X-Git-Url: http://plrg.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FTransforms%2FUtils%2FCodeExtractor.cpp;h=fa6641fae51aa5236b01aa6bd5d34ffa76449fbe;hb=f362affa3a695164a94d275fb44d18f44ebb855a;hp=dbc358bf61d90c6d2f6808ff1aecd492223829e2;hpb=7848e68c1635ccba5a08d55314d4e5aed5ab54b9;p=oota-llvm.git diff --git a/lib/Transforms/Utils/CodeExtractor.cpp b/lib/Transforms/Utils/CodeExtractor.cpp index dbc358bf61d..fa6641fae51 100644 --- a/lib/Transforms/Utils/CodeExtractor.cpp +++ b/lib/Transforms/Utils/CodeExtractor.cpp @@ -1,10 +1,10 @@ //===- CodeExtractor.cpp - Pull code region into a new function -----------===// -// +// // The LLVM Compiler Infrastructure // -// This file was developed by the LLVM research group and is distributed under -// the University of Illinois Open Source License. See LICENSE.TXT for details. -// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// //===----------------------------------------------------------------------===// // // This file implements the interface to tear out a code region, such as an @@ -18,21 +18,25 @@ #include "llvm/DerivedTypes.h" #include "llvm/Instructions.h" #include "llvm/Intrinsics.h" +#include "llvm/LLVMContext.h" #include "llvm/Module.h" #include "llvm/Pass.h" #include "llvm/Analysis/Dominators.h" #include "llvm/Analysis/LoopInfo.h" #include "llvm/Analysis/Verifier.h" #include "llvm/Transforms/Utils/BasicBlockUtils.h" -#include "Support/CommandLine.h" -#include "Support/Debug.h" -#include "Support/StringExtras.h" +#include "llvm/Support/CommandLine.h" +#include "llvm/Support/Debug.h" +#include "llvm/Support/ErrorHandling.h" +#include "llvm/Support/raw_ostream.h" +#include "llvm/ADT/SetVector.h" +#include "llvm/ADT/StringExtras.h" #include #include using namespace llvm; // Provide a command-line option to aggregate function arguments into a struct -// for functions produced by the code extrator. This is useful when converting +// for functions produced by the code extractor. This is useful when converting // extracted functions to pthread-based code, as only one argument (void*) can // be passed in to pthread_create(). static cl::opt @@ -41,15 +45,15 @@ AggregateArgsOpt("aggregate-extracted-args", cl::Hidden, namespace { class CodeExtractor { - typedef std::vector Values; - std::set BlocksToExtract; - DominatorSet *DS; + typedef SetVector Values; + SetVector BlocksToExtract; + DominatorTree* DT; bool AggregateArgs; unsigned NumExitBlocks; const Type *RetTy; public: - CodeExtractor(DominatorSet *ds = 0, bool AggArgs = false) - : DS(ds), AggregateArgs(AggregateArgsOpt), NumExitBlocks(~0U) {} + CodeExtractor(DominatorTree* dt = 0, bool AggArgs = false) + : DT(dt), AggregateArgs(AggArgs||AggregateArgsOpt), NumExitBlocks(~0U) {} Function *ExtractCodeRegion(const std::vector &code); @@ -64,7 +68,7 @@ namespace { return true; return false; } - + /// definedInCaller - Return true if the specified value is defined in the /// function being code extracted, but not in the region being extracted. /// These values must be passed in as live-ins to the function. @@ -100,10 +104,10 @@ namespace { /// region, we need to split the entry block of the region so that the PHI node /// is easier to deal with. void CodeExtractor::severSplitPHINodes(BasicBlock *&Header) { - bool HasPredsFromRegion = false; + unsigned NumPredsFromRegion = 0; unsigned NumPredsOutsideRegion = 0; - if (Header != &Header->getParent()->front()) { + if (Header != &Header->getParent()->getEntryBlock()) { PHINode *PN = dyn_cast(Header->begin()); if (!PN) return; // No PHI nodes. @@ -112,7 +116,7 @@ void CodeExtractor::severSplitPHINodes(BasicBlock *&Header) { // header block into two. for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) if (BlocksToExtract.count(PN->getIncomingBlock(i))) - HasPredsFromRegion = true; + ++NumPredsFromRegion; else ++NumPredsOutsideRegion; @@ -125,36 +129,25 @@ void CodeExtractor::severSplitPHINodes(BasicBlock *&Header) { // containing PHI nodes merging values from outside of the region, and a // second that contains all of the code for the block and merges back any // incoming values from inside of the region. - BasicBlock::iterator AfterPHIs = Header->begin(); - while (isa(AfterPHIs)) ++AfterPHIs; + BasicBlock::iterator AfterPHIs = Header->getFirstNonPHI(); BasicBlock *NewBB = Header->splitBasicBlock(AfterPHIs, Header->getName()+".ce"); // We only want to code extract the second block now, and it becomes the new // header of the region. BasicBlock *OldPred = Header; - BlocksToExtract.erase(OldPred); + BlocksToExtract.remove(OldPred); BlocksToExtract.insert(NewBB); Header = NewBB; // Okay, update dominator sets. The blocks that dominate the new one are the // blocks that dominate TIBB plus the new block itself. - if (DS) { - DominatorSet::DomSetType DomSet = DS->getDominators(OldPred); - DomSet.insert(NewBB); // A block always dominates itself. - DS->addBasicBlock(NewBB, DomSet); - - // Additionally, NewBB dominates all blocks in the function that are - // dominated by OldPred. - Function *F = Header->getParent(); - for (Function::iterator I = F->begin(), E = F->end(); I != E; ++I) - if (DS->properlyDominates(OldPred, I)) - DS->addDominator(I, NewBB); - } + if (DT) + DT->splitBlock(NewBB); // Okay, now we need to adjust the PHI nodes and any branches from within the // region to go to the new header block instead of the old header block. - if (HasPredsFromRegion) { + if (NumPredsFromRegion) { PHINode *PN = cast(OldPred->begin()); // Loop over all of the predecessors of OldPred that are in the region, // changing them to branch to NewBB instead. @@ -164,14 +157,14 @@ void CodeExtractor::severSplitPHINodes(BasicBlock *&Header) { TI->replaceUsesOfWith(OldPred, NewBB); } - // Okay, everthing within the region is now branching to the right block, we + // Okay, everything within the region is now branching to the right block, we // just have to update the PHI nodes now, inserting PHI nodes into NewBB. - for (AfterPHIs = OldPred->begin(); - PHINode *PN = dyn_cast(AfterPHIs); ++AfterPHIs) { + for (AfterPHIs = OldPred->begin(); isa(AfterPHIs); ++AfterPHIs) { + PHINode *PN = cast(AfterPHIs); // Create a new PHI node in the new region, which has an incoming value // from OldPred of PN. - PHINode *NewPN = new PHINode(PN->getType(), PN->getName()+".ce", - NewBB->begin()); + PHINode *NewPN = PHINode::Create(PN->getType(), 1 + NumPredsFromRegion, + PN->getName()+".ce", NewBB->begin()); NewPN->addIncoming(PN, OldPred); // Loop over all of the incoming value in PN, moving them to NewPN if they @@ -188,17 +181,33 @@ void CodeExtractor::severSplitPHINodes(BasicBlock *&Header) { } void CodeExtractor::splitReturnBlocks() { - for (std::set::iterator I = BlocksToExtract.begin(), + for (SetVector::iterator I = BlocksToExtract.begin(), E = BlocksToExtract.end(); I != E; ++I) - if (ReturnInst *RI = dyn_cast((*I)->getTerminator())) - (*I)->splitBasicBlock(RI, (*I)->getName()+".ret"); + if (ReturnInst *RI = dyn_cast((*I)->getTerminator())) { + BasicBlock *New = (*I)->splitBasicBlock(RI, (*I)->getName()+".ret"); + if (DT) { + // Old dominates New. New node dominates all other nodes dominated + // by Old. + DomTreeNode *OldNode = DT->getNode(*I); + SmallVector Children; + for (DomTreeNode::iterator DI = OldNode->begin(), DE = OldNode->end(); + DI != DE; ++DI) + Children.push_back(*DI); + + DomTreeNode *NewNode = DT->addNewBlock(New, *I); + + for (SmallVector::iterator I = Children.begin(), + E = Children.end(); I != E; ++I) + DT->changeImmediateDominator(*I, NewNode); + } + } } // findInputsOutputs - Find inputs to, outputs from the code region. // void CodeExtractor::findInputsOutputs(Values &inputs, Values &outputs) { std::set ExitBlocks; - for (std::set::const_iterator ci = BlocksToExtract.begin(), + for (SetVector::const_iterator ci = BlocksToExtract.begin(), ce = BlocksToExtract.end(); ci != ce; ++ci) { BasicBlock *BB = *ci; @@ -207,13 +216,13 @@ void CodeExtractor::findInputsOutputs(Values &inputs, Values &outputs) { // instruction is used outside the region, it's an output. for (User::op_iterator O = I->op_begin(), E = I->op_end(); O != E; ++O) if (definedInCaller(*O)) - inputs.push_back(*O); - + inputs.insert(*O); + // Consider uses of this instruction (outputs). for (Value::use_iterator UI = I->use_begin(), E = I->use_end(); UI != E; ++UI) if (!definedInRegion(*UI)) { - outputs.push_back(I); + outputs.insert(I); break; } } // for: insts @@ -226,12 +235,6 @@ void CodeExtractor::findInputsOutputs(Values &inputs, Values &outputs) { } // for: basic blocks NumExitBlocks = ExitBlocks.size(); - - // Eliminate duplicates. - std::sort(inputs.begin(), inputs.end()); - inputs.erase(std::unique(inputs.begin(), inputs.end()), inputs.end()); - std::sort(outputs.begin(), outputs.end()); - outputs.erase(std::unique(outputs.begin(), outputs.end()), outputs.end()); } /// constructFunction - make a function based on inputs and outputs, as follows: @@ -244,71 +247,79 @@ Function *CodeExtractor::constructFunction(const Values &inputs, BasicBlock *newHeader, Function *oldFunction, Module *M) { - DEBUG(std::cerr << "inputs: " << inputs.size() << "\n"); - DEBUG(std::cerr << "outputs: " << outputs.size() << "\n"); + DEBUG(dbgs() << "inputs: " << inputs.size() << "\n"); + DEBUG(dbgs() << "outputs: " << outputs.size() << "\n"); // This function returns unsigned, outputs will go back by reference. switch (NumExitBlocks) { case 0: - case 1: RetTy = Type::VoidTy; break; - case 2: RetTy = Type::BoolTy; break; - default: RetTy = Type::UShortTy; break; + case 1: RetTy = Type::getVoidTy(header->getContext()); break; + case 2: RetTy = Type::getInt1Ty(header->getContext()); break; + default: RetTy = Type::getInt16Ty(header->getContext()); break; } - std::vector paramTy; + std::vector paramTy; // Add the types of the input values to the function's argument list for (Values::const_iterator i = inputs.begin(), e = inputs.end(); i != e; ++i) { const Value *value = *i; - DEBUG(std::cerr << "value used in func: " << *value << "\n"); + DEBUG(dbgs() << "value used in func: " << *value << "\n"); paramTy.push_back(value->getType()); } // Add the types of the output values to the function's argument list. for (Values::const_iterator I = outputs.begin(), E = outputs.end(); I != E; ++I) { - DEBUG(std::cerr << "instr used in func: " << **I << "\n"); + DEBUG(dbgs() << "instr used in func: " << **I << "\n"); if (AggregateArgs) paramTy.push_back((*I)->getType()); else - paramTy.push_back(PointerType::get((*I)->getType())); + paramTy.push_back(PointerType::getUnqual((*I)->getType())); } - DEBUG(std::cerr << "Function type: " << *RetTy << " f("); - DEBUG(for (std::vector::iterator i = paramTy.begin(), - e = paramTy.end(); i != e; ++i) std::cerr << **i << ", "); - DEBUG(std::cerr << ")\n"); + DEBUG(dbgs() << "Function type: " << *RetTy << " f("); + for (std::vector::iterator i = paramTy.begin(), + e = paramTy.end(); i != e; ++i) + DEBUG(dbgs() << **i << ", "); + DEBUG(dbgs() << ")\n"); if (AggregateArgs && (inputs.size() + outputs.size() > 0)) { - PointerType *StructPtr = PointerType::get(StructType::get(paramTy)); + PointerType *StructPtr = + PointerType::getUnqual(StructType::get(M->getContext(), paramTy)); paramTy.clear(); paramTy.push_back(StructPtr); } - const FunctionType *funcType = FunctionType::get(RetTy, paramTy, false); + const FunctionType *funcType = + FunctionType::get(RetTy, paramTy, false); // Create the new function - Function *newFunction = new Function(funcType, - GlobalValue::InternalLinkage, - oldFunction->getName() + "_" + - header->getName(), M); + Function *newFunction = Function::Create(funcType, + GlobalValue::InternalLinkage, + oldFunction->getName() + "_" + + header->getName(), M); + // If the old function is no-throw, so is the new one. + if (oldFunction->doesNotThrow()) + newFunction->setDoesNotThrow(true); + newFunction->getBasicBlockList().push_back(newRootNode); // Create an iterator to name all of the arguments we inserted. - Function::aiterator AI = newFunction->abegin(); + Function::arg_iterator AI = newFunction->arg_begin(); // Rewrite all users of the inputs in the extracted region to use the // arguments (or appropriate addressing into struct) instead. for (unsigned i = 0, e = inputs.size(); i != e; ++i) { Value *RewriteVal; if (AggregateArgs) { - std::vector Indices; - Indices.push_back(Constant::getNullValue(Type::UIntTy)); - Indices.push_back(ConstantUInt::get(Type::UIntTy, i)); - std::string GEPname = "gep_" + inputs[i]->getName(); + Value *Idx[2]; + Idx[0] = Constant::getNullValue(Type::getInt32Ty(header->getContext())); + Idx[1] = ConstantInt::get(Type::getInt32Ty(header->getContext()), i); TerminatorInst *TI = newFunction->begin()->getTerminator(); - GetElementPtrInst *GEP = new GetElementPtrInst(AI, Indices, GEPname, TI); - RewriteVal = new LoadInst(GEP, "load" + GEPname, TI); + GetElementPtrInst *GEP = + GetElementPtrInst::Create(AI, Idx, Idx+2, + "gep_" + inputs[i]->getName(), TI); + RewriteVal = new LoadInst(GEP, "loadgep_" + inputs[i]->getName(), TI); } else RewriteVal = AI++; @@ -322,11 +333,11 @@ Function *CodeExtractor::constructFunction(const Values &inputs, // Set names for input and output arguments. if (!AggregateArgs) { - AI = newFunction->abegin(); + AI = newFunction->arg_begin(); for (unsigned i = 0, e = inputs.size(); i != e; ++i, ++AI) AI->setName(inputs[i]->getName()); for (unsigned i = 0, e = outputs.size(); i != e; ++i, ++AI) - AI->setName(outputs[i]->getName()+".out"); + AI->setName(outputs[i]->getName()+".out"); } // Rewrite branches to basic blocks outside of the loop to new dummy blocks @@ -344,6 +355,20 @@ Function *CodeExtractor::constructFunction(const Values &inputs, return newFunction; } +/// FindPhiPredForUseInBlock - Given a value and a basic block, find a PHI +/// that uses the value within the basic block, and return the predecessor +/// block associated with that use, or return 0 if none is found. +static BasicBlock* FindPhiPredForUseInBlock(Value* Used, BasicBlock* BB) { + for (Value::use_iterator UI = Used->use_begin(), + UE = Used->use_end(); UI != UE; ++UI) { + PHINode *P = dyn_cast(*UI); + if (P && P->getParent() == BB) + return P->getIncomingBlock(UI); + } + + return 0; +} + /// emitCallAndSwitchStatement - This method sets up the caller side by adding /// the call instruction, splitting any PHI nodes in the header block as /// necessary. @@ -352,7 +377,9 @@ emitCallAndSwitchStatement(Function *newFunction, BasicBlock *codeReplacer, Values &inputs, Values &outputs) { // Emit a call to the new function, passing in: *pointer to struct (if // aggregating parameters), or plan inputs and allocated memory for outputs - std::vector params, StructValues, ReloadOutputs; + std::vector params, StructValues, ReloadOutputs, Reloads; + + LLVMContext &Context = newFunction->getContext(); // Add inputs as params, or to be filled into the struct for (Values::iterator i = inputs.begin(), e = inputs.end(); i != e; ++i) @@ -382,31 +409,31 @@ emitCallAndSwitchStatement(Function *newFunction, BasicBlock *codeReplacer, ArgTypes.push_back((*v)->getType()); // Allocate a struct at the beginning of this function - Type *StructArgTy = StructType::get(ArgTypes); - Struct = - new AllocaInst(StructArgTy, 0, "structArg", + Type *StructArgTy = StructType::get(newFunction->getContext(), ArgTypes); + Struct = + new AllocaInst(StructArgTy, 0, "structArg", codeReplacer->getParent()->begin()->begin()); params.push_back(Struct); for (unsigned i = 0, e = inputs.size(); i != e; ++i) { - std::vector Indices; - Indices.push_back(Constant::getNullValue(Type::UIntTy)); - Indices.push_back(ConstantUInt::get(Type::UIntTy, i)); + Value *Idx[2]; + Idx[0] = Constant::getNullValue(Type::getInt32Ty(Context)); + Idx[1] = ConstantInt::get(Type::getInt32Ty(Context), i); GetElementPtrInst *GEP = - new GetElementPtrInst(Struct, Indices, - "gep_" + StructValues[i]->getName()); + GetElementPtrInst::Create(Struct, Idx, Idx + 2, + "gep_" + StructValues[i]->getName()); codeReplacer->getInstList().push_back(GEP); StoreInst *SI = new StoreInst(StructValues[i], GEP); codeReplacer->getInstList().push_back(SI); } - } + } // Emit the call to the function - CallInst *call = new CallInst(newFunction, params, - NumExitBlocks > 1 ? "targetBlock": ""); + CallInst *call = CallInst::Create(newFunction, params.begin(), params.end(), + NumExitBlocks > 1 ? "targetBlock" : ""); codeReplacer->getInstList().push_back(call); - Function::aiterator OutputArgBegin = newFunction->abegin(); + Function::arg_iterator OutputArgBegin = newFunction->arg_begin(); unsigned FirstOut = inputs.size(); if (!AggregateArgs) std::advance(OutputArgBegin, inputs.size()); @@ -415,18 +442,19 @@ emitCallAndSwitchStatement(Function *newFunction, BasicBlock *codeReplacer, for (unsigned i = 0, e = outputs.size(); i != e; ++i) { Value *Output = 0; if (AggregateArgs) { - std::vector Indices; - Indices.push_back(Constant::getNullValue(Type::UIntTy)); - Indices.push_back(ConstantUInt::get(Type::UIntTy, FirstOut + i)); - GetElementPtrInst *GEP - = new GetElementPtrInst(Struct, Indices, - "gep_reload_" + outputs[i]->getName()); + Value *Idx[2]; + Idx[0] = Constant::getNullValue(Type::getInt32Ty(Context)); + Idx[1] = ConstantInt::get(Type::getInt32Ty(Context), FirstOut + i); + GetElementPtrInst *GEP + = GetElementPtrInst::Create(Struct, Idx, Idx + 2, + "gep_reload_" + outputs[i]->getName()); codeReplacer->getInstList().push_back(GEP); Output = GEP; } else { Output = ReloadOutputs[i]; } LoadInst *load = new LoadInst(Output, outputs[i]->getName()+".reload"); + Reloads.push_back(load); codeReplacer->getInstList().push_back(load); std::vector Users(outputs[i]->use_begin(), outputs[i]->use_end()); for (unsigned u = 0, e = Users.size(); u != e; ++u) { @@ -438,8 +466,8 @@ emitCallAndSwitchStatement(Function *newFunction, BasicBlock *codeReplacer, // Now we can emit a switch statement using the call as a value. SwitchInst *TheSwitch = - new SwitchInst(ConstantUInt::getNullValue(Type::UShortTy), - codeReplacer, codeReplacer); + SwitchInst::Create(Constant::getNullValue(Type::getInt16Ty(Context)), + codeReplacer, 0, codeReplacer); // Since there may be multiple exits from the original region, make the new // function return an unsigned, switch on that number. This loop iterates @@ -449,7 +477,7 @@ emitCallAndSwitchStatement(Function *newFunction, BasicBlock *codeReplacer, std::map ExitBlockMap; unsigned switchVal = 0; - for (std::set::const_iterator i = BlocksToExtract.begin(), + for (SetVector::const_iterator i = BlocksToExtract.begin(), e = BlocksToExtract.end(); i != e; ++i) { TerminatorInst *TI = (*i)->getTerminator(); for (unsigned i = 0, e = TI->getNumSuccessors(); i != e; ++i) @@ -460,8 +488,9 @@ emitCallAndSwitchStatement(Function *newFunction, BasicBlock *codeReplacer, if (!NewTarget) { // If we don't already have an exit stub for this non-extracted // destination, create one now! - NewTarget = new BasicBlock(OldTarget->getName() + ".exitStub", - newFunction); + NewTarget = BasicBlock::Create(Context, + OldTarget->getName() + ".exitStub", + newFunction); unsigned SuccNum = switchVal++; Value *brVal = 0; @@ -469,39 +498,77 @@ emitCallAndSwitchStatement(Function *newFunction, BasicBlock *codeReplacer, case 0: case 1: break; // No value needed. case 2: // Conditional branch, return a bool - brVal = SuccNum ? ConstantBool::False : ConstantBool::True; + brVal = ConstantInt::get(Type::getInt1Ty(Context), !SuccNum); break; default: - brVal = ConstantUInt::get(Type::UShortTy, SuccNum); + brVal = ConstantInt::get(Type::getInt16Ty(Context), SuccNum); break; } - ReturnInst *NTRet = new ReturnInst(brVal, NewTarget); + ReturnInst *NTRet = ReturnInst::Create(Context, brVal, NewTarget); // Update the switch instruction. - TheSwitch->addCase(ConstantUInt::get(Type::UShortTy, SuccNum), + TheSwitch->addCase(ConstantInt::get(Type::getInt16Ty(Context), + SuccNum), OldTarget); // Restore values just before we exit - Function::aiterator OAI = OutputArgBegin; + Function::arg_iterator OAI = OutputArgBegin; for (unsigned out = 0, e = outputs.size(); out != e; ++out) { // For an invoke, the normal destination is the only one that is // dominated by the result of the invocation BasicBlock *DefBlock = cast(outputs[out])->getParent(); - if (InvokeInst *Invoke = dyn_cast(outputs[out])) + + bool DominatesDef = true; + + if (InvokeInst *Invoke = dyn_cast(outputs[out])) { DefBlock = Invoke->getNormalDest(); - if (!DS || DS->dominates(DefBlock, TI->getParent())) + + // Make sure we are looking at the original successor block, not + // at a newly inserted exit block, which won't be in the dominator + // info. + for (std::map::iterator I = + ExitBlockMap.begin(), E = ExitBlockMap.end(); I != E; ++I) + if (DefBlock == I->second) { + DefBlock = I->first; + break; + } + + // In the extract block case, if the block we are extracting ends + // with an invoke instruction, make sure that we don't emit a + // store of the invoke value for the unwind block. + if (!DT && DefBlock != OldTarget) + DominatesDef = false; + } + + if (DT) { + DominatesDef = DT->dominates(DefBlock, OldTarget); + + // If the output value is used by a phi in the target block, + // then we need to test for dominance of the phi's predecessor + // instead. Unfortunately, this a little complicated since we + // have already rewritten uses of the value to uses of the reload. + BasicBlock* pred = FindPhiPredForUseInBlock(Reloads[out], + OldTarget); + if (pred && DT && DT->dominates(DefBlock, pred)) + DominatesDef = true; + } + + if (DominatesDef) { if (AggregateArgs) { - std::vector Indices; - Indices.push_back(Constant::getNullValue(Type::UIntTy)); - Indices.push_back(ConstantUInt::get(Type::UIntTy,FirstOut+out)); + Value *Idx[2]; + Idx[0] = Constant::getNullValue(Type::getInt32Ty(Context)); + Idx[1] = ConstantInt::get(Type::getInt32Ty(Context), + FirstOut+out); GetElementPtrInst *GEP = - new GetElementPtrInst(OAI, Indices, - "gep_" + outputs[out]->getName(), - NTRet); + GetElementPtrInst::Create(OAI, Idx, Idx + 2, + "gep_" + outputs[out]->getName(), + NTRet); new StoreInst(outputs[out], GEP, NTRet); - } else + } else { new StoreInst(outputs[out], OAI, NTRet); + } + } // Advance output iterator even if we don't emit a store if (!AggregateArgs) ++OAI; } @@ -513,34 +580,38 @@ emitCallAndSwitchStatement(Function *newFunction, BasicBlock *codeReplacer, } // Now that we've done the deed, simplify the switch instruction. + const Type *OldFnRetTy = TheSwitch->getParent()->getParent()->getReturnType(); switch (NumExitBlocks) { case 0: - // There is only 1 successor (the block containing the switch itself), which + // There are no successors (the block containing the switch itself), which // means that previously this was the last part of the function, and hence // this should be rewritten as a `ret' - + // Check if the function should return a value - if (TheSwitch->getParent()->getParent()->getReturnType() != Type::VoidTy && - TheSwitch->getParent()->getParent()->getReturnType() == - TheSwitch->getCondition()->getType()) + if (OldFnRetTy->isVoidTy()) { + ReturnInst::Create(Context, 0, TheSwitch); // Return void + } else if (OldFnRetTy == TheSwitch->getCondition()->getType()) { // return what we have - new ReturnInst(TheSwitch->getCondition(), TheSwitch); - else - // just return - new ReturnInst(0, TheSwitch); + ReturnInst::Create(Context, TheSwitch->getCondition(), TheSwitch); + } else { + // Otherwise we must have code extracted an unwind or something, just + // return whatever we want. + ReturnInst::Create(Context, + Constant::getNullValue(OldFnRetTy), TheSwitch); + } - TheSwitch->getParent()->getInstList().erase(TheSwitch); + TheSwitch->eraseFromParent(); break; case 1: // Only a single destination, change the switch into an unconditional // branch. - new BranchInst(TheSwitch->getSuccessor(1), TheSwitch); - TheSwitch->getParent()->getInstList().erase(TheSwitch); + BranchInst::Create(TheSwitch->getSuccessor(1), TheSwitch); + TheSwitch->eraseFromParent(); break; case 2: - new BranchInst(TheSwitch->getSuccessor(1), TheSwitch->getSuccessor(2), - call, TheSwitch); - TheSwitch->getParent()->getInstList().erase(TheSwitch); + BranchInst::Create(TheSwitch->getSuccessor(1), TheSwitch->getSuccessor(2), + call, TheSwitch); + TheSwitch->eraseFromParent(); break; default: // Otherwise, make the default destination of the switch instruction be one @@ -557,7 +628,7 @@ void CodeExtractor::moveCodeToFunction(Function *newFunction) { Function::BasicBlockListType &oldBlocks = oldFunc->getBasicBlockList(); Function::BasicBlockListType &newBlocks = newFunction->getBasicBlockList(); - for (std::set::const_iterator i = BlocksToExtract.begin(), + for (SetVector::const_iterator i = BlocksToExtract.begin(), e = BlocksToExtract.end(); i != e; ++i) { // Delete the basic block from the old function, and the list of blocks oldBlocks.remove(*i); @@ -574,17 +645,17 @@ void CodeExtractor::moveCodeToFunction(Function *newFunction) { /// /// find inputs and outputs for the region /// -/// for inputs: add to function as args, map input instr* to arg# -/// for outputs: add allocas for scalars, +/// for inputs: add to function as args, map input instr* to arg# +/// for outputs: add allocas for scalars, /// add to func as args, map output instr* to arg# /// /// rewrite func to use argument #s instead of instr* /// -/// for each scalar output in the function: at every exit, store intermediate +/// for each scalar output in the function: at every exit, store intermediate /// computed result back into memory. /// -Function *CodeExtractor::ExtractCodeRegion(const std::vector &code) -{ +Function *CodeExtractor:: +ExtractCodeRegion(const std::vector &code) { if (!isEligible(code)) return 0; @@ -608,7 +679,7 @@ Function *CodeExtractor::ExtractCodeRegion(const std::vector &code) assert(BlocksToExtract.count(*PI) && "No blocks in this region may have entries from outside the region" " except for the first block!"); - + // If we have to split PHI nodes or the entry block, do so now. severSplitPHINodes(header); @@ -619,19 +690,22 @@ Function *CodeExtractor::ExtractCodeRegion(const std::vector &code) Function *oldFunction = header->getParent(); // This takes place of the original loop - BasicBlock *codeReplacer = new BasicBlock("codeRepl", oldFunction); + BasicBlock *codeReplacer = BasicBlock::Create(header->getContext(), + "codeRepl", oldFunction, + header); // The new function needs a root node because other nodes can branch to the // head of the region, but the entry node of a function cannot have preds. - BasicBlock *newFuncRoot = new BasicBlock("newFuncRoot"); - newFuncRoot->getInstList().push_back(new BranchInst(header)); + BasicBlock *newFuncRoot = BasicBlock::Create(header->getContext(), + "newFuncRoot"); + newFuncRoot->getInstList().push_back(BranchInst::Create(header)); // Find inputs to, outputs from the code region. findInputsOutputs(inputs, outputs); // Construct new function based on inputs/outputs & add allocas for all defs. Function *newFunction = constructFunction(inputs, outputs, header, - newFuncRoot, + newFuncRoot, codeReplacer, oldFunction, oldFunction->getParent()); @@ -641,11 +715,12 @@ Function *CodeExtractor::ExtractCodeRegion(const std::vector &code) // Loop over all of the PHI nodes in the header block, and change any // references to the old incoming edge to be the new incoming edge. - for (BasicBlock::iterator I = header->begin(); - PHINode *PN = dyn_cast(I); ++I) + for (BasicBlock::iterator I = header->begin(); isa(I); ++I) { + PHINode *PN = cast(I); for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) if (!BlocksToExtract.count(PN->getIncomingBlock(i))) PN->setIncomingBlock(i, newFuncRoot); + } // Look at all successors of the codeReplacer block. If any of these blocks // had PHI nodes in them, we need to update the "from" block to be the code @@ -653,19 +728,30 @@ Function *CodeExtractor::ExtractCodeRegion(const std::vector &code) std::vector Succs(succ_begin(codeReplacer), succ_end(codeReplacer)); for (unsigned i = 0, e = Succs.size(); i != e; ++i) - for (BasicBlock::iterator I = Succs[i]->begin(); - PHINode *PN = dyn_cast(I); ++I) + for (BasicBlock::iterator I = Succs[i]->begin(); isa(I); ++I) { + PHINode *PN = cast(I); + std::set ProcessedPreds; for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) - if (BlocksToExtract.count(PN->getIncomingBlock(i))) - PN->setIncomingBlock(i, codeReplacer); - - //std::cerr << "NEW FUNCTION: " << *newFunction; + if (BlocksToExtract.count(PN->getIncomingBlock(i))) { + if (ProcessedPreds.insert(PN->getIncomingBlock(i)).second) + PN->setIncomingBlock(i, codeReplacer); + else { + // There were multiple entries in the PHI for this block, now there + // is only one, so remove the duplicated entries. + PN->removeIncomingValue(i, false); + --i; --e; + } + } + } + + //cerr << "NEW FUNCTION: " << *newFunction; // verifyFunction(*newFunction); - // std::cerr << "OLD FUNCTION: " << *oldFunction; + // cerr << "OLD FUNCTION: " << *oldFunction; // verifyFunction(*oldFunction); - DEBUG(if (verifyFunction(*newFunction)) abort()); + DEBUG(if (verifyFunction(*newFunction)) + report_fatal_error("verifyFunction failed!")); return newFunction; } @@ -688,16 +774,16 @@ bool CodeExtractor::isEligible(const std::vector &code) { /// ExtractCodeRegion - slurp a sequence of basic blocks into a brand new /// function /// -Function* llvm::ExtractCodeRegion(DominatorSet &DS, +Function* llvm::ExtractCodeRegion(DominatorTree &DT, const std::vector &code, bool AggregateArgs) { - return CodeExtractor(&DS, AggregateArgs).ExtractCodeRegion(code); + return CodeExtractor(&DT, AggregateArgs).ExtractCodeRegion(code); } /// ExtractBasicBlock - slurp a natural loop into a brand new function /// -Function* llvm::ExtractLoop(DominatorSet &DS, Loop *L, bool AggregateArgs) { - return CodeExtractor(&DS, AggregateArgs).ExtractCodeRegion(L->getBlocks()); +Function* llvm::ExtractLoop(DominatorTree &DT, Loop *L, bool AggregateArgs) { + return CodeExtractor(&DT, AggregateArgs).ExtractCodeRegion(L->getBlocks()); } /// ExtractBasicBlock - slurp a basic block into a brand new function @@ -705,5 +791,5 @@ Function* llvm::ExtractLoop(DominatorSet &DS, Loop *L, bool AggregateArgs) { Function* llvm::ExtractBasicBlock(BasicBlock *BB, bool AggregateArgs) { std::vector Blocks; Blocks.push_back(BB); - return CodeExtractor(0, AggregateArgs).ExtractCodeRegion(Blocks); + return CodeExtractor(0, AggregateArgs).ExtractCodeRegion(Blocks); }