlyStore && "Should only have load/stores"); continue; } LoadInst *LI = cast<LoadInst>(UserInst); // Okay, if we have a load from the alloca, we want to replace it with the // only value stored to the alloca. We can do this if the value is // dominated by the store. If not, we use the rest of the mem2reg machinery // to insert the phi nodes as needed. if (!StoringGlobalVal) { // Non-instructions are always dominated. if (LI->getParent() == StoreBB) { // If we have a use that is in the same block as the store, compare the // indices of the two instructions to see which one came first. If the // load came before the store, we can't handle it. if (StoreIndex == -1) StoreIndex = LBI.getInstructionIndex(OnlyStore); if (unsigned(StoreIndex) > LBI.getInstructionIndex(LI)) { // Can't handle this load, bail out. Info.UsingBlocks.push_back(StoreBB); continue; } } else if (LI->getParent() != StoreBB && !DT.dominates(StoreBB, LI->getParent())) { // If the load and store are in different blocks, use BB dominance to // check their relationships. If the store doesn't dom the use, bail // out. Info.UsingBlocks.push_back(LI->getParent()); continue; } } // Otherwise, we *can* safely rewrite this load. Value *ReplVal = OnlyStore->getOperand(0); // If the replacement value is the load, this must occur in unreachable // code. if (ReplVal == LI) ReplVal = UndefValue::get(LI->getType()); // If the load was marked as nonnull we don't want to lose // that information when we erase this Load. So we preserve // it with an assume. if (AC && LI->getMetadata(LLVMContext::MD_nonnull) && !isKnownNonZero(ReplVal, DL, 0, AC, LI, &DT)) addAssumeNonNull(AC, LI); LI->replaceAllUsesWith(ReplVal); LI->eraseFromParent(); LBI.deleteValue(LI); } // Finally, after the scan, check to see if the store is all that is left. if (!Info.UsingBlocks.empty()) return false; // If not, we'll have to fall back for the remainder. // Record debuginfo for the store and remove the declaration's // debuginfo. for (DbgVariableIntrinsic *DII : Info.DbgDeclares) { DIBuilder DIB(*AI->getModule(), /*AllowUnresolved*/ false); ConvertDebugDeclareToDebugValue(DII, Info.OnlyStore, DIB); DII->eraseFromParent(); LBI.deleteValue(DII); } // Remove the (now dead) store and alloca. Info.OnlyStore->eraseFromParent(); LBI.deleteValue(Info.OnlyStore); AI->eraseFromParent(); LBI.deleteValue(AI); return true; }