bool CallAnalyzer::visitAlloca(AllocaInst &I) {
// Check whether inlining will turn a dynamic alloca into a static
// alloca, and handle that case.
if (I.isArrayAllocation()) {
if (Constant *Size = SimplifiedValues.lookup(I.getArraySize())) {
ConstantInt *AllocSize = dyn_cast<ConstantInt>(Size);
assert(AllocSize && "Allocation size not a constant int?");
Type *Ty = I.getAllocatedType();
AllocatedSize += Ty->getPrimitiveSizeInBits() * AllocSize->getZExtValue();
return Base::visitAlloca(I);
}
}
// Accumulate the allocated size.
if (I.isStaticAlloca()) {
Type *Ty = I.getAllocatedType();
AllocatedSize += (DL ? DL->getTypeAllocSize(Ty) :
Ty->getPrimitiveSizeInBits());
}
// We will happily inline static alloca instructions.
if (I.isStaticAlloca())
return Base::visitAlloca(I);
// FIXME: This is overly conservative. Dynamic allocas are inefficient for
// a variety of reasons, and so we would like to not inline them into
// functions which don't currently have a dynamic alloca. This simply
// disables inlining altogether in the presence of a dynamic alloca.
HasDynamicAlloca = true;
return false;
}
SizeOffsetEvalType ObjectSizeOffsetEvaluator::visitAllocaInst(AllocaInst &I) {
if (!I.getAllocatedType()->isSized())
return unknown();
// must be a VLA
assert(I.isArrayAllocation());
Value *ArraySize = I.getArraySize();
Value *Size = ConstantInt::get(ArraySize->getType(),
TD->getTypeAllocSize(I.getAllocatedType()));
Size = Builder.CreateMul(Size, ArraySize);
return std::make_pair(Size, Zero);
}
void AllocaMergingBase::analyzeBlock(const cheerp::Registerize& registerize, BasicBlock& BB,
AllocaInfos& allocaInfos)
{
for(Instruction& I: BB)
{
if(I.getOpcode() == Instruction::Alloca)
{
AllocaInst* AI = cast<AllocaInst>(&I);
if(AI->isArrayAllocation())
continue;
allocaInfos.push_back(std::make_pair(AI, registerize.getLiveRangeForAlloca(AI)));
}
}
}
SizeOffsetType ObjectSizeOffsetVisitor::visitAllocaInst(AllocaInst &I) {
if (!I.getAllocatedType()->isSized())
return unknown();
APInt Size(IntTyBits, TD->getTypeAllocSize(I.getAllocatedType()));
if (!I.isArrayAllocation())
return std::make_pair(align(Size, I.getAlignment()), Zero);
Value *ArraySize = I.getArraySize();
if (const ConstantInt *C = dyn_cast<ConstantInt>(ArraySize)) {
Size *= C->getValue().zextOrSelf(IntTyBits);
return std::make_pair(align(Size, I.getAlignment()), Zero);
}
return unknown();
}
SizeOffsetType ObjectSizeOffsetVisitor::visitAllocaInst(AllocaInst &I) {
if (!I.getAllocatedType()->isSized())
return unknown();
APInt Size(IntTyBits, DL.getTypeAllocSize(I.getAllocatedType()));
if (!I.isArrayAllocation())
return std::make_pair(align(Size, I.getAlignment()), Zero);
Value *ArraySize = I.getArraySize();
if (const ConstantInt *C = dyn_cast<ConstantInt>(ArraySize)) {
APInt NumElems = C->getValue();
if (!CheckedZextOrTrunc(NumElems))
return unknown();
bool Overflow;
Size = Size.umul_ov(NumElems, Overflow);
return Overflow ? unknown() : std::make_pair(align(Size, I.getAlignment()),
Zero);
}
return unknown();
}
void AMDGPUPromoteAlloca::handleAlloca(AllocaInst &I) {
// Array allocations are probably not worth handling, since an allocation of
// the array type is the canonical form.
if (!I.isStaticAlloca() || I.isArrayAllocation())
return;
IRBuilder<> Builder(&I);
// First try to replace the alloca with a vector
Type *AllocaTy = I.getAllocatedType();
DEBUG(dbgs() << "Trying to promote " << I << '\n');
if (tryPromoteAllocaToVector(&I))
return;
DEBUG(dbgs() << " alloca is not a candidate for vectorization.\n");
const Function &ContainingFunction = *I.getParent()->getParent();
// FIXME: We should also try to get this value from the reqd_work_group_size
// function attribute if it is available.
unsigned WorkGroupSize = AMDGPU::getMaximumWorkGroupSize(ContainingFunction);
int AllocaSize =
WorkGroupSize * Mod->getDataLayout().getTypeAllocSize(AllocaTy);
if (AllocaSize > LocalMemAvailable) {
DEBUG(dbgs() << " Not enough local memory to promote alloca.\n");
return;
}
std::vector<Value*> WorkList;
if (!collectUsesWithPtrTypes(&I, WorkList)) {
DEBUG(dbgs() << " Do not know how to convert all uses\n");
return;
}
DEBUG(dbgs() << "Promoting alloca to local memory\n");
LocalMemAvailable -= AllocaSize;
Function *F = I.getParent()->getParent();
Type *GVTy = ArrayType::get(I.getAllocatedType(), WorkGroupSize);
GlobalVariable *GV = new GlobalVariable(
*Mod, GVTy, false, GlobalValue::InternalLinkage,
UndefValue::get(GVTy),
Twine(F->getName()) + Twine('.') + I.getName(),
nullptr,
GlobalVariable::NotThreadLocal,
AMDGPUAS::LOCAL_ADDRESS);
GV->setUnnamedAddr(true);
GV->setAlignment(I.getAlignment());
Value *TCntY, *TCntZ;
std::tie(TCntY, TCntZ) = getLocalSizeYZ(Builder);
Value *TIdX = getWorkitemID(Builder, 0);
Value *TIdY = getWorkitemID(Builder, 1);
Value *TIdZ = getWorkitemID(Builder, 2);
Value *Tmp0 = Builder.CreateMul(TCntY, TCntZ, "", true, true);
Tmp0 = Builder.CreateMul(Tmp0, TIdX);
Value *Tmp1 = Builder.CreateMul(TIdY, TCntZ, "", true, true);
Value *TID = Builder.CreateAdd(Tmp0, Tmp1);
TID = Builder.CreateAdd(TID, TIdZ);
Value *Indices[] = {
Constant::getNullValue(Type::getInt32Ty(Mod->getContext())),
TID
};
Value *Offset = Builder.CreateInBoundsGEP(GVTy, GV, Indices);
I.mutateType(Offset->getType());
I.replaceAllUsesWith(Offset);
I.eraseFromParent();
for (Value *V : WorkList) {
CallInst *Call = dyn_cast<CallInst>(V);
if (!Call) {
Type *EltTy = V->getType()->getPointerElementType();
PointerType *NewTy = PointerType::get(EltTy, AMDGPUAS::LOCAL_ADDRESS);
// The operand's value should be corrected on its own.
if (isa<AddrSpaceCastInst>(V))
continue;
// FIXME: It doesn't really make sense to try to do this for all
// instructions.
V->mutateType(NewTy);
continue;
}
IntrinsicInst *Intr = dyn_cast<IntrinsicInst>(Call);
if (!Intr) {
// FIXME: What is this for? It doesn't make sense to promote arbitrary
// function calls. If the call is to a defined function that can also be
// promoted, we should be able to do this once that function is also
// rewritten.
std::vector<Type*> ArgTypes;
for (unsigned ArgIdx = 0, ArgEnd = Call->getNumArgOperands();
ArgIdx != ArgEnd; ++ArgIdx) {
ArgTypes.push_back(Call->getArgOperand(ArgIdx)->getType());
}
Function *F = Call->getCalledFunction();
FunctionType *NewType = FunctionType::get(Call->getType(), ArgTypes,
F->isVarArg());
Constant *C = Mod->getOrInsertFunction((F->getName() + ".local").str(),
NewType, F->getAttributes());
Function *NewF = cast<Function>(C);
Call->setCalledFunction(NewF);
continue;
}
Builder.SetInsertPoint(Intr);
switch (Intr->getIntrinsicID()) {
case Intrinsic::lifetime_start:
case Intrinsic::lifetime_end:
// These intrinsics are for address space 0 only
Intr->eraseFromParent();
continue;
case Intrinsic::memcpy: {
MemCpyInst *MemCpy = cast<MemCpyInst>(Intr);
Builder.CreateMemCpy(MemCpy->getRawDest(), MemCpy->getRawSource(),
MemCpy->getLength(), MemCpy->getAlignment(),
MemCpy->isVolatile());
Intr->eraseFromParent();
continue;
}
case Intrinsic::memmove: {
MemMoveInst *MemMove = cast<MemMoveInst>(Intr);
Builder.CreateMemMove(MemMove->getRawDest(), MemMove->getRawSource(),
MemMove->getLength(), MemMove->getAlignment(),
MemMove->isVolatile());
Intr->eraseFromParent();
continue;
}
case Intrinsic::memset: {
MemSetInst *MemSet = cast<MemSetInst>(Intr);
Builder.CreateMemSet(MemSet->getRawDest(), MemSet->getValue(),
MemSet->getLength(), MemSet->getAlignment(),
MemSet->isVolatile());
Intr->eraseFromParent();
continue;
}
case Intrinsic::invariant_start:
case Intrinsic::invariant_end:
case Intrinsic::invariant_group_barrier:
Intr->eraseFromParent();
// FIXME: I think the invariant marker should still theoretically apply,
// but the intrinsics need to be changed to accept pointers with any
// address space.
continue;
case Intrinsic::objectsize: {
Value *Src = Intr->getOperand(0);
Type *SrcTy = Src->getType()->getPointerElementType();
Function *ObjectSize = Intrinsic::getDeclaration(Mod,
Intrinsic::objectsize,
{ Intr->getType(), PointerType::get(SrcTy, AMDGPUAS::LOCAL_ADDRESS) }
);
CallInst *NewCall
= Builder.CreateCall(ObjectSize, { Src, Intr->getOperand(1) });
Intr->replaceAllUsesWith(NewCall);
Intr->eraseFromParent();
continue;
}
default:
Intr->dump();
llvm_unreachable("Don't know how to promote alloca intrinsic use.");
}
}
}
// emit code to unpack a raw value from a box into registers or a stack slot
static Value *emit_unbox(jl_codectx_t &ctx, Type *to, const jl_cgval_t &x, jl_value_t *jt, Value *dest, bool volatile_store)
{
assert(to != T_void);
// TODO: fully validate that x.typ == jt?
if (x.isghost) {
// this can happen when a branch yielding a different type ends
// up being dead code, and type inference knows that the other
// branch's type is the only one that matters.
if (type_is_ghost(to)) {
return NULL;
}
//emit_error(ctx, "emit_unbox: a type mismatch error in occurred during codegen");
return UndefValue::get(to); // type mismatch error
}
Constant *c = x.constant ? julia_const_to_llvm(x.constant) : NULL;
if (!x.ispointer() || c) { // already unboxed, but sometimes need conversion
Value *unboxed = emit_unboxed_coercion(ctx, to, c ? c : x.V);
if (!dest)
return unboxed;
Type *dest_ty = unboxed->getType()->getPointerTo();
if (dest->getType() != dest_ty)
dest = emit_bitcast(ctx, dest, dest_ty);
ctx.builder.CreateStore(unboxed, dest, volatile_store);
return NULL;
}
// bools stored as int8, so an extra Trunc is needed to get an int1
Value *p = x.constant ? literal_pointer_val(ctx, x.constant) : x.V;
Type *ptype = (to == T_int1 ? T_pint8 : to->getPointerTo());
Value *unboxed = NULL;
if (to == T_int1)
unboxed = ctx.builder.CreateTrunc(tbaa_decorate(x.tbaa, ctx.builder.CreateLoad(maybe_bitcast(ctx, p, ptype))), T_int1);
else if (jt == (jl_value_t*)jl_bool_type)
unboxed = ctx.builder.CreateZExt(ctx.builder.CreateTrunc(tbaa_decorate(x.tbaa, ctx.builder.CreateLoad(maybe_bitcast(ctx, p, ptype))), T_int1), to);
if (unboxed) {
if (!dest)
return unboxed;
ctx.builder.CreateStore(unboxed, dest);
return NULL;
}
unsigned alignment = julia_alignment(jt, 0);
if (dest) {
MDNode *tbaa = x.tbaa;
// the memcpy intrinsic does not allow to specify different alias tags
// for the load part (x.tbaa) and the store part (tbaa_stack).
// since the tbaa lattice has to be a tree we have unfortunately
// x.tbaa ∪ tbaa_stack = tbaa_root if x.tbaa != tbaa_stack
if (tbaa != tbaa_stack)
tbaa = NULL;
emit_memcpy(ctx, dest, p, jl_datatype_size(jt), alignment, volatile_store, tbaa);
return NULL;
}
else {
if (p->getType() != ptype && isa<AllocaInst>(p)) {
// LLVM's mem2reg can't handle coercion if the load/store type does
// not match the type of the alloca. As such, it is better to
// perform the load using the alloca's type and then perform the
// appropriate coercion manually.
AllocaInst *AI = cast<AllocaInst>(p);
Type *AllocType = AI->getAllocatedType();
#if JL_LLVM_VERSION >= 40000
const DataLayout &DL = jl_data_layout;
#else
const DataLayout &DL = jl_ExecutionEngine->getDataLayout();
#endif
if (!AI->isArrayAllocation() &&
(AllocType->isFloatingPointTy() || AllocType->isIntegerTy() || AllocType->isPointerTy()) &&
(to->isFloatingPointTy() || to->isIntegerTy() || to->isPointerTy()) &&
DL.getTypeSizeInBits(AllocType) == DL.getTypeSizeInBits(to)) {
Instruction *load = ctx.builder.CreateAlignedLoad(p, alignment);
return emit_unboxed_coercion(ctx, to, tbaa_decorate(x.tbaa, load));
}
}
p = maybe_bitcast(ctx, p, ptype);
Instruction *load = ctx.builder.CreateAlignedLoad(p, alignment);
return tbaa_decorate(x.tbaa, load);
}
}
/// If it is possible to inline the specified call site,
/// do so and update the CallGraph for this operation.
///
/// This function also does some basic book-keeping to update the IR. The
/// InlinedArrayAllocas map keeps track of any allocas that are already
/// available from other functions inlined into the caller. If we are able to
/// inline this call site we attempt to reuse already available allocas or add
/// any new allocas to the set if not possible.
static bool InlineCallIfPossible(CallSite CS, InlineFunctionInfo &IFI,
InlinedArrayAllocasTy &InlinedArrayAllocas,
int InlineHistory, bool InsertLifetime) {
Function *Callee = CS.getCalledFunction();
Function *Caller = CS.getCaller();
// Try to inline the function. Get the list of static allocas that were
// inlined.
if (!InlineFunction(CS, IFI, InsertLifetime))
return false;
AdjustCallerSSPLevel(Caller, Callee);
// Look at all of the allocas that we inlined through this call site. If we
// have already inlined other allocas through other calls into this function,
// then we know that they have disjoint lifetimes and that we can merge them.
//
// There are many heuristics possible for merging these allocas, and the
// different options have different tradeoffs. One thing that we *really*
// don't want to hurt is SRoA: once inlining happens, often allocas are no
// longer address taken and so they can be promoted.
//
// Our "solution" for that is to only merge allocas whose outermost type is an
// array type. These are usually not promoted because someone is using a
// variable index into them. These are also often the most important ones to
// merge.
//
// A better solution would be to have real memory lifetime markers in the IR
// and not have the inliner do any merging of allocas at all. This would
// allow the backend to do proper stack slot coloring of all allocas that
// *actually make it to the backend*, which is really what we want.
//
// Because we don't have this information, we do this simple and useful hack.
//
SmallPtrSet<AllocaInst*, 16> UsedAllocas;
// When processing our SCC, check to see if CS was inlined from some other
// call site. For example, if we're processing "A" in this code:
// A() { B() }
// B() { x = alloca ... C() }
// C() { y = alloca ... }
// Assume that C was not inlined into B initially, and so we're processing A
// and decide to inline B into A. Doing this makes an alloca available for
// reuse and makes a callsite (C) available for inlining. When we process
// the C call site we don't want to do any alloca merging between X and Y
// because their scopes are not disjoint. We could make this smarter by
// keeping track of the inline history for each alloca in the
// InlinedArrayAllocas but this isn't likely to be a significant win.
if (InlineHistory != -1) // Only do merging for top-level call sites in SCC.
return true;
// Loop over all the allocas we have so far and see if they can be merged with
// a previously inlined alloca. If not, remember that we had it.
for (unsigned AllocaNo = 0, e = IFI.StaticAllocas.size();
AllocaNo != e; ++AllocaNo) {
AllocaInst *AI = IFI.StaticAllocas[AllocaNo];
// Don't bother trying to merge array allocations (they will usually be
// canonicalized to be an allocation *of* an array), or allocations whose
// type is not itself an array (because we're afraid of pessimizing SRoA).
ArrayType *ATy = dyn_cast<ArrayType>(AI->getAllocatedType());
if (!ATy || AI->isArrayAllocation())
continue;
// Get the list of all available allocas for this array type.
std::vector<AllocaInst*> &AllocasForType = InlinedArrayAllocas[ATy];
// Loop over the allocas in AllocasForType to see if we can reuse one. Note
// that we have to be careful not to reuse the same "available" alloca for
// multiple different allocas that we just inlined, we use the 'UsedAllocas'
// set to keep track of which "available" allocas are being used by this
// function. Also, AllocasForType can be empty of course!
bool MergedAwayAlloca = false;
for (unsigned i = 0, e = AllocasForType.size(); i != e; ++i) {
AllocaInst *AvailableAlloca = AllocasForType[i];
unsigned Align1 = AI->getAlignment(),
Align2 = AvailableAlloca->getAlignment();
// The available alloca has to be in the right function, not in some other
// function in this SCC.
if (AvailableAlloca->getParent() != AI->getParent())
continue;
// If the inlined function already uses this alloca then we can't reuse
// it.
if (!UsedAllocas.insert(AvailableAlloca).second)
continue;
// Otherwise, we *can* reuse it, RAUW AI into AvailableAlloca and declare
// success!
DEBUG(dbgs() << " ***MERGED ALLOCA: " << *AI << "\n\t\tINTO: "
<< *AvailableAlloca << '\n');
AI->replaceAllUsesWith(AvailableAlloca);
if (Align1 != Align2) {
if (!Align1 || !Align2) {
const DataLayout &DL = Caller->getParent()->getDataLayout();
unsigned TypeAlign = DL.getABITypeAlignment(AI->getAllocatedType());
Align1 = Align1 ? Align1 : TypeAlign;
Align2 = Align2 ? Align2 : TypeAlign;
}
if (Align1 > Align2)
AvailableAlloca->setAlignment(AI->getAlignment());
}
AI->eraseFromParent();
MergedAwayAlloca = true;
++NumMergedAllocas;
IFI.StaticAllocas[AllocaNo] = nullptr;
break;
}
// If we already nuked the alloca, we're done with it.
if (MergedAwayAlloca)
continue;
// If we were unable to merge away the alloca either because there are no
// allocas of the right type available or because we reused them all
// already, remember that this alloca came from an inlined function and mark
// it used so we don't reuse it for other allocas from this inline
// operation.
AllocasForType.push_back(AI);
UsedAllocas.insert(AI);
}
return true;
}
// FIXME: Should try to pick the most likely to be profitable allocas first.
bool AMDGPUPromoteAlloca::handleAlloca(AllocaInst &I, bool SufficientLDS) {
// Array allocations are probably not worth handling, since an allocation of
// the array type is the canonical form.
if (!I.isStaticAlloca() || I.isArrayAllocation())
return false;
IRBuilder<> Builder(&I);
// First try to replace the alloca with a vector
Type *AllocaTy = I.getAllocatedType();
DEBUG(dbgs() << "Trying to promote " << I << '\n');
if (tryPromoteAllocaToVector(&I, AS))
return true; // Promoted to vector.
const Function &ContainingFunction = *I.getParent()->getParent();
CallingConv::ID CC = ContainingFunction.getCallingConv();
// Don't promote the alloca to LDS for shader calling conventions as the work
// item ID intrinsics are not supported for these calling conventions.
// Furthermore not all LDS is available for some of the stages.
switch (CC) {
case CallingConv::AMDGPU_KERNEL:
case CallingConv::SPIR_KERNEL:
break;
default:
DEBUG(dbgs() << " promote alloca to LDS not supported with calling convention.\n");
return false;
}
// Not likely to have sufficient local memory for promotion.
if (!SufficientLDS)
return false;
const AMDGPUSubtarget &ST =
TM->getSubtarget<AMDGPUSubtarget>(ContainingFunction);
unsigned WorkGroupSize = ST.getFlatWorkGroupSizes(ContainingFunction).second;
const DataLayout &DL = Mod->getDataLayout();
unsigned Align = I.getAlignment();
if (Align == 0)
Align = DL.getABITypeAlignment(I.getAllocatedType());
// FIXME: This computed padding is likely wrong since it depends on inverse
// usage order.
//
// FIXME: It is also possible that if we're allowed to use all of the memory
// could could end up using more than the maximum due to alignment padding.
uint32_t NewSize = alignTo(CurrentLocalMemUsage, Align);
uint32_t AllocSize = WorkGroupSize * DL.getTypeAllocSize(AllocaTy);
NewSize += AllocSize;
if (NewSize > LocalMemLimit) {
DEBUG(dbgs() << " " << AllocSize
<< " bytes of local memory not available to promote\n");
return false;
}
CurrentLocalMemUsage = NewSize;
std::vector<Value*> WorkList;
if (!collectUsesWithPtrTypes(&I, &I, WorkList)) {
DEBUG(dbgs() << " Do not know how to convert all uses\n");
return false;
}
DEBUG(dbgs() << "Promoting alloca to local memory\n");
Function *F = I.getParent()->getParent();
Type *GVTy = ArrayType::get(I.getAllocatedType(), WorkGroupSize);
GlobalVariable *GV = new GlobalVariable(
*Mod, GVTy, false, GlobalValue::InternalLinkage,
UndefValue::get(GVTy),
Twine(F->getName()) + Twine('.') + I.getName(),
nullptr,
GlobalVariable::NotThreadLocal,
AS.LOCAL_ADDRESS);
GV->setUnnamedAddr(GlobalValue::UnnamedAddr::Global);
GV->setAlignment(I.getAlignment());
Value *TCntY, *TCntZ;
std::tie(TCntY, TCntZ) = getLocalSizeYZ(Builder);
Value *TIdX = getWorkitemID(Builder, 0);
Value *TIdY = getWorkitemID(Builder, 1);
Value *TIdZ = getWorkitemID(Builder, 2);
Value *Tmp0 = Builder.CreateMul(TCntY, TCntZ, "", true, true);
Tmp0 = Builder.CreateMul(Tmp0, TIdX);
Value *Tmp1 = Builder.CreateMul(TIdY, TCntZ, "", true, true);
Value *TID = Builder.CreateAdd(Tmp0, Tmp1);
TID = Builder.CreateAdd(TID, TIdZ);
Value *Indices[] = {
Constant::getNullValue(Type::getInt32Ty(Mod->getContext())),
TID
};
Value *Offset = Builder.CreateInBoundsGEP(GVTy, GV, Indices);
I.mutateType(Offset->getType());
I.replaceAllUsesWith(Offset);
I.eraseFromParent();
for (Value *V : WorkList) {
CallInst *Call = dyn_cast<CallInst>(V);
if (!Call) {
if (ICmpInst *CI = dyn_cast<ICmpInst>(V)) {
Value *Src0 = CI->getOperand(0);
Type *EltTy = Src0->getType()->getPointerElementType();
PointerType *NewTy = PointerType::get(EltTy, AS.LOCAL_ADDRESS);
if (isa<ConstantPointerNull>(CI->getOperand(0)))
CI->setOperand(0, ConstantPointerNull::get(NewTy));
if (isa<ConstantPointerNull>(CI->getOperand(1)))
CI->setOperand(1, ConstantPointerNull::get(NewTy));
continue;
}
// The operand's value should be corrected on its own and we don't want to
// touch the users.
if (isa<AddrSpaceCastInst>(V))
continue;
Type *EltTy = V->getType()->getPointerElementType();
PointerType *NewTy = PointerType::get(EltTy, AS.LOCAL_ADDRESS);
// FIXME: It doesn't really make sense to try to do this for all
// instructions.
V->mutateType(NewTy);
// Adjust the types of any constant operands.
if (SelectInst *SI = dyn_cast<SelectInst>(V)) {
if (isa<ConstantPointerNull>(SI->getOperand(1)))
SI->setOperand(1, ConstantPointerNull::get(NewTy));
if (isa<ConstantPointerNull>(SI->getOperand(2)))
SI->setOperand(2, ConstantPointerNull::get(NewTy));
} else if (PHINode *Phi = dyn_cast<PHINode>(V)) {
for (unsigned I = 0, E = Phi->getNumIncomingValues(); I != E; ++I) {
if (isa<ConstantPointerNull>(Phi->getIncomingValue(I)))
Phi->setIncomingValue(I, ConstantPointerNull::get(NewTy));
}
}
continue;
}
IntrinsicInst *Intr = cast<IntrinsicInst>(Call);
Builder.SetInsertPoint(Intr);
switch (Intr->getIntrinsicID()) {
case Intrinsic::lifetime_start:
case Intrinsic::lifetime_end:
// These intrinsics are for address space 0 only
Intr->eraseFromParent();
continue;
case Intrinsic::memcpy: {
MemCpyInst *MemCpy = cast<MemCpyInst>(Intr);
Builder.CreateMemCpy(MemCpy->getRawDest(), MemCpy->getDestAlignment(),
MemCpy->getRawSource(), MemCpy->getSourceAlignment(),
MemCpy->getLength(), MemCpy->isVolatile());
Intr->eraseFromParent();
continue;
}
case Intrinsic::memmove: {
MemMoveInst *MemMove = cast<MemMoveInst>(Intr);
Builder.CreateMemMove(MemMove->getRawDest(), MemMove->getDestAlignment(),
MemMove->getRawSource(), MemMove->getSourceAlignment(),
MemMove->getLength(), MemMove->isVolatile());
Intr->eraseFromParent();
continue;
}
case Intrinsic::memset: {
MemSetInst *MemSet = cast<MemSetInst>(Intr);
Builder.CreateMemSet(MemSet->getRawDest(), MemSet->getValue(),
MemSet->getLength(), MemSet->getDestAlignment(),
MemSet->isVolatile());
Intr->eraseFromParent();
continue;
}
case Intrinsic::invariant_start:
case Intrinsic::invariant_end:
case Intrinsic::invariant_group_barrier:
Intr->eraseFromParent();
// FIXME: I think the invariant marker should still theoretically apply,
// but the intrinsics need to be changed to accept pointers with any
// address space.
continue;
case Intrinsic::objectsize: {
Value *Src = Intr->getOperand(0);
Type *SrcTy = Src->getType()->getPointerElementType();
Function *ObjectSize = Intrinsic::getDeclaration(Mod,
Intrinsic::objectsize,
{ Intr->getType(), PointerType::get(SrcTy, AS.LOCAL_ADDRESS) }
);
CallInst *NewCall = Builder.CreateCall(
ObjectSize, {Src, Intr->getOperand(1), Intr->getOperand(2)});
Intr->replaceAllUsesWith(NewCall);
Intr->eraseFromParent();
continue;
}
default:
Intr->print(errs());
llvm_unreachable("Don't know how to promote alloca intrinsic use.");
}
}
return true;
}
/// InlineCallIfPossible - If it is possible to inline the specified call site,
/// do so and update the CallGraph for this operation.
///
/// This function also does some basic book-keeping to update the IR. The
/// InlinedArrayAllocas map keeps track of any allocas that are already
/// available from other functions inlined into the caller. If we are able to
/// inline this call site we attempt to reuse already available allocas or add
/// any new allocas to the set if not possible.
static bool InlineCallIfPossible(CallSite CS, InlineFunctionInfo &IFI,
InlinedArrayAllocasTy &InlinedArrayAllocas) {
Function *Callee = CS.getCalledFunction();
Function *Caller = CS.getCaller();
// Try to inline the function. Get the list of static allocas that were
// inlined.
if (!InlineFunction(CS, IFI))
return false;
// If the inlined function had a higher stack protection level than the
// calling function, then bump up the caller's stack protection level.
if (Callee->hasFnAttr(Attribute::StackProtectReq))
Caller->addFnAttr(Attribute::StackProtectReq);
else if (Callee->hasFnAttr(Attribute::StackProtect) &&
!Caller->hasFnAttr(Attribute::StackProtectReq))
Caller->addFnAttr(Attribute::StackProtect);
// Look at all of the allocas that we inlined through this call site. If we
// have already inlined other allocas through other calls into this function,
// then we know that they have disjoint lifetimes and that we can merge them.
//
// There are many heuristics possible for merging these allocas, and the
// different options have different tradeoffs. One thing that we *really*
// don't want to hurt is SRoA: once inlining happens, often allocas are no
// longer address taken and so they can be promoted.
//
// Our "solution" for that is to only merge allocas whose outermost type is an
// array type. These are usually not promoted because someone is using a
// variable index into them. These are also often the most important ones to
// merge.
//
// A better solution would be to have real memory lifetime markers in the IR
// and not have the inliner do any merging of allocas at all. This would
// allow the backend to do proper stack slot coloring of all allocas that
// *actually make it to the backend*, which is really what we want.
//
// Because we don't have this information, we do this simple and useful hack.
//
SmallPtrSet<AllocaInst*, 16> UsedAllocas;
// Loop over all the allocas we have so far and see if they can be merged with
// a previously inlined alloca. If not, remember that we had it.
for (unsigned AllocaNo = 0, e = IFI.StaticAllocas.size();
AllocaNo != e; ++AllocaNo) {
AllocaInst *AI = IFI.StaticAllocas[AllocaNo];
// Don't bother trying to merge array allocations (they will usually be
// canonicalized to be an allocation *of* an array), or allocations whose
// type is not itself an array (because we're afraid of pessimizing SRoA).
const ArrayType *ATy = dyn_cast<ArrayType>(AI->getAllocatedType());
if (ATy == 0 || AI->isArrayAllocation())
continue;
// Get the list of all available allocas for this array type.
std::vector<AllocaInst*> &AllocasForType = InlinedArrayAllocas[ATy];
// Loop over the allocas in AllocasForType to see if we can reuse one. Note
// that we have to be careful not to reuse the same "available" alloca for
// multiple different allocas that we just inlined, we use the 'UsedAllocas'
// set to keep track of which "available" allocas are being used by this
// function. Also, AllocasForType can be empty of course!
bool MergedAwayAlloca = false;
for (unsigned i = 0, e = AllocasForType.size(); i != e; ++i) {
AllocaInst *AvailableAlloca = AllocasForType[i];
// The available alloca has to be in the right function, not in some other
// function in this SCC.
if (AvailableAlloca->getParent() != AI->getParent())
continue;
// If the inlined function already uses this alloca then we can't reuse
// it.
if (!UsedAllocas.insert(AvailableAlloca))
continue;
// Otherwise, we *can* reuse it, RAUW AI into AvailableAlloca and declare
// success!
DEBUG(dbgs() << " ***MERGED ALLOCA: " << *AI);
AI->replaceAllUsesWith(AvailableAlloca);
AI->eraseFromParent();
MergedAwayAlloca = true;
++NumMergedAllocas;
break;
}
// If we already nuked the alloca, we're done with it.
if (MergedAwayAlloca)
continue;
// If we were unable to merge away the alloca either because there are no
// allocas of the right type available or because we reused them all
// already, remember that this alloca came from an inlined function and mark
// it used so we don't reuse it for other allocas from this inline
// operation.
AllocasForType.push_back(AI);
UsedAllocas.insert(AI);
}
return true;
}
/// Look at all of the allocas that we inlined through this call site. If we
/// have already inlined other allocas through other calls into this function,
/// then we know that they have disjoint lifetimes and that we can merge them.
///
/// There are many heuristics possible for merging these allocas, and the
/// different options have different tradeoffs. One thing that we *really*
/// don't want to hurt is SRoA: once inlining happens, often allocas are no
/// longer address taken and so they can be promoted.
///
/// Our "solution" for that is to only merge allocas whose outermost type is an
/// array type. These are usually not promoted because someone is using a
/// variable index into them. These are also often the most important ones to
/// merge.
///
/// A better solution would be to have real memory lifetime markers in the IR
/// and not have the inliner do any merging of allocas at all. This would
/// allow the backend to do proper stack slot coloring of all allocas that
/// *actually make it to the backend*, which is really what we want.
///
/// Because we don't have this information, we do this simple and useful hack.
static void mergeInlinedArrayAllocas(
Function *Caller, InlineFunctionInfo &IFI,
InlinedArrayAllocasTy &InlinedArrayAllocas, int InlineHistory) {
SmallPtrSet<AllocaInst *, 16> UsedAllocas;
// When processing our SCC, check to see if CS was inlined from some other
// call site. For example, if we're processing "A" in this code:
// A() { B() }
// B() { x = alloca ... C() }
// C() { y = alloca ... }
// Assume that C was not inlined into B initially, and so we're processing A
// and decide to inline B into A. Doing this makes an alloca available for
// reuse and makes a callsite (C) available for inlining. When we process
// the C call site we don't want to do any alloca merging between X and Y
// because their scopes are not disjoint. We could make this smarter by
// keeping track of the inline history for each alloca in the
// InlinedArrayAllocas but this isn't likely to be a significant win.
if (InlineHistory != -1) // Only do merging for top-level call sites in SCC.
return;
// Loop over all the allocas we have so far and see if they can be merged with
// a previously inlined alloca. If not, remember that we had it.
for (unsigned AllocaNo = 0, e = IFI.StaticAllocas.size(); AllocaNo != e;
++AllocaNo) {
AllocaInst *AI = IFI.StaticAllocas[AllocaNo];
// Don't bother trying to merge array allocations (they will usually be
// canonicalized to be an allocation *of* an array), or allocations whose
// type is not itself an array (because we're afraid of pessimizing SRoA).
ArrayType *ATy = dyn_cast<ArrayType>(AI->getAllocatedType());
if (!ATy || AI->isArrayAllocation())
continue;
// Get the list of all available allocas for this array type.
std::vector<AllocaInst *> &AllocasForType = InlinedArrayAllocas[ATy];
// Loop over the allocas in AllocasForType to see if we can reuse one. Note
// that we have to be careful not to reuse the same "available" alloca for
// multiple different allocas that we just inlined, we use the 'UsedAllocas'
// set to keep track of which "available" allocas are being used by this
// function. Also, AllocasForType can be empty of course!
bool MergedAwayAlloca = false;
for (AllocaInst *AvailableAlloca : AllocasForType) {
unsigned Align1 = AI->getAlignment(),
Align2 = AvailableAlloca->getAlignment();
// The available alloca has to be in the right function, not in some other
// function in this SCC.
if (AvailableAlloca->getParent() != AI->getParent())
continue;
// If the inlined function already uses this alloca then we can't reuse
// it.
if (!UsedAllocas.insert(AvailableAlloca).second)
continue;
// Otherwise, we *can* reuse it, RAUW AI into AvailableAlloca and declare
// success!
LLVM_DEBUG(dbgs() << " ***MERGED ALLOCA: " << *AI
<< "\n\t\tINTO: " << *AvailableAlloca << '\n');
// Move affected dbg.declare calls immediately after the new alloca to
// avoid the situation when a dbg.declare precedes its alloca.
if (auto *L = LocalAsMetadata::getIfExists(AI))
if (auto *MDV = MetadataAsValue::getIfExists(AI->getContext(), L))
for (User *U : MDV->users())
if (DbgDeclareInst *DDI = dyn_cast<DbgDeclareInst>(U))
DDI->moveBefore(AvailableAlloca->getNextNode());
AI->replaceAllUsesWith(AvailableAlloca);
if (Align1 != Align2) {
if (!Align1 || !Align2) {
const DataLayout &DL = Caller->getParent()->getDataLayout();
unsigned TypeAlign = DL.getABITypeAlignment(AI->getAllocatedType());
Align1 = Align1 ? Align1 : TypeAlign;
Align2 = Align2 ? Align2 : TypeAlign;
}
if (Align1 > Align2)
AvailableAlloca->setAlignment(AI->getAlignment());
}
AI->eraseFromParent();
MergedAwayAlloca = true;
++NumMergedAllocas;
IFI.StaticAllocas[AllocaNo] = nullptr;
break;
}
// If we already nuked the alloca, we're done with it.
if (MergedAwayAlloca)
continue;
// If we were unable to merge away the alloca either because there are no
// allocas of the right type available or because we reused them all
// already, remember that this alloca came from an inlined function and mark
// it used so we don't reuse it for other allocas from this inline
// operation.
AllocasForType.push_back(AI);
UsedAllocas.insert(AI);
}
}
