bool CallAnalyzer::visitBitCast(BitCastInst &I) {
// Propagate constants through bitcasts.
Constant *COp = dyn_cast<Constant>(I.getOperand(0));
if (!COp)
COp = SimplifiedValues.lookup(I.getOperand(0));
if (COp)
if (Constant *C = ConstantExpr::getBitCast(COp, I.getType())) {
SimplifiedValues[&I] = C;
return true;
}
// Track base/offsets through casts
std::pair<Value *, APInt> BaseAndOffset
= ConstantOffsetPtrs.lookup(I.getOperand(0));
// Casts don't change the offset, just wrap it up.
if (BaseAndOffset.first)
ConstantOffsetPtrs[&I] = BaseAndOffset;
// Also look for SROA candidates here.
Value *SROAArg;
DenseMap<Value *, int>::iterator CostIt;
if (lookupSROAArgAndCost(I.getOperand(0), SROAArg, CostIt))
SROAArgValues[&I] = SROAArg;
// Bitcasts are always zero cost.
return true;
}
bool Scalarizer::visitBitCastInst(BitCastInst &BCI) {
VectorType *DstVT = dyn_cast<VectorType>(BCI.getDestTy());
VectorType *SrcVT = dyn_cast<VectorType>(BCI.getSrcTy());
if (!DstVT || !SrcVT)
return false;
unsigned DstNumElems = DstVT->getNumElements();
unsigned SrcNumElems = SrcVT->getNumElements();
IRBuilder<> Builder(BCI.getParent(), &BCI);
Scatterer Op0 = scatter(&BCI, BCI.getOperand(0));
ValueVector Res;
Res.resize(DstNumElems);
if (DstNumElems == SrcNumElems) {
for (unsigned I = 0; I < DstNumElems; ++I)
Res[I] = Builder.CreateBitCast(Op0[I], DstVT->getElementType(),
BCI.getName() + ".i" + Twine(I));
} else if (DstNumElems > SrcNumElems) {
// <M x t1> -> <N*M x t2>. Convert each t1 to <N x t2> and copy the
// individual elements to the destination.
unsigned FanOut = DstNumElems / SrcNumElems;
Type *MidTy = VectorType::get(DstVT->getElementType(), FanOut);
unsigned ResI = 0;
for (unsigned Op0I = 0; Op0I < SrcNumElems; ++Op0I) {
Value *V = Op0[Op0I];
Instruction *VI;
// Look through any existing bitcasts before converting to <N x t2>.
// In the best case, the resulting conversion might be a no-op.
while ((VI = dyn_cast<Instruction>(V)) &&
VI->getOpcode() == Instruction::BitCast)
V = VI->getOperand(0);
V = Builder.CreateBitCast(V, MidTy, V->getName() + ".cast");
Scatterer Mid = scatter(&BCI, V);
for (unsigned MidI = 0; MidI < FanOut; ++MidI)
Res[ResI++] = Mid[MidI];
}
} else {
// <N*M x t1> -> <M x t2>. Convert each group of <N x t1> into a t2.
unsigned FanIn = SrcNumElems / DstNumElems;
Type *MidTy = VectorType::get(SrcVT->getElementType(), FanIn);
unsigned Op0I = 0;
for (unsigned ResI = 0; ResI < DstNumElems; ++ResI) {
Value *V = UndefValue::get(MidTy);
for (unsigned MidI = 0; MidI < FanIn; ++MidI)
V = Builder.CreateInsertElement(V, Op0[Op0I++], Builder.getInt32(MidI),
BCI.getName() + ".i" + Twine(ResI)
+ ".upto" + Twine(MidI));
Res[ResI] = Builder.CreateBitCast(V, DstVT->getElementType(),
BCI.getName() + ".i" + Twine(ResI));
}
}
gather(&BCI, Res);
return true;
}
/// DupRetToEnableTailCallOpts - Look for opportunities to duplicate return
/// instructions to the predecessor to enable tail call optimizations. The
/// case it is currently looking for is:
/// @code
/// bb0:
/// %tmp0 = tail call i32 @f0()
/// br label %return
/// bb1:
/// %tmp1 = tail call i32 @f1()
/// br label %return
/// bb2:
/// %tmp2 = tail call i32 @f2()
/// br label %return
/// return:
/// %retval = phi i32 [ %tmp0, %bb0 ], [ %tmp1, %bb1 ], [ %tmp2, %bb2 ]
/// ret i32 %retval
/// @endcode
///
/// =>
///
/// @code
/// bb0:
/// %tmp0 = tail call i32 @f0()
/// ret i32 %tmp0
/// bb1:
/// %tmp1 = tail call i32 @f1()
/// ret i32 %tmp1
/// bb2:
/// %tmp2 = tail call i32 @f2()
/// ret i32 %tmp2
/// @endcode
bool CodeGenPrepare::DupRetToEnableTailCallOpts(ReturnInst *RI) {
if (!TLI)
return false;
PHINode *PN = 0;
BitCastInst *BCI = 0;
Value *V = RI->getReturnValue();
if (V) {
BCI = dyn_cast<BitCastInst>(V);
if (BCI)
V = BCI->getOperand(0);
PN = dyn_cast<PHINode>(V);
if (!PN)
return false;
}
BasicBlock *BB = RI->getParent();
if (PN && PN->getParent() != BB)
return false;
// It's not safe to eliminate the sign / zero extension of the return value.
// See llvm::isInTailCallPosition().
const Function *F = BB->getParent();
Attributes CallerRetAttr = F->getAttributes().getRetAttributes();
if (CallerRetAttr.hasAttribute(Attributes::ZExt) ||
CallerRetAttr.hasAttribute(Attributes::SExt))
return false;
// Make sure there are no instructions between the PHI and return, or that the
// return is the first instruction in the block.
if (PN) {
BasicBlock::iterator BI = BB->begin();
do { ++BI; } while (isa<DbgInfoIntrinsic>(BI));
if (&*BI == BCI)
// Also skip over the bitcast.
++BI;
if (&*BI != RI)
return false;
} else {
BasicBlock::iterator BI = BB->begin();
while (isa<DbgInfoIntrinsic>(BI)) ++BI;
if (&*BI != RI)
return false;
}
/// Only dup the ReturnInst if the CallInst is likely to be emitted as a tail
/// call.
SmallVector<CallInst*, 4> TailCalls;
if (PN) {
for (unsigned I = 0, E = PN->getNumIncomingValues(); I != E; ++I) {
CallInst *CI = dyn_cast<CallInst>(PN->getIncomingValue(I));
// Make sure the phi value is indeed produced by the tail call.
if (CI && CI->hasOneUse() && CI->getParent() == PN->getIncomingBlock(I) &&
TLI->mayBeEmittedAsTailCall(CI))
TailCalls.push_back(CI);
}
} else {
SmallPtrSet<BasicBlock*, 4> VisitedBBs;
for (pred_iterator PI = pred_begin(BB), PE = pred_end(BB); PI != PE; ++PI) {
if (!VisitedBBs.insert(*PI))
continue;
BasicBlock::InstListType &InstList = (*PI)->getInstList();
BasicBlock::InstListType::reverse_iterator RI = InstList.rbegin();
BasicBlock::InstListType::reverse_iterator RE = InstList.rend();
do { ++RI; } while (RI != RE && isa<DbgInfoIntrinsic>(&*RI));
if (RI == RE)
continue;
CallInst *CI = dyn_cast<CallInst>(&*RI);
if (CI && CI->use_empty() && TLI->mayBeEmittedAsTailCall(CI))
TailCalls.push_back(CI);
}
}
bool Changed = false;
for (unsigned i = 0, e = TailCalls.size(); i != e; ++i) {
CallInst *CI = TailCalls[i];
CallSite CS(CI);
// Conservatively require the attributes of the call to match those of the
// return. Ignore noalias because it doesn't affect the call sequence.
Attributes CalleeRetAttr = CS.getAttributes().getRetAttributes();
if (AttrBuilder(CalleeRetAttr).
removeAttribute(Attributes::NoAlias) !=
AttrBuilder(CallerRetAttr).
removeAttribute(Attributes::NoAlias))
continue;
// Make sure the call instruction is followed by an unconditional branch to
// the return block.
BasicBlock *CallBB = CI->getParent();
BranchInst *BI = dyn_cast<BranchInst>(CallBB->getTerminator());
if (!BI || !BI->isUnconditional() || BI->getSuccessor(0) != BB)
continue;
// Duplicate the return into CallBB.
(void)FoldReturnIntoUncondBranch(RI, BB, CallBB);
ModifiedDT = Changed = true;
++NumRetsDup;
}
// If we eliminated all predecessors of the block, delete the block now.
if (Changed && !BB->hasAddressTaken() && pred_begin(BB) == pred_end(BB))
BB->eraseFromParent();
return Changed;
}
void GraphBuilder::visitBitCastInst(BitCastInst &I) {
if (!isa<PointerType>(I.getType())) return; // Only pointers
DSNodeHandle Ptr = getValueDest(I.getOperand(0));
if (Ptr.isNull()) return;
setDestTo(I, Ptr);
}
bool CombineNoopCasts::runOnBasicBlock(BasicBlock &BB) {
bool Changed = false;
for (BasicBlock::iterator BBI = BB.begin(), BBE = BB.end(); BBI != BBE;) {
Instruction* Inst = BBI++;
if (IntToPtrInst *Cast1 = dyn_cast<IntToPtrInst>(Inst)) {
if (isa<PtrToIntInst>(Cast1->getOperand(0)) ||
(isa<ConstantExpr>(Cast1->getOperand(0)) &&
cast<ConstantExpr>(Cast1->getOperand(0))->getOpcode() == Instruction::PtrToInt)) {
User *Cast2 = cast<User>(Cast1->getOperand(0));
Value *V = Cast2->getOperand(0);
if(Cast2->getType() != IntPtrType) {
continue;
}
if (V->getType() != Cast1->getType())
V = CopyDebug(new BitCastInst(V, Cast1->getType(),
V->getName() + ".bc", Cast1),
Cast2);
Cast1->replaceAllUsesWith(V);
if (Cast1->use_empty())
Cast1->eraseFromParent();
if (Cast2->use_empty() && isa<Instruction>(Cast2))
cast<Instruction>(Cast2)->eraseFromParent();
Changed = true;
}
} else if(PtrToIntInst *Cast1 = dyn_cast<PtrToIntInst>(Inst)) {
if(Cast1->getType() != IntPtrType) {
continue;
}
if (isa<IntToPtrInst>(Cast1->getOperand(0)) ||
(isa<ConstantExpr>(Cast1->getOperand(0)) &&
cast<ConstantExpr>(Cast1->getOperand(0))->getOpcode() == Instruction::IntToPtr)) {
User *Cast2 = cast<User>(Cast1->getOperand(0));
Value *V = Cast2->getOperand(0);
Cast1->replaceAllUsesWith(V);
if (Cast1->use_empty())
Cast1->eraseFromParent();
if (Cast2->use_empty() && isa<Instruction>(Cast2))
cast<Instruction>(Cast2)->eraseFromParent();
Changed = true;
} else if(BitCastInst *Cast2 = dyn_cast<BitCastInst>(Cast1->getOperand(0))) {
Cast1->setOperand(0, Cast2->getOperand(0));
if (Cast2->use_empty())
Cast2->eraseFromParent();
}
} else if(BitCastInst* Cast1 = dyn_cast<BitCastInst>(Inst)) {
if(Cast1->getOperand(0)->getType() == Cast1->getType()) {
Cast1->replaceAllUsesWith(Cast1->getOperand(0));
Cast1->eraseFromParent();
} else if(BitCastInst* Cast2 = dyn_cast<BitCastInst>(Cast1->getOperand(0))) {
if(Cast1->getType() == Cast2->getOperand(0)->getType()) {
Cast1->replaceAllUsesWith(Cast2->getOperand(0));
Cast1->eraseFromParent();
if (Cast2->use_empty()) {
Cast2->eraseFromParent();
}
}
}
}
}
// de-duplicate bitcasts:
DenseMap<std::pair<Value*, Type*>, BitCastInst*> BCs;
for (BasicBlock::iterator BBI = BB.begin(), BBE = BB.end(); BBI != BBE;) {
Instruction* Inst = BBI++;
if(PtrToIntInst *Cast1 = dyn_cast<PtrToIntInst>(Inst)) {
if(Cast1->use_empty()) {
Cast1->eraseFromParent();
continue;
}
}
if(!isa<BitCastInst>(Inst)) {
continue;
}
BitCastInst* BC = cast<BitCastInst>(Inst);
auto Val = std::make_pair(BC->getOperand(0), BC->getType());
auto BI = BCs.find(Val);
if(BI == BCs.end()) {
BCs.insert(std::make_pair(Val, BC));
} else {
BC->replaceAllUsesWith(BI->second);
BC->eraseFromParent();
}
}
return Changed;
}
/// extractMallocCallFromBitCast - Returns the corresponding CallInst if the
/// instruction is a bitcast of the result of a malloc call.
CallInst *llvm::extractMallocCallFromBitCast(Value *I) {
BitCastInst *BCI = dyn_cast<BitCastInst>(I);
return (isBitCastOfMallocCall(BCI)) ? cast<CallInst>(BCI->getOperand(0))
: NULL;
}
/*
* Rewrite OpenMP call sites and their associated kernel functions -- the folloiwng pattern
call void @GOMP_parallel_start(void (i8*)* @_Z20initialize_variablesiPfS_.omp_fn.4, i8* %.omp_data_o.5571, i32 0) nounwind
call void @_Z20initialize_variablesiPfS_.omp_fn.4(i8* %.omp_data_o.5571) nounwind
call void @GOMP_parallel_end() nounwind
*/
void HeteroOMPTransform::rewrite_omp_call_sites(Module &M) {
SmallVector<Instruction *, 16> toDelete;
DenseMap<Value *, Value *> ValueMap;
for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I){
if (!I->isDeclaration()) {
for (Function::iterator BBI = I->begin(), BBE = I->end(); BBI != BBE; ++BBI) {
bool match = false;
for (BasicBlock::iterator INSNI = BBI->begin(), INSNE = BBI->end(); INSNI != INSNE; ++INSNI) {
if (isa<CallInst>(INSNI)) {
CallSite CI(cast<Instruction>(INSNI));
if (CI.getCalledFunction() != NULL){
string called_func_name = CI.getCalledFunction()->getName();
if (called_func_name == OMP_PARALLEL_START_NAME && CI.arg_size() == 3) {
// change alloc to malloc_shared
// %5 = call i8* @_Z13malloc_sharedm(i64 20) ; <i8*> [#uses=5]
// %6 = bitcast i8* %5 to float* ; <float*> [#uses=2]
AllocaInst *AllocCall;
Value *arg_0 = CI.getArgument(0); // function
Value *arg_1 = CI.getArgument(1); // context
Value *loop_ub = NULL;
Function *function;
BitCastInst* BCI;
Function *kernel_function;
BasicBlock::iterator iI(*INSNI);
//BasicBlock::iterator iJ = iI+1;
iI++; iI++;
//BasicBlock::iterator iK = iI;
CallInst /**next,*/ *next_next;
if (arg_0 != NULL && arg_1 != NULL /*&& (next = dyn_cast<CallInst>(*iJ))*/
&& (next_next = dyn_cast<CallInst>(iI)) && (next_next->getCalledFunction() != NULL)
&& (next_next->getCalledFunction()->getName() == OMP_PARALLEL_END_NAME)
&& (BCI = dyn_cast<BitCastInst>(arg_1)) && (AllocCall = dyn_cast<AllocaInst>(BCI->getOperand(0)))
&& (function = dyn_cast<Function>(arg_0)) && (loop_ub = find_loop_upper_bound (AllocCall))
&& (kernel_function=convert_to_kernel_function (M, function))){
SmallVector<Value*, 16> Args;
Args.push_back(AllocCall->getArraySize());
Instruction *MallocCall = CallInst::Create(mallocFnTy, Args, "", AllocCall);
CastInst *MallocCast = CastInst::Create(Instruction::BitCast, MallocCall, AllocCall->getType(), "", AllocCall);
ValueMap[AllocCall] = MallocCast;
//AllocCall->replaceAllUsesWith(MallocCall);
// Add offload function
Args.clear();
Args.push_back(loop_ub);
Args.push_back(BCI);
Args.push_back(kernel_function);
if (offloadFnTy == NULL) {
init_offload_type(M, kernel_function);
}
Instruction *call = CallInst::Create(offloadFnTy, Args, "", INSNI);
if (find(toDelete.begin(), toDelete.end(), AllocCall) == toDelete.end()){
toDelete.push_back(AllocCall);
}
toDelete.push_back(&(*INSNI));
match = true;
}
}
else if (called_func_name == OMP_PARALLEL_END_NAME && CI.arg_size() == 0 && match) {
toDelete.push_back(&(*INSNI));
match = false;
}
else if (match) {
toDelete.push_back(&(*INSNI));
}
}
}
}
}
}
}
/* Replace AllocCalls by MallocCalls */
for(DenseMap<Value *, Value *>::iterator I = ValueMap.begin(), E = ValueMap.end(); I != E; I++) {
I->first->replaceAllUsesWith(I->second);
}
/* delete the instructions for get_omp_num_thread and get_omp_thread_num */
while(!toDelete.empty()) {
Instruction *g = toDelete.back();
toDelete.pop_back();
g->eraseFromParent();
}
}
