void CallGraph::addToCallGraph(Function *F) {
CallGraphNode *Node = getOrInsertFunction(F);
// If this function has external linkage or has its address taken, anything
// could call it.
if (!F->hasLocalLinkage() || F->hasAddressTaken())
ExternalCallingNode->addCalledFunction(CallSite(), Node);
// If this function is not defined in this translation unit, it could call
// anything.
if (F->isDeclaration() && !F->isIntrinsic())
Node->addCalledFunction(CallSite(), CallsExternalNode.get());
// Look for calls by this function.
for (BasicBlock &BB : *F)
for (Instruction &I : BB) {
if (auto CS = CallSite(&I)) {
const Function *Callee = CS.getCalledFunction();
if (!Callee || !Intrinsic::isLeaf(Callee->getIntrinsicID()))
// Indirect calls of intrinsics are not allowed so no need to check.
// We can be more precise here by using TargetArg returned by
// Intrinsic::isLeaf.
Node->addCalledFunction(CS, CallsExternalNode.get());
else if (!Callee->isIntrinsic())
Node->addCalledFunction(CS, getOrInsertFunction(Callee));
}
}
}
/// UpdateCallGraphAfterInlining - Once we have cloned code over from a callee
/// into the caller, update the specified callgraph to reflect the changes we
/// made. Note that it's possible that not all code was copied over, so only
/// some edges of the callgraph may remain.
static void UpdateCallGraphAfterInlining(CallSite CS,
Function::iterator FirstNewBlock,
ValueToValueMapTy &VMap,
InlineFunctionInfo &IFI) {
CallGraph &CG = *IFI.CG;
const Function *Caller = CS.getInstruction()->getParent()->getParent();
const Function *Callee = CS.getCalledFunction();
CallGraphNode *CalleeNode = CG[Callee];
CallGraphNode *CallerNode = CG[Caller];
// Since we inlined some uninlined call sites in the callee into the caller,
// add edges from the caller to all of the callees of the callee.
CallGraphNode::iterator I = CalleeNode->begin(), E = CalleeNode->end();
// Consider the case where CalleeNode == CallerNode.
CallGraphNode::CalledFunctionsVector CallCache;
if (CalleeNode == CallerNode) {
CallCache.assign(I, E);
I = CallCache.begin();
E = CallCache.end();
}
for (; I != E; ++I) {
const Value *OrigCall = I->first;
ValueToValueMapTy::iterator VMI = VMap.find(OrigCall);
// Only copy the edge if the call was inlined!
if (VMI == VMap.end() || VMI->second == 0)
continue;
// If the call was inlined, but then constant folded, there is no edge to
// add. Check for this case.
Instruction *NewCall = dyn_cast<Instruction>(VMI->second);
if (NewCall == 0) continue;
// Remember that this call site got inlined for the client of
// InlineFunction.
IFI.InlinedCalls.push_back(NewCall);
// It's possible that inlining the callsite will cause it to go from an
// indirect to a direct call by resolving a function pointer. If this
// happens, set the callee of the new call site to a more precise
// destination. This can also happen if the call graph node of the caller
// was just unnecessarily imprecise.
if (I->second->getFunction() == 0)
if (Function *F = CallSite(NewCall).getCalledFunction()) {
// Indirect call site resolved to direct call.
CallerNode->addCalledFunction(CallSite(NewCall), CG[F]);
continue;
}
CallerNode->addCalledFunction(CallSite(NewCall), I->second);
}
// Update the call graph by deleting the edge from Callee to Caller. We must
// do this after the loop above in case Caller and Callee are the same.
CallerNode->removeCallEdgeFor(CS);
}
/// UpdateCallGraphAfterInlining - Once we have cloned code over from a callee
/// into the caller, update the specified callgraph to reflect the changes we
/// made. Note that it's possible that not all code was copied over, so only
/// some edges of the callgraph will be remain.
static void UpdateCallGraphAfterInlining(const Function *Caller,
const Function *Callee,
Function::iterator FirstNewBlock,
DenseMap<const Value*, Value*> &ValueMap,
CallGraph &CG) {
// Update the call graph by deleting the edge from Callee to Caller
CallGraphNode *CalleeNode = CG[Callee];
CallGraphNode *CallerNode = CG[Caller];
CallerNode->removeCallEdgeTo(CalleeNode);
// Since we inlined some uninlined call sites in the callee into the caller,
// add edges from the caller to all of the callees of the callee.
for (CallGraphNode::iterator I = CalleeNode->begin(),
E = CalleeNode->end(); I != E; ++I) {
const Instruction *OrigCall = I->first.getInstruction();
DenseMap<const Value*, Value*>::iterator VMI = ValueMap.find(OrigCall);
// Only copy the edge if the call was inlined!
if (VMI != ValueMap.end() && VMI->second) {
// If the call was inlined, but then constant folded, there is no edge to
// add. Check for this case.
if (Instruction *NewCall = dyn_cast<Instruction>(VMI->second))
CallerNode->addCalledFunction(CallSite::get(NewCall), I->second);
}
}
}
void ManualInliner::run(std::vector<Function *>::iterator fbegin, std::vector<Function *>::iterator fend) {
std::vector<CallGraphNode*> nodes;
//the inliner requires an up to date callgraph, so we add the functions in the SCC
//to the callgraph. If needed, we can do this during function creation to make it faster
for(std::vector<Function *>::iterator fp = fbegin; fp != fend; ++fp) {
Function * F = *fp;
CallGraphNode * n = CG->getOrInsertFunction(F);
for (Function::iterator BB = F->begin(), BBE = F->end(); BB != BBE; ++BB)
for (BasicBlock::iterator II = BB->begin(), IE = BB->end(); II != IE; ++II) {
CallSite CS(cast<Value>(II));
if (CS) {
const Function *Callee = CS.getCalledFunction();
if (Callee && !Callee->isIntrinsic()) {
CallGraphNode * n2 = CG->getOrInsertFunction(Callee);
n->addCalledFunction(CS,n2);
}
}
}
nodes.push_back(n);
}
//create a fake SCC node and manually run the inliner pass on it.
CallGraphSCC SCC(NULL);
SCC.initialize(&nodes[0], &nodes[0]+nodes.size());
SI->runOnSCC(SCC);
//We optimize the function now, which will invalidate the call graph,
//removing called functions makes sure that further inlining passes don't attempt to add invalid callsites as inlining candidates
for(std::vector<Function *>::iterator fp = fbegin; fp != fend; ++fp) {
CG->getOrInsertFunction(*fp)->removeAllCalledFunctions();
}
}
// addToCallGraph - Add a function to the call graph, and link the node to all
// of the functions that it calls.
//
void addToCallGraph(Function *F) {
CallGraphNode *Node = getOrInsertFunction(F);
// If this function has external linkage, anything could call it.
if (!F->hasLocalLinkage()) {
ExternalCallingNode->addCalledFunction(CallSite(), Node);
// Found the entry point?
if (F->getName() == "main") {
if (Root) // Found multiple external mains? Don't pick one.
Root = ExternalCallingNode;
else
Root = Node; // Found a main, keep track of it!
}
}
// Loop over all of the users of the function, looking for non-call uses.
for (Value::use_iterator I = F->use_begin(), E = F->use_end(); I != E; ++I)
if ((!isa<CallInst>(I) && !isa<InvokeInst>(I))
|| !CallSite(cast<Instruction>(I)).isCallee(I)) {
// Not a call, or being used as a parameter rather than as the callee.
ExternalCallingNode->addCalledFunction(CallSite(), Node);
break;
}
// If this function is not defined in this translation unit, it could call
// anything.
if (F->isDeclaration() && !F->isIntrinsic())
Node->addCalledFunction(CallSite(), CallsExternalNode);
// Look for calls by this function.
for (Function::iterator BB = F->begin(), BBE = F->end(); BB != BBE; ++BB)
for (BasicBlock::iterator II = BB->begin(), IE = BB->end();
II != IE; ++II) {
CallSite CS = CallSite::get(II);
if (CS.getInstruction() && !isa<DbgInfoIntrinsic>(II)) {
const Function *Callee = CS.getCalledFunction();
if (Callee)
Node->addCalledFunction(CS, getOrInsertFunction(Callee));
else
Node->addCalledFunction(CS, CallsExternalNode);
}
}
}
void CallGraph::addToCallGraph(Function *F) {
CallGraphNode *Node = getOrInsertFunction(F);
// If this function has external linkage, anything could call it.
if (!F->hasLocalLinkage()) {
ExternalCallingNode->addCalledFunction(CallSite(), Node);
// Found the entry point?
if (F->getName() == "main") {
if (Root) // Found multiple external mains? Don't pick one.
Root = ExternalCallingNode;
else
Root = Node; // Found a main, keep track of it!
}
}
// If this function has its address taken, anything could call it.
if (F->hasAddressTaken())
ExternalCallingNode->addCalledFunction(CallSite(), Node);
// If this function is not defined in this translation unit, it could call
// anything.
if (F->isDeclaration() && !F->isIntrinsic())
Node->addCalledFunction(CallSite(), CallsExternalNode.get());
// Look for calls by this function.
for (Function::iterator BB = F->begin(), BBE = F->end(); BB != BBE; ++BB)
for (BasicBlock::iterator II = BB->begin(), IE = BB->end(); II != IE;
++II) {
CallSite CS(cast<Value>(II));
if (CS) {
const Function *Callee = CS.getCalledFunction();
if (!Callee || !Intrinsic::isLeaf(Callee->getIntrinsicID()))
// Indirect calls of intrinsics are not allowed so no need to check.
// We can be more precise here by using TargetArg returned by
// Intrinsic::isLeaf.
Node->addCalledFunction(CS, CallsExternalNode.get());
else if (!Callee->isIntrinsic())
Node->addCalledFunction(CS, getOrInsertFunction(Callee));
}
}
}
Function * StructuredModuleEditor::cloneFunc(Function * Original) {
if (Original == NULL)
return NULL;
ValueMap<const Value*, WeakVH> VMap;
// Creates a clone of the function we are cloning
Function *Clone = CloneFunction(Original, VMap, false);
Clone->setName(Original->getName() + "-cloned");
// Adds the clone to the Module
M->getFunctionList().push_back(Clone);
// Adds the clone to the CFG
CG->getOrInsertFunction(Clone);
// Adds each of the original function's CFG node's interprocedural out-edges
// to the clone's node. All of the original function's intraprocedural in-edges are redirected to the cloned function.
// The clone will have no interprocedural in-edges as it
// was just created.
CallGraphNode *CloneNode = CG->getOrInsertFunction(Clone);
for (Function::iterator BBI = Clone->begin(), BBE = Clone->end();
BBI != BBE; ++BBI) {
for (BasicBlock::iterator II = BBI->begin(), IE = BBI->end(); II != IE;
++II) {
CallSite CS(cast<Value>(II));
// If this isn't a call, or it is a call to an intrinsic...
if (!CS || isa<IntrinsicInst>(II))
continue;
Function *Callee = CS.getCalledFunction();
if (Callee == Original) {
Callee = Clone;
CS.setCalledFunction(Clone);
}
CloneNode->addCalledFunction(CS, CG->getOrInsertFunction(Callee));
}
}
return Clone;
}
/// UpdateCallGraphAfterInlining - Once we have cloned code over from a callee
/// into the caller, update the specified callgraph to reflect the changes we
/// made. Note that it's possible that not all code was copied over, so only
/// some edges of the callgraph may remain.
static void UpdateCallGraphAfterInlining(CallSite CS,
Function::iterator FirstNewBlock,
DenseMap<const Value*, Value*> &ValueMap,
CallGraph &CG) {
const Function *Caller = CS.getInstruction()->getParent()->getParent();
const Function *Callee = CS.getCalledFunction();
CallGraphNode *CalleeNode = CG[Callee];
CallGraphNode *CallerNode = CG[Caller];
// Since we inlined some uninlined call sites in the callee into the caller,
// add edges from the caller to all of the callees of the callee.
CallGraphNode::iterator I = CalleeNode->begin(), E = CalleeNode->end();
// Consider the case where CalleeNode == CallerNode.
CallGraphNode::CalledFunctionsVector CallCache;
if (CalleeNode == CallerNode) {
CallCache.assign(I, E);
I = CallCache.begin();
E = CallCache.end();
}
for (; I != E; ++I) {
const Instruction *OrigCall = I->first.getInstruction();
DenseMap<const Value*, Value*>::iterator VMI = ValueMap.find(OrigCall);
// Only copy the edge if the call was inlined!
if (VMI != ValueMap.end() && VMI->second) {
// If the call was inlined, but then constant folded, there is no edge to
// add. Check for this case.
if (Instruction *NewCall = dyn_cast<Instruction>(VMI->second))
CallerNode->addCalledFunction(CallSite::get(NewCall), I->second);
}
}
// Update the call graph by deleting the edge from Callee to Caller. We must
// do this after the loop above in case Caller and Callee are the same.
CallerNode->removeCallEdgeFor(CS);
}
// InlineFunction - This function inlines the called function into the basic
// block of the caller. This returns false if it is not possible to inline this
// call. The program is still in a well defined state if this occurs though.
//
// Note that this only does one level of inlining. For example, if the
// instruction 'call B' is inlined, and 'B' calls 'C', then the call to 'C' now
// exists in the instruction stream. Similiarly this will inline a recursive
// function by one level.
//
bool llvm::InlineFunction(CallSite CS, CallGraph *CG, const TargetData *TD) {
Instruction *TheCall = CS.getInstruction();
assert(TheCall->getParent() && TheCall->getParent()->getParent() &&
"Instruction not in function!");
const Function *CalledFunc = CS.getCalledFunction();
if (CalledFunc == 0 || // Can't inline external function or indirect
CalledFunc->isDeclaration() || // call, or call to a vararg function!
CalledFunc->getFunctionType()->isVarArg()) return false;
// If the call to the callee is not a tail call, we must clear the 'tail'
// flags on any calls that we inline.
bool MustClearTailCallFlags =
!(isa<CallInst>(TheCall) && cast<CallInst>(TheCall)->isTailCall());
// If the call to the callee cannot throw, set the 'nounwind' flag on any
// calls that we inline.
bool MarkNoUnwind = CS.doesNotThrow();
BasicBlock *OrigBB = TheCall->getParent();
Function *Caller = OrigBB->getParent();
// GC poses two hazards to inlining, which only occur when the callee has GC:
// 1. If the caller has no GC, then the callee's GC must be propagated to the
// caller.
// 2. If the caller has a differing GC, it is invalid to inline.
if (CalledFunc->hasGC()) {
if (!Caller->hasGC())
Caller->setGC(CalledFunc->getGC());
else if (CalledFunc->getGC() != Caller->getGC())
return false;
}
// Get an iterator to the last basic block in the function, which will have
// the new function inlined after it.
//
Function::iterator LastBlock = &Caller->back();
// Make sure to capture all of the return instructions from the cloned
// function.
std::vector<ReturnInst*> Returns;
ClonedCodeInfo InlinedFunctionInfo;
Function::iterator FirstNewBlock;
{ // Scope to destroy ValueMap after cloning.
DenseMap<const Value*, Value*> ValueMap;
assert(CalledFunc->arg_size() == CS.arg_size() &&
"No varargs calls can be inlined!");
// Calculate the vector of arguments to pass into the function cloner, which
// matches up the formal to the actual argument values.
CallSite::arg_iterator AI = CS.arg_begin();
unsigned ArgNo = 0;
for (Function::const_arg_iterator I = CalledFunc->arg_begin(),
E = CalledFunc->arg_end(); I != E; ++I, ++AI, ++ArgNo) {
Value *ActualArg = *AI;
// When byval arguments actually inlined, we need to make the copy implied
// by them explicit. However, we don't do this if the callee is readonly
// or readnone, because the copy would be unneeded: the callee doesn't
// modify the struct.
if (CalledFunc->paramHasAttr(ArgNo+1, Attribute::ByVal) &&
!CalledFunc->onlyReadsMemory()) {
const Type *AggTy = cast<PointerType>(I->getType())->getElementType();
const Type *VoidPtrTy = PointerType::getUnqual(Type::Int8Ty);
// Create the alloca. If we have TargetData, use nice alignment.
unsigned Align = 1;
if (TD) Align = TD->getPrefTypeAlignment(AggTy);
Value *NewAlloca = new AllocaInst(AggTy, 0, Align, I->getName(),
Caller->begin()->begin());
// Emit a memcpy.
const Type *Tys[] = { Type::Int64Ty };
Function *MemCpyFn = Intrinsic::getDeclaration(Caller->getParent(),
Intrinsic::memcpy,
Tys, 1);
Value *DestCast = new BitCastInst(NewAlloca, VoidPtrTy, "tmp", TheCall);
Value *SrcCast = new BitCastInst(*AI, VoidPtrTy, "tmp", TheCall);
Value *Size;
if (TD == 0)
Size = ConstantExpr::getSizeOf(AggTy);
else
Size = ConstantInt::get(Type::Int64Ty, TD->getTypeStoreSize(AggTy));
// Always generate a memcpy of alignment 1 here because we don't know
// the alignment of the src pointer. Other optimizations can infer
// better alignment.
Value *CallArgs[] = {
DestCast, SrcCast, Size, ConstantInt::get(Type::Int32Ty, 1)
};
CallInst *TheMemCpy =
CallInst::Create(MemCpyFn, CallArgs, CallArgs+4, "", TheCall);
// If we have a call graph, update it.
if (CG) {
CallGraphNode *MemCpyCGN = CG->getOrInsertFunction(MemCpyFn);
CallGraphNode *CallerNode = (*CG)[Caller];
CallerNode->addCalledFunction(TheMemCpy, MemCpyCGN);
}
// Uses of the argument in the function should use our new alloca
// instead.
ActualArg = NewAlloca;
}
ValueMap[I] = ActualArg;
}
// We want the inliner to prune the code as it copies. We would LOVE to
// have no dead or constant instructions leftover after inlining occurs
// (which can happen, e.g., because an argument was constant), but we'll be
// happy with whatever the cloner can do.
CloneAndPruneFunctionInto(Caller, CalledFunc, ValueMap, Returns, ".i",
&InlinedFunctionInfo, TD);
// Remember the first block that is newly cloned over.
FirstNewBlock = LastBlock; ++FirstNewBlock;
// Update the callgraph if requested.
if (CG)
UpdateCallGraphAfterInlining(CS, FirstNewBlock, ValueMap, *CG);
}
// If there are any alloca instructions in the block that used to be the entry
// block for the callee, move them to the entry block of the caller. First
// calculate which instruction they should be inserted before. We insert the
// instructions at the end of the current alloca list.
//
{
BasicBlock::iterator InsertPoint = Caller->begin()->begin();
for (BasicBlock::iterator I = FirstNewBlock->begin(),
E = FirstNewBlock->end(); I != E; )
if (AllocaInst *AI = dyn_cast<AllocaInst>(I++)) {
// If the alloca is now dead, remove it. This often occurs due to code
// specialization.
if (AI->use_empty()) {
AI->eraseFromParent();
continue;
}
if (isa<Constant>(AI->getArraySize())) {
// Scan for the block of allocas that we can move over, and move them
// all at once.
while (isa<AllocaInst>(I) &&
isa<Constant>(cast<AllocaInst>(I)->getArraySize()))
++I;
// Transfer all of the allocas over in a block. Using splice means
// that the instructions aren't removed from the symbol table, then
// reinserted.
Caller->getEntryBlock().getInstList().splice(
InsertPoint,
FirstNewBlock->getInstList(),
AI, I);
}
}
}
// If the inlined code contained dynamic alloca instructions, wrap the inlined
// code with llvm.stacksave/llvm.stackrestore intrinsics.
if (InlinedFunctionInfo.ContainsDynamicAllocas) {
Module *M = Caller->getParent();
// Get the two intrinsics we care about.
Constant *StackSave, *StackRestore;
StackSave = Intrinsic::getDeclaration(M, Intrinsic::stacksave);
StackRestore = Intrinsic::getDeclaration(M, Intrinsic::stackrestore);
// If we are preserving the callgraph, add edges to the stacksave/restore
// functions for the calls we insert.
CallGraphNode *StackSaveCGN = 0, *StackRestoreCGN = 0, *CallerNode = 0;
if (CG) {
// We know that StackSave/StackRestore are Function*'s, because they are
// intrinsics which must have the right types.
StackSaveCGN = CG->getOrInsertFunction(cast<Function>(StackSave));
StackRestoreCGN = CG->getOrInsertFunction(cast<Function>(StackRestore));
CallerNode = (*CG)[Caller];
}
// Insert the llvm.stacksave.
CallInst *SavedPtr = CallInst::Create(StackSave, "savedstack",
FirstNewBlock->begin());
if (CG) CallerNode->addCalledFunction(SavedPtr, StackSaveCGN);
// Insert a call to llvm.stackrestore before any return instructions in the
// inlined function.
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
CallInst *CI = CallInst::Create(StackRestore, SavedPtr, "", Returns[i]);
if (CG) CallerNode->addCalledFunction(CI, StackRestoreCGN);
}
// Count the number of StackRestore calls we insert.
unsigned NumStackRestores = Returns.size();
// If we are inlining an invoke instruction, insert restores before each
// unwind. These unwinds will be rewritten into branches later.
if (InlinedFunctionInfo.ContainsUnwinds && isa<InvokeInst>(TheCall)) {
for (Function::iterator BB = FirstNewBlock, E = Caller->end();
BB != E; ++BB)
if (UnwindInst *UI = dyn_cast<UnwindInst>(BB->getTerminator())) {
CallInst::Create(StackRestore, SavedPtr, "", UI);
++NumStackRestores;
}
}
}
// If we are inlining tail call instruction through a call site that isn't
// marked 'tail', we must remove the tail marker for any calls in the inlined
// code. Also, calls inlined through a 'nounwind' call site should be marked
// 'nounwind'.
if (InlinedFunctionInfo.ContainsCalls &&
(MustClearTailCallFlags || MarkNoUnwind)) {
for (Function::iterator BB = FirstNewBlock, E = Caller->end();
BB != E; ++BB)
for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
if (CallInst *CI = dyn_cast<CallInst>(I)) {
if (MustClearTailCallFlags)
CI->setTailCall(false);
if (MarkNoUnwind)
CI->setDoesNotThrow();
}
}
// If we are inlining through a 'nounwind' call site then any inlined 'unwind'
// instructions are unreachable.
if (InlinedFunctionInfo.ContainsUnwinds && MarkNoUnwind)
for (Function::iterator BB = FirstNewBlock, E = Caller->end();
BB != E; ++BB) {
TerminatorInst *Term = BB->getTerminator();
if (isa<UnwindInst>(Term)) {
new UnreachableInst(Term);
BB->getInstList().erase(Term);
}
}
// If we are inlining for an invoke instruction, we must make sure to rewrite
// any inlined 'unwind' instructions into branches to the invoke exception
// destination, and call instructions into invoke instructions.
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall))
HandleInlinedInvoke(II, FirstNewBlock, InlinedFunctionInfo);
// If we cloned in _exactly one_ basic block, and if that block ends in a
// return instruction, we splice the body of the inlined callee directly into
// the calling basic block.
if (Returns.size() == 1 && std::distance(FirstNewBlock, Caller->end()) == 1) {
// Move all of the instructions right before the call.
OrigBB->getInstList().splice(TheCall, FirstNewBlock->getInstList(),
FirstNewBlock->begin(), FirstNewBlock->end());
// Remove the cloned basic block.
Caller->getBasicBlockList().pop_back();
// If the call site was an invoke instruction, add a branch to the normal
// destination.
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall))
BranchInst::Create(II->getNormalDest(), TheCall);
// If the return instruction returned a value, replace uses of the call with
// uses of the returned value.
if (!TheCall->use_empty()) {
ReturnInst *R = Returns[0];
TheCall->replaceAllUsesWith(R->getReturnValue());
}
// Since we are now done with the Call/Invoke, we can delete it.
TheCall->eraseFromParent();
// Since we are now done with the return instruction, delete it also.
Returns[0]->eraseFromParent();
// We are now done with the inlining.
return true;
}
// Otherwise, we have the normal case, of more than one block to inline or
// multiple return sites.
// We want to clone the entire callee function into the hole between the
// "starter" and "ender" blocks. How we accomplish this depends on whether
// this is an invoke instruction or a call instruction.
BasicBlock *AfterCallBB;
if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall)) {
// Add an unconditional branch to make this look like the CallInst case...
BranchInst *NewBr = BranchInst::Create(II->getNormalDest(), TheCall);
// Split the basic block. This guarantees that no PHI nodes will have to be
// updated due to new incoming edges, and make the invoke case more
// symmetric to the call case.
AfterCallBB = OrigBB->splitBasicBlock(NewBr,
CalledFunc->getName()+".exit");
} else { // It's a call
// If this is a call instruction, we need to split the basic block that
// the call lives in.
//
AfterCallBB = OrigBB->splitBasicBlock(TheCall,
CalledFunc->getName()+".exit");
}
// Change the branch that used to go to AfterCallBB to branch to the first
// basic block of the inlined function.
//
TerminatorInst *Br = OrigBB->getTerminator();
assert(Br && Br->getOpcode() == Instruction::Br &&
"splitBasicBlock broken!");
Br->setOperand(0, FirstNewBlock);
// Now that the function is correct, make it a little bit nicer. In
// particular, move the basic blocks inserted from the end of the function
// into the space made by splitting the source basic block.
Caller->getBasicBlockList().splice(AfterCallBB, Caller->getBasicBlockList(),
FirstNewBlock, Caller->end());
// Handle all of the return instructions that we just cloned in, and eliminate
// any users of the original call/invoke instruction.
const Type *RTy = CalledFunc->getReturnType();
if (Returns.size() > 1) {
// The PHI node should go at the front of the new basic block to merge all
// possible incoming values.
PHINode *PHI = 0;
if (!TheCall->use_empty()) {
PHI = PHINode::Create(RTy, TheCall->getName(),
AfterCallBB->begin());
// Anything that used the result of the function call should now use the
// PHI node as their operand.
TheCall->replaceAllUsesWith(PHI);
}
// Loop over all of the return instructions adding entries to the PHI node
// as appropriate.
if (PHI) {
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
ReturnInst *RI = Returns[i];
assert(RI->getReturnValue()->getType() == PHI->getType() &&
"Ret value not consistent in function!");
PHI->addIncoming(RI->getReturnValue(), RI->getParent());
}
}
// Add a branch to the merge points and remove return instructions.
for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
ReturnInst *RI = Returns[i];
BranchInst::Create(AfterCallBB, RI);
RI->eraseFromParent();
}
} else if (!Returns.empty()) {
// Otherwise, if there is exactly one return value, just replace anything
// using the return value of the call with the computed value.
if (!TheCall->use_empty())
TheCall->replaceAllUsesWith(Returns[0]->getReturnValue());
// Splice the code from the return block into the block that it will return
// to, which contains the code that was after the call.
BasicBlock *ReturnBB = Returns[0]->getParent();
AfterCallBB->getInstList().splice(AfterCallBB->begin(),
ReturnBB->getInstList());
// Update PHI nodes that use the ReturnBB to use the AfterCallBB.
ReturnBB->replaceAllUsesWith(AfterCallBB);
// Delete the return instruction now and empty ReturnBB now.
Returns[0]->eraseFromParent();
ReturnBB->eraseFromParent();
} else if (!TheCall->use_empty()) {
// No returns, but something is using the return value of the call. Just
// nuke the result.
TheCall->replaceAllUsesWith(UndefValue::get(TheCall->getType()));
}
// Since we are now done with the Call/Invoke, we can delete it.
TheCall->eraseFromParent();
// We should always be able to fold the entry block of the function into the
// single predecessor of the block...
assert(cast<BranchInst>(Br)->isUnconditional() && "splitBasicBlock broken!");
BasicBlock *CalleeEntry = cast<BranchInst>(Br)->getSuccessor(0);
// Splice the code entry block into calling block, right before the
// unconditional branch.
OrigBB->getInstList().splice(Br, CalleeEntry->getInstList());
CalleeEntry->replaceAllUsesWith(OrigBB); // Update PHI nodes
// Remove the unconditional branch.
OrigBB->getInstList().erase(Br);
// Now we can remove the CalleeEntry block, which is now empty.
Caller->getBasicBlockList().erase(CalleeEntry);
return true;
}
/// updateCallSites - Update all sites that call F to use NF.
CallGraphNode *SRETPromotion::updateCallSites(Function *F, Function *NF) {
CallGraph &CG = getAnalysis<CallGraph>();
SmallVector<Value*, 16> Args;
// Attributes - Keep track of the parameter attributes for the arguments.
SmallVector<AttributeWithIndex, 8> ArgAttrsVec;
// Get a new callgraph node for NF.
CallGraphNode *NF_CGN = CG.getOrInsertFunction(NF);
while (!F->use_empty()) {
CallSite CS(*F->use_begin());
Instruction *Call = CS.getInstruction();
const AttrListPtr &PAL = F->getAttributes();
// Add any return attributes.
if (Attributes attrs = PAL.getRetAttributes())
ArgAttrsVec.push_back(AttributeWithIndex::get(0, attrs));
// Copy arguments, however skip first one.
CallSite::arg_iterator AI = CS.arg_begin(), AE = CS.arg_end();
Value *FirstCArg = *AI;
++AI;
// 0th parameter attribute is reserved for return type.
// 1th parameter attribute is for first 1st sret argument.
unsigned ParamIndex = 2;
while (AI != AE) {
Args.push_back(*AI);
if (Attributes Attrs = PAL.getParamAttributes(ParamIndex))
ArgAttrsVec.push_back(AttributeWithIndex::get(ParamIndex - 1, Attrs));
++ParamIndex;
++AI;
}
// Add any function attributes.
if (Attributes attrs = PAL.getFnAttributes())
ArgAttrsVec.push_back(AttributeWithIndex::get(~0, attrs));
AttrListPtr NewPAL = AttrListPtr::get(ArgAttrsVec.begin(), ArgAttrsVec.end());
// Build new call instruction.
Instruction *New;
if (InvokeInst *II = dyn_cast<InvokeInst>(Call)) {
New = InvokeInst::Create(NF, II->getNormalDest(), II->getUnwindDest(),
Args.begin(), Args.end(), "", Call);
cast<InvokeInst>(New)->setCallingConv(CS.getCallingConv());
cast<InvokeInst>(New)->setAttributes(NewPAL);
} else {
New = CallInst::Create(NF, Args.begin(), Args.end(), "", Call);
cast<CallInst>(New)->setCallingConv(CS.getCallingConv());
cast<CallInst>(New)->setAttributes(NewPAL);
if (cast<CallInst>(Call)->isTailCall())
cast<CallInst>(New)->setTailCall();
}
Args.clear();
ArgAttrsVec.clear();
New->takeName(Call);
// Update the callgraph to know that the callsite has been transformed.
CallGraphNode *CalleeNode = CG[Call->getParent()->getParent()];
CalleeNode->removeCallEdgeFor(Call);
CalleeNode->addCalledFunction(New, NF_CGN);
// Update all users of sret parameter to extract value using extractvalue.
for (Value::use_iterator UI = FirstCArg->use_begin(),
UE = FirstCArg->use_end(); UI != UE; ) {
User *U2 = *UI++;
CallInst *C2 = dyn_cast<CallInst>(U2);
if (C2 && (C2 == Call))
continue;
GetElementPtrInst *UGEP = cast<GetElementPtrInst>(U2);
ConstantInt *Idx = cast<ConstantInt>(UGEP->getOperand(2));
Value *GR = ExtractValueInst::Create(New, Idx->getZExtValue(),
"evi", UGEP);
while(!UGEP->use_empty()) {
// isSafeToUpdateAllCallers has checked that all GEP uses are
// LoadInsts
LoadInst *L = cast<LoadInst>(*UGEP->use_begin());
L->replaceAllUsesWith(GR);
L->eraseFromParent();
}
UGEP->eraseFromParent();
continue;
}
Call->eraseFromParent();
}
return NF_CGN;
}