//===- 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
#include "llvm/Analysis/Verifier.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/Compiler.h"
#include "llvm/Support/Debug.h"
#include "llvm/ADT/StringExtras.h"
#include <algorithm>
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<bool>
cl::desc("Aggregate arguments to code-extracted functions"));
namespace {
- class CodeExtractor {
+ class VISIBILITY_HIDDEN CodeExtractor {
typedef std::vector<Value*> Values;
std::set<BasicBlock*> BlocksToExtract;
- DominatorSet *DS;
+ DominatorTree* DT;
bool AggregateArgs;
unsigned NumExitBlocks;
const Type *RetTy;
public:
- CodeExtractor(DominatorSet *ds = 0, bool AggArgs = false)
- : DS(ds), AggregateArgs(AggArgs||AggregateArgsOpt), NumExitBlocks(~0U) {}
+ CodeExtractor(DominatorTree* dt = 0, bool AggArgs = false)
+ : DT(dt), AggregateArgs(AggArgs||AggregateArgsOpt), NumExitBlocks(~0U) {}
Function *ExtractCodeRegion(const std::vector<BasicBlock*> &code);
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.
bool HasPredsFromRegion = false;
unsigned NumPredsOutsideRegion = 0;
- if (Header != &Header->getParent()->front()) {
+ if (Header != &Header->getParent()->getEntryBlock()) {
PHINode *PN = dyn_cast<PHINode>(Header->begin());
if (!PN) return; // No PHI nodes.
// 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<PHINode>(AfterPHIs)) ++AfterPHIs;
+ BasicBlock::iterator AfterPHIs = Header->getFirstNonPHI();
BasicBlock *NewBB = Header->splitBasicBlock(AfterPHIs,
Header->getName()+".ce");
// 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.
PHINode *PN = cast<PHINode>(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(), PN->getName()+".ce",
+ NewBB->begin());
NewPN->addIncoming(PN, OldPred);
// Loop over all of the incoming value in PN, moving them to NewPN if they
//
void CodeExtractor::findInputsOutputs(Values &inputs, Values &outputs) {
std::set<BasicBlock*> ExitBlocks;
- for (std::set<BasicBlock*>::const_iterator ci = BlocksToExtract.begin(),
+ for (std::set<BasicBlock*>::const_iterator ci = BlocksToExtract.begin(),
ce = BlocksToExtract.end(); ci != ce; ++ci) {
BasicBlock *BB = *ci;
for (User::op_iterator O = I->op_begin(), E = I->op_end(); O != E; ++O)
if (definedInCaller(*O))
inputs.push_back(*O);
-
+
// Consider uses of this instruction (outputs).
for (Value::use_iterator UI = I->use_begin(), E = I->use_end();
UI != E; ++UI)
BasicBlock *newHeader,
Function *oldFunction,
Module *M) {
- DEBUG(std::cerr << "inputs: " << inputs.size() << "\n");
- DEBUG(std::cerr << "outputs: " << outputs.size() << "\n");
+ DOUT << "inputs: " << inputs.size() << "\n";
+ DOUT << "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 2: RetTy = Type::Int1Ty; break;
+ default: RetTy = Type::Int16Ty; break;
}
std::vector<const Type*> paramTy;
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");
+ DOUT << "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");
+ DOUT << "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<const Type*>::iterator i = paramTy.begin(),
- e = paramTy.end(); i != e; ++i) std::cerr << **i << ", ");
- DEBUG(std::cerr << ")\n");
+ DOUT << "Function type: " << *RetTy << " f(";
+ for (std::vector<const Type*>::iterator i = paramTy.begin(),
+ e = paramTy.end(); i != e; ++i)
+ DOUT << **i << ", ";
+ DOUT << ")\n";
if (AggregateArgs && (inputs.size() + outputs.size() > 0)) {
- PointerType *StructPtr = PointerType::get(StructType::get(paramTy));
+ PointerType *StructPtr = PointerType::getUnqual(StructType::get(paramTy));
paramTy.clear();
paramTy.push_back(StructPtr);
}
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.
for (unsigned i = 0, e = inputs.size(); i != e; ++i) {
Value *RewriteVal;
if (AggregateArgs) {
- std::vector<Value*> Indices;
- Indices.push_back(Constant::getNullValue(Type::UIntTy));
- Indices.push_back(ConstantUInt::get(Type::UIntTy, i));
+ Value *Idx[2];
+ Idx[0] = Constant::getNullValue(Type::Int32Ty);
+ Idx[1] = ConstantInt::get(Type::Int32Ty, i);
std::string GEPname = "gep_" + inputs[i]->getName();
TerminatorInst *TI = newFunction->begin()->getTerminator();
- GetElementPtrInst *GEP = new GetElementPtrInst(AI, Indices, GEPname, TI);
+ GetElementPtrInst *GEP = GetElementPtrInst::Create(AI, Idx, Idx+2,
+ GEPname, TI);
RewriteVal = new LoadInst(GEP, "load" + GEPname, TI);
} else
RewriteVal = AI++;
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
// Allocate a struct at the beginning of this function
Type *StructArgTy = StructType::get(ArgTypes);
- Struct =
- new AllocaInst(StructArgTy, 0, "structArg",
+ 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<Value*> Indices;
- Indices.push_back(Constant::getNullValue(Type::UIntTy));
- Indices.push_back(ConstantUInt::get(Type::UIntTy, i));
+ Value *Idx[2];
+ Idx[0] = Constant::getNullValue(Type::Int32Ty);
+ Idx[1] = ConstantInt::get(Type::Int32Ty, 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::arg_iterator OutputArgBegin = newFunction->arg_begin();
for (unsigned i = 0, e = outputs.size(); i != e; ++i) {
Value *Output = 0;
if (AggregateArgs) {
- std::vector<Value*> 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::Int32Ty);
+ Idx[1] = ConstantInt::get(Type::Int32Ty, FirstOut + i);
+ GetElementPtrInst *GEP
+ = GetElementPtrInst::Create(Struct, Idx, Idx + 2,
+ "gep_reload_" + outputs[i]->getName());
codeReplacer->getInstList().push_back(GEP);
Output = GEP;
} else {
// Now we can emit a switch statement using the call as a value.
SwitchInst *TheSwitch =
- new SwitchInst(ConstantUInt::getNullValue(Type::UShortTy),
- codeReplacer, 0, codeReplacer);
+ SwitchInst::Create(ConstantInt::getNullValue(Type::Int16Ty),
+ 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
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(OldTarget->getName() + ".exitStub",
+ newFunction);
unsigned SuccNum = switchVal++;
Value *brVal = 0;
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::Int1Ty, !SuccNum);
break;
default:
- brVal = ConstantUInt::get(Type::UShortTy, SuccNum);
+ brVal = ConstantInt::get(Type::Int16Ty, SuccNum);
break;
}
- ReturnInst *NTRet = new ReturnInst(brVal, NewTarget);
+ ReturnInst *NTRet = ReturnInst::Create(brVal, NewTarget);
// Update the switch instruction.
- TheSwitch->addCase(ConstantUInt::get(Type::UShortTy, SuccNum),
+ TheSwitch->addCase(ConstantInt::get(Type::Int16Ty, SuccNum),
OldTarget);
// Restore values just before we exit
// 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 (!DS && DefBlock != OldTarget)
+ if (!DT && DefBlock != OldTarget)
DominatesDef = false;
}
- if (DS)
- DominatesDef = DS->dominates(DefBlock, OldTarget);
+ if (DT)
+ DominatesDef = DT->dominates(DefBlock, OldTarget);
if (DominatesDef) {
if (AggregateArgs) {
- std::vector<Value*> 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::Int32Ty);
+ Idx[1] = ConstantInt::get(Type::Int32Ty,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 {
new StoreInst(outputs[out], OAI, NTRet);
// 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 (OldFnRetTy == Type::VoidTy) {
- new ReturnInst(0, TheSwitch); // Return void
+ ReturnInst::Create(0, TheSwitch); // Return void
} else if (OldFnRetTy == TheSwitch->getCondition()->getType()) {
// return what we have
- new ReturnInst(TheSwitch->getCondition(), TheSwitch);
+ ReturnInst::Create(TheSwitch->getCondition(), TheSwitch);
} else {
// Otherwise we must have code extracted an unwind or something, just
// return whatever we want.
- new ReturnInst(Constant::getNullValue(OldFnRetTy), TheSwitch);
+ ReturnInst::Create(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
///
/// 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::
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);
Function *oldFunction = header->getParent();
// This takes place of the original loop
- BasicBlock *codeReplacer = new BasicBlock("codeRepl", oldFunction, header);
+ BasicBlock *codeReplacer = BasicBlock::Create("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("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());
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
// replacer, not the original block in the extracted region.
PHINode *PN = cast<PHINode>(I);
std::set<BasicBlock*> ProcessedPreds;
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
- if (BlocksToExtract.count(PN->getIncomingBlock(i)))
+ if (BlocksToExtract.count(PN->getIncomingBlock(i))) {
if (ProcessedPreds.insert(PN->getIncomingBlock(i)).second)
PN->setIncomingBlock(i, codeReplacer);
else {
PN->removeIncomingValue(i, false);
--i; --e;
}
+ }
}
-
- //std::cerr << "NEW FUNCTION: " << *newFunction;
+
+ //cerr << "NEW FUNCTION: " << *newFunction;
// verifyFunction(*newFunction);
- // std::cerr << "OLD FUNCTION: " << *oldFunction;
+ // cerr << "OLD FUNCTION: " << *oldFunction;
// verifyFunction(*oldFunction);
DEBUG(if (verifyFunction(*newFunction)) abort());
/// 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<BasicBlock*> &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
Function* llvm::ExtractBasicBlock(BasicBlock *BB, bool AggregateArgs) {
std::vector<BasicBlock*> Blocks;
Blocks.push_back(BB);
- return CodeExtractor(0, AggregateArgs).ExtractCodeRegion(Blocks);
+ return CodeExtractor(0, AggregateArgs).ExtractCodeRegion(Blocks);
}