SDValue
AlphaTargetLowering::LowerReturn(SDValue Chain,
CallingConv::ID CallConv, bool isVarArg,
const SmallVectorImpl<ISD::OutputArg> &Outs,
DebugLoc dl, SelectionDAG &DAG) const {
SDValue Copy = DAG.getCopyToReg(Chain, dl, Alpha::R26,
DAG.getNode(AlphaISD::GlobalRetAddr,
DebugLoc(), MVT::i64),
SDValue());
switch (Outs.size()) {
default:
llvm_unreachable("Do not know how to return this many arguments!");
case 0:
break;
//return SDValue(); // ret void is legal
case 1: {
EVT ArgVT = Outs[0].Val.getValueType();
unsigned ArgReg;
if (ArgVT.isInteger())
ArgReg = Alpha::R0;
else {
assert(ArgVT.isFloatingPoint());
ArgReg = Alpha::F0;
}
Copy = DAG.getCopyToReg(Copy, dl, ArgReg,
Outs[0].Val, Copy.getValue(1));
if (DAG.getMachineFunction().getRegInfo().liveout_empty())
DAG.getMachineFunction().getRegInfo().addLiveOut(ArgReg);
break;
}
case 2: {
EVT ArgVT = Outs[0].Val.getValueType();
unsigned ArgReg1, ArgReg2;
if (ArgVT.isInteger()) {
ArgReg1 = Alpha::R0;
ArgReg2 = Alpha::R1;
} else {
assert(ArgVT.isFloatingPoint());
ArgReg1 = Alpha::F0;
ArgReg2 = Alpha::F1;
}
Copy = DAG.getCopyToReg(Copy, dl, ArgReg1,
Outs[0].Val, Copy.getValue(1));
if (std::find(DAG.getMachineFunction().getRegInfo().liveout_begin(),
DAG.getMachineFunction().getRegInfo().liveout_end(), ArgReg1)
== DAG.getMachineFunction().getRegInfo().liveout_end())
DAG.getMachineFunction().getRegInfo().addLiveOut(ArgReg1);
Copy = DAG.getCopyToReg(Copy, dl, ArgReg2,
Outs[1].Val, Copy.getValue(1));
if (std::find(DAG.getMachineFunction().getRegInfo().liveout_begin(),
DAG.getMachineFunction().getRegInfo().liveout_end(), ArgReg2)
== DAG.getMachineFunction().getRegInfo().liveout_end())
DAG.getMachineFunction().getRegInfo().addLiveOut(ArgReg2);
break;
}
}
return DAG.getNode(AlphaISD::RET_FLAG, dl,
MVT::Other, Copy, Copy.getValue(1));
}
Value *GenericToNVVM::getOrInsertCVTA(Module *M, Function *F,
GlobalVariable *GV,
IRBuilder<> &Builder) {
PointerType *GVType = GV->getType();
Value *CVTA = nullptr;
// See if the address space conversion requires the operand to be bitcast
// to i8 addrspace(n)* first.
EVT ExtendedGVType = EVT::getEVT(GVType->getElementType(), true);
if (!ExtendedGVType.isInteger() && !ExtendedGVType.isFloatingPoint()) {
// A bitcast to i8 addrspace(n)* on the operand is needed.
LLVMContext &Context = M->getContext();
unsigned int AddrSpace = GVType->getAddressSpace();
Type *DestTy = PointerType::get(Type::getInt8Ty(Context), AddrSpace);
CVTA = Builder.CreateBitCast(GV, DestTy, "cvta");
// Insert the address space conversion.
Type *ResultType =
PointerType::get(Type::getInt8Ty(Context), llvm::ADDRESS_SPACE_GENERIC);
SmallVector<Type *, 2> ParamTypes;
ParamTypes.push_back(ResultType);
ParamTypes.push_back(DestTy);
Function *CVTAFunction = Intrinsic::getDeclaration(
M, Intrinsic::nvvm_ptr_global_to_gen, ParamTypes);
CVTA = Builder.CreateCall(CVTAFunction, CVTA, "cvta");
// Another bitcast from i8 * to <the element type of GVType> * is
// required.
DestTy =
PointerType::get(GVType->getElementType(), llvm::ADDRESS_SPACE_GENERIC);
CVTA = Builder.CreateBitCast(CVTA, DestTy, "cvta");
} else {
// A simple CVTA is enough.
SmallVector<Type *, 2> ParamTypes;
ParamTypes.push_back(PointerType::get(GVType->getElementType(),
llvm::ADDRESS_SPACE_GENERIC));
ParamTypes.push_back(GVType);
Function *CVTAFunction = Intrinsic::getDeclaration(
M, Intrinsic::nvvm_ptr_global_to_gen, ParamTypes);
CVTA = Builder.CreateCall(CVTAFunction, GV, "cvta");
}
return CVTA;
}
unsigned ARMTTI::getCastInstrCost(unsigned Opcode, Type *Dst,
Type *Src) const {
int ISD = TLI->InstructionOpcodeToISD(Opcode);
assert(ISD && "Invalid opcode");
// Single to/from double precision conversions.
static const CostTblEntry<MVT::SimpleValueType> NEONFltDblTbl[] = {
// Vector fptrunc/fpext conversions.
{ ISD::FP_ROUND, MVT::v2f64, 2 },
{ ISD::FP_EXTEND, MVT::v2f32, 2 },
{ ISD::FP_EXTEND, MVT::v4f32, 4 }
};
if (Src->isVectorTy() && ST->hasNEON() && (ISD == ISD::FP_ROUND ||
ISD == ISD::FP_EXTEND)) {
std::pair<unsigned, MVT> LT = TLI->getTypeLegalizationCost(Src);
int Idx = CostTableLookup(NEONFltDblTbl, ISD, LT.second);
if (Idx != -1)
return LT.first * NEONFltDblTbl[Idx].Cost;
}
EVT SrcTy = TLI->getValueType(Src);
EVT DstTy = TLI->getValueType(Dst);
if (!SrcTy.isSimple() || !DstTy.isSimple())
return TargetTransformInfo::getCastInstrCost(Opcode, Dst, Src);
// Some arithmetic, load and store operations have specific instructions
// to cast up/down their types automatically at no extra cost.
// TODO: Get these tables to know at least what the related operations are.
static const TypeConversionCostTblEntry<MVT::SimpleValueType>
NEONVectorConversionTbl[] = {
{ ISD::SIGN_EXTEND, MVT::v4i32, MVT::v4i16, 0 },
{ ISD::ZERO_EXTEND, MVT::v4i32, MVT::v4i16, 0 },
{ ISD::SIGN_EXTEND, MVT::v2i64, MVT::v2i32, 1 },
{ ISD::ZERO_EXTEND, MVT::v2i64, MVT::v2i32, 1 },
{ ISD::TRUNCATE, MVT::v4i32, MVT::v4i64, 0 },
{ ISD::TRUNCATE, MVT::v4i16, MVT::v4i32, 1 },
// The number of vmovl instructions for the extension.
{ ISD::SIGN_EXTEND, MVT::v4i64, MVT::v4i16, 3 },
{ ISD::ZERO_EXTEND, MVT::v4i64, MVT::v4i16, 3 },
{ ISD::SIGN_EXTEND, MVT::v8i32, MVT::v8i8, 3 },
{ ISD::ZERO_EXTEND, MVT::v8i32, MVT::v8i8, 3 },
{ ISD::SIGN_EXTEND, MVT::v8i64, MVT::v8i8, 7 },
{ ISD::ZERO_EXTEND, MVT::v8i64, MVT::v8i8, 7 },
{ ISD::SIGN_EXTEND, MVT::v8i64, MVT::v8i16, 6 },
{ ISD::ZERO_EXTEND, MVT::v8i64, MVT::v8i16, 6 },
{ ISD::SIGN_EXTEND, MVT::v16i32, MVT::v16i8, 6 },
{ ISD::ZERO_EXTEND, MVT::v16i32, MVT::v16i8, 6 },
// Operations that we legalize using splitting.
{ ISD::TRUNCATE, MVT::v16i8, MVT::v16i32, 6 },
{ ISD::TRUNCATE, MVT::v8i8, MVT::v8i32, 3 },
// Vector float <-> i32 conversions.
{ ISD::SINT_TO_FP, MVT::v4f32, MVT::v4i32, 1 },
{ ISD::UINT_TO_FP, MVT::v4f32, MVT::v4i32, 1 },
{ ISD::SINT_TO_FP, MVT::v2f32, MVT::v2i8, 3 },
{ ISD::UINT_TO_FP, MVT::v2f32, MVT::v2i8, 3 },
{ ISD::SINT_TO_FP, MVT::v2f32, MVT::v2i16, 2 },
{ ISD::UINT_TO_FP, MVT::v2f32, MVT::v2i16, 2 },
{ ISD::SINT_TO_FP, MVT::v2f32, MVT::v2i32, 1 },
{ ISD::UINT_TO_FP, MVT::v2f32, MVT::v2i32, 1 },
{ ISD::SINT_TO_FP, MVT::v4f32, MVT::v4i1, 3 },
{ ISD::UINT_TO_FP, MVT::v4f32, MVT::v4i1, 3 },
{ ISD::SINT_TO_FP, MVT::v4f32, MVT::v4i8, 3 },
{ ISD::UINT_TO_FP, MVT::v4f32, MVT::v4i8, 3 },
{ ISD::SINT_TO_FP, MVT::v4f32, MVT::v4i16, 2 },
{ ISD::UINT_TO_FP, MVT::v4f32, MVT::v4i16, 2 },
{ ISD::SINT_TO_FP, MVT::v8f32, MVT::v8i16, 4 },
{ ISD::UINT_TO_FP, MVT::v8f32, MVT::v8i16, 4 },
{ ISD::SINT_TO_FP, MVT::v8f32, MVT::v8i32, 2 },
{ ISD::UINT_TO_FP, MVT::v8f32, MVT::v8i32, 2 },
{ ISD::SINT_TO_FP, MVT::v16f32, MVT::v16i16, 8 },
{ ISD::UINT_TO_FP, MVT::v16f32, MVT::v16i16, 8 },
{ ISD::SINT_TO_FP, MVT::v16f32, MVT::v16i32, 4 },
{ ISD::UINT_TO_FP, MVT::v16f32, MVT::v16i32, 4 },
{ ISD::FP_TO_SINT, MVT::v4i32, MVT::v4f32, 1 },
{ ISD::FP_TO_UINT, MVT::v4i32, MVT::v4f32, 1 },
{ ISD::FP_TO_SINT, MVT::v4i8, MVT::v4f32, 3 },
{ ISD::FP_TO_UINT, MVT::v4i8, MVT::v4f32, 3 },
{ ISD::FP_TO_SINT, MVT::v4i16, MVT::v4f32, 2 },
{ ISD::FP_TO_UINT, MVT::v4i16, MVT::v4f32, 2 },
// Vector double <-> i32 conversions.
{ ISD::SINT_TO_FP, MVT::v2f64, MVT::v2i32, 2 },
{ ISD::UINT_TO_FP, MVT::v2f64, MVT::v2i32, 2 },
{ ISD::SINT_TO_FP, MVT::v2f64, MVT::v2i8, 4 },
{ ISD::UINT_TO_FP, MVT::v2f64, MVT::v2i8, 4 },
{ ISD::SINT_TO_FP, MVT::v2f64, MVT::v2i16, 3 },
{ ISD::UINT_TO_FP, MVT::v2f64, MVT::v2i16, 3 },
{ ISD::SINT_TO_FP, MVT::v2f64, MVT::v2i32, 2 },
{ ISD::UINT_TO_FP, MVT::v2f64, MVT::v2i32, 2 },
{ ISD::FP_TO_SINT, MVT::v2i32, MVT::v2f64, 2 },
{ ISD::FP_TO_UINT, MVT::v2i32, MVT::v2f64, 2 },
{ ISD::FP_TO_SINT, MVT::v8i16, MVT::v8f32, 4 },
{ ISD::FP_TO_UINT, MVT::v8i16, MVT::v8f32, 4 },
{ ISD::FP_TO_SINT, MVT::v16i16, MVT::v16f32, 8 },
{ ISD::FP_TO_UINT, MVT::v16i16, MVT::v16f32, 8 }
};
if (SrcTy.isVector() && ST->hasNEON()) {
int Idx = ConvertCostTableLookup(NEONVectorConversionTbl, ISD,
DstTy.getSimpleVT(), SrcTy.getSimpleVT());
if (Idx != -1)
return NEONVectorConversionTbl[Idx].Cost;
}
// Scalar float to integer conversions.
static const TypeConversionCostTblEntry<MVT::SimpleValueType>
NEONFloatConversionTbl[] = {
{ ISD::FP_TO_SINT, MVT::i1, MVT::f32, 2 },
{ ISD::FP_TO_UINT, MVT::i1, MVT::f32, 2 },
{ ISD::FP_TO_SINT, MVT::i1, MVT::f64, 2 },
{ ISD::FP_TO_UINT, MVT::i1, MVT::f64, 2 },
{ ISD::FP_TO_SINT, MVT::i8, MVT::f32, 2 },
{ ISD::FP_TO_UINT, MVT::i8, MVT::f32, 2 },
{ ISD::FP_TO_SINT, MVT::i8, MVT::f64, 2 },
{ ISD::FP_TO_UINT, MVT::i8, MVT::f64, 2 },
{ ISD::FP_TO_SINT, MVT::i16, MVT::f32, 2 },
{ ISD::FP_TO_UINT, MVT::i16, MVT::f32, 2 },
{ ISD::FP_TO_SINT, MVT::i16, MVT::f64, 2 },
{ ISD::FP_TO_UINT, MVT::i16, MVT::f64, 2 },
{ ISD::FP_TO_SINT, MVT::i32, MVT::f32, 2 },
{ ISD::FP_TO_UINT, MVT::i32, MVT::f32, 2 },
{ ISD::FP_TO_SINT, MVT::i32, MVT::f64, 2 },
{ ISD::FP_TO_UINT, MVT::i32, MVT::f64, 2 },
{ ISD::FP_TO_SINT, MVT::i64, MVT::f32, 10 },
{ ISD::FP_TO_UINT, MVT::i64, MVT::f32, 10 },
{ ISD::FP_TO_SINT, MVT::i64, MVT::f64, 10 },
{ ISD::FP_TO_UINT, MVT::i64, MVT::f64, 10 }
};
if (SrcTy.isFloatingPoint() && ST->hasNEON()) {
int Idx = ConvertCostTableLookup(NEONFloatConversionTbl, ISD,
DstTy.getSimpleVT(), SrcTy.getSimpleVT());
if (Idx != -1)
return NEONFloatConversionTbl[Idx].Cost;
}
// Scalar integer to float conversions.
static const TypeConversionCostTblEntry<MVT::SimpleValueType>
NEONIntegerConversionTbl[] = {
{ ISD::SINT_TO_FP, MVT::f32, MVT::i1, 2 },
{ ISD::UINT_TO_FP, MVT::f32, MVT::i1, 2 },
{ ISD::SINT_TO_FP, MVT::f64, MVT::i1, 2 },
{ ISD::UINT_TO_FP, MVT::f64, MVT::i1, 2 },
{ ISD::SINT_TO_FP, MVT::f32, MVT::i8, 2 },
{ ISD::UINT_TO_FP, MVT::f32, MVT::i8, 2 },
{ ISD::SINT_TO_FP, MVT::f64, MVT::i8, 2 },
{ ISD::UINT_TO_FP, MVT::f64, MVT::i8, 2 },
{ ISD::SINT_TO_FP, MVT::f32, MVT::i16, 2 },
{ ISD::UINT_TO_FP, MVT::f32, MVT::i16, 2 },
{ ISD::SINT_TO_FP, MVT::f64, MVT::i16, 2 },
{ ISD::UINT_TO_FP, MVT::f64, MVT::i16, 2 },
{ ISD::SINT_TO_FP, MVT::f32, MVT::i32, 2 },
{ ISD::UINT_TO_FP, MVT::f32, MVT::i32, 2 },
{ ISD::SINT_TO_FP, MVT::f64, MVT::i32, 2 },
{ ISD::UINT_TO_FP, MVT::f64, MVT::i32, 2 },
{ ISD::SINT_TO_FP, MVT::f32, MVT::i64, 10 },
{ ISD::UINT_TO_FP, MVT::f32, MVT::i64, 10 },
{ ISD::SINT_TO_FP, MVT::f64, MVT::i64, 10 },
{ ISD::UINT_TO_FP, MVT::f64, MVT::i64, 10 }
};
if (SrcTy.isInteger() && ST->hasNEON()) {
int Idx = ConvertCostTableLookup(NEONIntegerConversionTbl, ISD,
DstTy.getSimpleVT(), SrcTy.getSimpleVT());
if (Idx != -1)
return NEONIntegerConversionTbl[Idx].Cost;
}
// Scalar integer conversion costs.
static const TypeConversionCostTblEntry<MVT::SimpleValueType>
ARMIntegerConversionTbl[] = {
// i16 -> i64 requires two dependent operations.
{ ISD::SIGN_EXTEND, MVT::i64, MVT::i16, 2 },
// Truncates on i64 are assumed to be free.
{ ISD::TRUNCATE, MVT::i32, MVT::i64, 0 },
{ ISD::TRUNCATE, MVT::i16, MVT::i64, 0 },
{ ISD::TRUNCATE, MVT::i8, MVT::i64, 0 },
{ ISD::TRUNCATE, MVT::i1, MVT::i64, 0 }
};
if (SrcTy.isInteger()) {
int Idx = ConvertCostTableLookup(ARMIntegerConversionTbl, ISD,
DstTy.getSimpleVT(), SrcTy.getSimpleVT());
if (Idx != -1)
return ARMIntegerConversionTbl[Idx].Cost;
}
return TargetTransformInfo::getCastInstrCost(Opcode, Dst, Src);
}
static bool CC_MipsO32(unsigned ValNo, EVT ValVT,
EVT LocVT, CCValAssign::LocInfo LocInfo,
ISD::ArgFlagsTy ArgFlags, CCState &State) {
static const unsigned IntRegsSize=4, FloatRegsSize=2;
static const unsigned IntRegs[] = {
Mips::A0, Mips::A1, Mips::A2, Mips::A3
};
static const unsigned F32Regs[] = {
Mips::F12, Mips::F14
};
static const unsigned F64Regs[] = {
Mips::D6, Mips::D7
};
unsigned Reg=0;
unsigned UnallocIntReg = State.getFirstUnallocated(IntRegs, IntRegsSize);
bool IntRegUsed = (IntRegs[UnallocIntReg] != (unsigned (Mips::A0)));
// Promote i8 and i16
if (LocVT == MVT::i8 || LocVT == MVT::i16) {
LocVT = MVT::i32;
if (ArgFlags.isSExt())
LocInfo = CCValAssign::SExt;
else if (ArgFlags.isZExt())
LocInfo = CCValAssign::ZExt;
else
LocInfo = CCValAssign::AExt;
}
if (ValVT == MVT::i32 || (ValVT == MVT::f32 && IntRegUsed)) {
Reg = State.AllocateReg(IntRegs, IntRegsSize);
IntRegUsed = true;
LocVT = MVT::i32;
}
if (ValVT.isFloatingPoint() && !IntRegUsed) {
if (ValVT == MVT::f32)
Reg = State.AllocateReg(F32Regs, FloatRegsSize);
else
Reg = State.AllocateReg(F64Regs, FloatRegsSize);
}
if (ValVT == MVT::f64 && IntRegUsed) {
if (UnallocIntReg != IntRegsSize) {
// If we hit register A3 as the first not allocated, we must
// mark it as allocated (shadow) and use the stack instead.
if (IntRegs[UnallocIntReg] != (unsigned (Mips::A3)))
Reg = Mips::A2;
for (;UnallocIntReg < IntRegsSize; ++UnallocIntReg)
State.AllocateReg(UnallocIntReg);
}
LocVT = MVT::i32;
}
if (!Reg) {
unsigned SizeInBytes = ValVT.getSizeInBits() >> 3;
unsigned Offset = State.AllocateStack(SizeInBytes, SizeInBytes);
State.addLoc(CCValAssign::getMem(ValNo, ValVT, Offset, LocVT, LocInfo));
} else
