SystemZISelLowering.cpp [plain text]
#include "SystemZISelLowering.h"
#include "SystemZCallingConv.h"
#include "SystemZConstantPoolValue.h"
#include "SystemZMachineFunctionInfo.h"
#include "SystemZTargetMachine.h"
#include "llvm/CodeGen/CallingConvLower.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/TargetLoweringObjectFileImpl.h"
#include <cctype>
using namespace llvm;
#define DEBUG_TYPE "systemz-lower"
namespace {
struct IPMConversion {
IPMConversion(unsigned xorValue, int64_t addValue, unsigned bit)
: XORValue(xorValue), AddValue(addValue), Bit(bit) {}
int64_t XORValue;
int64_t AddValue;
unsigned Bit;
};
struct Comparison {
Comparison(SDValue Op0In, SDValue Op1In)
: Op0(Op0In), Op1(Op1In), Opcode(0), ICmpType(0), CCValid(0), CCMask(0) {}
SDValue Op0, Op1;
unsigned Opcode;
unsigned ICmpType;
unsigned CCValid;
unsigned CCMask;
};
}
static bool is32Bit(EVT VT) {
switch (VT.getSimpleVT().SimpleTy) {
case MVT::i32:
return true;
case MVT::i64:
return false;
default:
llvm_unreachable("Unsupported type");
}
}
static MachineOperand earlyUseOperand(MachineOperand Op) {
if (Op.isReg())
Op.setIsKill(false);
return Op;
}
SystemZTargetLowering::SystemZTargetLowering(const TargetMachine &tm,
const SystemZSubtarget &STI)
: TargetLowering(tm), Subtarget(STI) {
MVT PtrVT = getPointerTy();
if (Subtarget.hasHighWord())
addRegisterClass(MVT::i32, &SystemZ::GRX32BitRegClass);
else
addRegisterClass(MVT::i32, &SystemZ::GR32BitRegClass);
addRegisterClass(MVT::i64, &SystemZ::GR64BitRegClass);
addRegisterClass(MVT::f32, &SystemZ::FP32BitRegClass);
addRegisterClass(MVT::f64, &SystemZ::FP64BitRegClass);
addRegisterClass(MVT::f128, &SystemZ::FP128BitRegClass);
computeRegisterProperties(Subtarget.getRegisterInfo());
setExceptionPointerRegister(SystemZ::R6D);
setExceptionSelectorRegister(SystemZ::R7D);
setStackPointerRegisterToSaveRestore(SystemZ::R15D);
setSchedulingPreference(Sched::RegPressure);
setBooleanContents(ZeroOrOneBooleanContent);
setBooleanVectorContents(ZeroOrOneBooleanContent);
setMinFunctionAlignment(2);
for (unsigned I = MVT::FIRST_INTEGER_VALUETYPE;
I <= MVT::LAST_FP_VALUETYPE;
++I) {
MVT VT = MVT::SimpleValueType(I);
if (isTypeLegal(VT)) {
setOperationAction(ISD::SETCC, VT, Custom);
setOperationAction(ISD::SELECT, VT, Expand);
setOperationAction(ISD::SELECT_CC, VT, Custom);
setOperationAction(ISD::BR_CC, VT, Custom);
}
}
setOperationAction(ISD::BR_JT, MVT::Other, Expand);
setOperationAction(ISD::BRCOND, MVT::Other, Expand);
for (unsigned I = MVT::FIRST_INTEGER_VALUETYPE;
I <= MVT::LAST_INTEGER_VALUETYPE;
++I) {
MVT VT = MVT::SimpleValueType(I);
if (isTypeLegal(VT)) {
setOperationAction(ISD::SDIV, VT, Expand);
setOperationAction(ISD::UDIV, VT, Expand);
setOperationAction(ISD::SREM, VT, Expand);
setOperationAction(ISD::UREM, VT, Expand);
setOperationAction(ISD::SDIVREM, VT, Custom);
setOperationAction(ISD::UDIVREM, VT, Custom);
setOperationAction(ISD::ATOMIC_LOAD, VT, Custom);
setOperationAction(ISD::ATOMIC_STORE, VT, Custom);
setOperationAction(ISD::ATOMIC_LOAD_SUB, VT, Custom);
setOperationAction(ISD::CTPOP, VT, Expand);
setOperationAction(ISD::CTTZ, VT, Expand);
setOperationAction(ISD::CTTZ_ZERO_UNDEF, VT, Expand);
setOperationAction(ISD::CTLZ_ZERO_UNDEF, VT, Expand);
setOperationAction(ISD::ROTR, VT, Expand);
setOperationAction(ISD::MULHS, VT, Expand);
setOperationAction(ISD::MULHU, VT, Expand);
setOperationAction(ISD::SMUL_LOHI, VT, Custom);
setOperationAction(ISD::UMUL_LOHI, VT, Custom);
if (!Subtarget.hasFPExtension())
setOperationAction(ISD::FP_TO_UINT, VT, Expand);
}
}
setOperationAction(ISD::ATOMIC_SWAP, MVT::i32, Custom);
setOperationAction(ISD::ATOMIC_LOAD_ADD, MVT::i32, Custom);
setOperationAction(ISD::ATOMIC_LOAD_SUB, MVT::i32, Custom);
setOperationAction(ISD::ATOMIC_LOAD_AND, MVT::i32, Custom);
setOperationAction(ISD::ATOMIC_LOAD_OR, MVT::i32, Custom);
setOperationAction(ISD::ATOMIC_LOAD_XOR, MVT::i32, Custom);
setOperationAction(ISD::ATOMIC_LOAD_NAND, MVT::i32, Custom);
setOperationAction(ISD::ATOMIC_LOAD_MIN, MVT::i32, Custom);
setOperationAction(ISD::ATOMIC_LOAD_MAX, MVT::i32, Custom);
setOperationAction(ISD::ATOMIC_LOAD_UMIN, MVT::i32, Custom);
setOperationAction(ISD::ATOMIC_LOAD_UMAX, MVT::i32, Custom);
setOperationAction(ISD::ATOMIC_CMP_SWAP, MVT::i32, Custom);
if (!Subtarget.hasFPExtension()) {
setOperationAction(ISD::UINT_TO_FP, MVT::i32, Promote);
setOperationAction(ISD::UINT_TO_FP, MVT::i64, Expand);
}
setOperationAction(ISD::CTLZ, MVT::i32, Promote);
setOperationAction(ISD::CTLZ, MVT::i64, Legal);
setOperationAction(ISD::OR, MVT::i64, Custom);
setOperationAction(ISD::SRL_PARTS, MVT::i64, Expand);
setOperationAction(ISD::SHL_PARTS, MVT::i64, Expand);
setOperationAction(ISD::SRA_PARTS, MVT::i64, Expand);
setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::i1, Expand);
for (MVT VT : MVT::integer_valuetypes()) {
setLoadExtAction(ISD::SEXTLOAD, VT, MVT::i1, Promote);
setLoadExtAction(ISD::ZEXTLOAD, VT, MVT::i1, Promote);
setLoadExtAction(ISD::EXTLOAD, VT, MVT::i1, Promote);
}
setOperationAction(ISD::ConstantPool, PtrVT, Custom);
setOperationAction(ISD::GlobalAddress, PtrVT, Custom);
setOperationAction(ISD::GlobalTLSAddress, PtrVT, Custom);
setOperationAction(ISD::BlockAddress, PtrVT, Custom);
setOperationAction(ISD::JumpTable, PtrVT, Custom);
setOperationAction(ISD::DYNAMIC_STACKALLOC, PtrVT, Custom);
setOperationAction(ISD::STACKSAVE, MVT::Other, Custom);
setOperationAction(ISD::STACKRESTORE, MVT::Other, Custom);
setOperationAction(ISD::PREFETCH, MVT::Other, Custom);
for (unsigned I = MVT::FIRST_FP_VALUETYPE;
I <= MVT::LAST_FP_VALUETYPE;
++I) {
MVT VT = MVT::SimpleValueType(I);
if (isTypeLegal(VT)) {
setOperationAction(ISD::FRINT, VT, Legal);
if (Subtarget.hasFPExtension()) {
setOperationAction(ISD::FNEARBYINT, VT, Legal);
setOperationAction(ISD::FFLOOR, VT, Legal);
setOperationAction(ISD::FCEIL, VT, Legal);
setOperationAction(ISD::FTRUNC, VT, Legal);
setOperationAction(ISD::FROUND, VT, Legal);
}
setOperationAction(ISD::FSIN, VT, Expand);
setOperationAction(ISD::FCOS, VT, Expand);
setOperationAction(ISD::FREM, VT, Expand);
}
}
setOperationAction(ISD::FMA, MVT::f32, Legal);
setOperationAction(ISD::FMA, MVT::f64, Legal);
setOperationAction(ISD::FMA, MVT::f128, Expand);
for (MVT VT : MVT::fp_valuetypes())
setLoadExtAction(ISD::EXTLOAD, VT, MVT::f80, Expand);
setTruncStoreAction(MVT::f64, MVT::f32, Expand);
setTruncStoreAction(MVT::f128, MVT::f32, Expand);
setTruncStoreAction(MVT::f128, MVT::f64, Expand);
setOperationAction(ISD::BITCAST, MVT::i32, Custom);
setOperationAction(ISD::BITCAST, MVT::f32, Custom);
setOperationAction(ISD::VASTART, MVT::Other, Custom);
setOperationAction(ISD::VACOPY, MVT::Other, Custom);
setOperationAction(ISD::VAEND, MVT::Other, Expand);
setTargetDAGCombine(ISD::SIGN_EXTEND);
MaxStoresPerMemcpy = 0;
MaxStoresPerMemcpyOptSize = 0;
MaxStoresPerMemset = 0;
MaxStoresPerMemsetOptSize = 0;
}
EVT SystemZTargetLowering::getSetCCResultType(LLVMContext &, EVT VT) const {
if (!VT.isVector())
return MVT::i32;
return VT.changeVectorElementTypeToInteger();
}
bool SystemZTargetLowering::isFMAFasterThanFMulAndFAdd(EVT VT) const {
VT = VT.getScalarType();
if (!VT.isSimple())
return false;
switch (VT.getSimpleVT().SimpleTy) {
case MVT::f32:
case MVT::f64:
return true;
case MVT::f128:
return false;
default:
break;
}
return false;
}
bool SystemZTargetLowering::isFPImmLegal(const APFloat &Imm, EVT VT) const {
return Imm.isZero() || Imm.isNegZero();
}
bool SystemZTargetLowering::allowsMisalignedMemoryAccesses(EVT VT,
unsigned,
unsigned,
bool *Fast) const {
if (Fast)
*Fast = true;
return true;
}
bool SystemZTargetLowering::isLegalAddressingMode(const AddrMode &AM,
Type *Ty) const {
if (AM.BaseGV)
return false;
if (!isInt<20>(AM.BaseOffs))
return false;
return AM.Scale == 0 || AM.Scale == 1;
}
bool SystemZTargetLowering::isTruncateFree(Type *FromType, Type *ToType) const {
if (!FromType->isIntegerTy() || !ToType->isIntegerTy())
return false;
unsigned FromBits = FromType->getPrimitiveSizeInBits();
unsigned ToBits = ToType->getPrimitiveSizeInBits();
return FromBits > ToBits;
}
bool SystemZTargetLowering::isTruncateFree(EVT FromVT, EVT ToVT) const {
if (!FromVT.isInteger() || !ToVT.isInteger())
return false;
unsigned FromBits = FromVT.getSizeInBits();
unsigned ToBits = ToVT.getSizeInBits();
return FromBits > ToBits;
}
TargetLowering::ConstraintType
SystemZTargetLowering::getConstraintType(const std::string &Constraint) const {
if (Constraint.size() == 1) {
switch (Constraint[0]) {
case 'a': case 'd': case 'f': case 'h': case 'r': return C_RegisterClass;
case 'Q': case 'R': case 'S': case 'T': case 'm': return C_Memory;
case 'I': case 'J': case 'K': case 'L': case 'M': return C_Other;
default:
break;
}
}
return TargetLowering::getConstraintType(Constraint);
}
TargetLowering::ConstraintWeight SystemZTargetLowering::
getSingleConstraintMatchWeight(AsmOperandInfo &info,
const char *constraint) const {
ConstraintWeight weight = CW_Invalid;
Value *CallOperandVal = info.CallOperandVal;
if (!CallOperandVal)
return CW_Default;
Type *type = CallOperandVal->getType();
switch (*constraint) {
default:
weight = TargetLowering::getSingleConstraintMatchWeight(info, constraint);
break;
case 'a': case 'd': case 'h': case 'r': if (CallOperandVal->getType()->isIntegerTy())
weight = CW_Register;
break;
case 'f': if (type->isFloatingPointTy())
weight = CW_Register;
break;
case 'I': if (auto *C = dyn_cast<ConstantInt>(CallOperandVal))
if (isUInt<8>(C->getZExtValue()))
weight = CW_Constant;
break;
case 'J': if (auto *C = dyn_cast<ConstantInt>(CallOperandVal))
if (isUInt<12>(C->getZExtValue()))
weight = CW_Constant;
break;
case 'K': if (auto *C = dyn_cast<ConstantInt>(CallOperandVal))
if (isInt<16>(C->getSExtValue()))
weight = CW_Constant;
break;
case 'L': if (auto *C = dyn_cast<ConstantInt>(CallOperandVal))
if (isInt<20>(C->getSExtValue()))
weight = CW_Constant;
break;
case 'M': if (auto *C = dyn_cast<ConstantInt>(CallOperandVal))
if (C->getZExtValue() == 0x7fffffff)
weight = CW_Constant;
break;
}
return weight;
}
static std::pair<unsigned, const TargetRegisterClass *>
parseRegisterNumber(const std::string &Constraint,
const TargetRegisterClass *RC, const unsigned *Map) {
assert(*(Constraint.end()-1) == '}' && "Missing '}'");
if (isdigit(Constraint[2])) {
std::string Suffix(Constraint.data() + 2, Constraint.size() - 2);
unsigned Index = atoi(Suffix.c_str());
if (Index < 16 && Map[Index])
return std::make_pair(Map[Index], RC);
}
return std::make_pair(0U, nullptr);
}
std::pair<unsigned, const TargetRegisterClass *>
SystemZTargetLowering::getRegForInlineAsmConstraint(
const TargetRegisterInfo *TRI, const std::string &Constraint,
MVT VT) const {
if (Constraint.size() == 1) {
switch (Constraint[0]) {
default: break;
case 'd': case 'r': if (VT == MVT::i64)
return std::make_pair(0U, &SystemZ::GR64BitRegClass);
else if (VT == MVT::i128)
return std::make_pair(0U, &SystemZ::GR128BitRegClass);
return std::make_pair(0U, &SystemZ::GR32BitRegClass);
case 'a': if (VT == MVT::i64)
return std::make_pair(0U, &SystemZ::ADDR64BitRegClass);
else if (VT == MVT::i128)
return std::make_pair(0U, &SystemZ::ADDR128BitRegClass);
return std::make_pair(0U, &SystemZ::ADDR32BitRegClass);
case 'h': return std::make_pair(0U, &SystemZ::GRH32BitRegClass);
case 'f': if (VT == MVT::f64)
return std::make_pair(0U, &SystemZ::FP64BitRegClass);
else if (VT == MVT::f128)
return std::make_pair(0U, &SystemZ::FP128BitRegClass);
return std::make_pair(0U, &SystemZ::FP32BitRegClass);
}
}
if (Constraint[0] == '{') {
if (Constraint[1] == 'r') {
if (VT == MVT::i32)
return parseRegisterNumber(Constraint, &SystemZ::GR32BitRegClass,
SystemZMC::GR32Regs);
if (VT == MVT::i128)
return parseRegisterNumber(Constraint, &SystemZ::GR128BitRegClass,
SystemZMC::GR128Regs);
return parseRegisterNumber(Constraint, &SystemZ::GR64BitRegClass,
SystemZMC::GR64Regs);
}
if (Constraint[1] == 'f') {
if (VT == MVT::f32)
return parseRegisterNumber(Constraint, &SystemZ::FP32BitRegClass,
SystemZMC::FP32Regs);
if (VT == MVT::f128)
return parseRegisterNumber(Constraint, &SystemZ::FP128BitRegClass,
SystemZMC::FP128Regs);
return parseRegisterNumber(Constraint, &SystemZ::FP64BitRegClass,
SystemZMC::FP64Regs);
}
}
return TargetLowering::getRegForInlineAsmConstraint(TRI, Constraint, VT);
}
void SystemZTargetLowering::
LowerAsmOperandForConstraint(SDValue Op, std::string &Constraint,
std::vector<SDValue> &Ops,
SelectionDAG &DAG) const {
if (Constraint.length() == 1) {
switch (Constraint[0]) {
case 'I': if (auto *C = dyn_cast<ConstantSDNode>(Op))
if (isUInt<8>(C->getZExtValue()))
Ops.push_back(DAG.getTargetConstant(C->getZExtValue(),
Op.getValueType()));
return;
case 'J': if (auto *C = dyn_cast<ConstantSDNode>(Op))
if (isUInt<12>(C->getZExtValue()))
Ops.push_back(DAG.getTargetConstant(C->getZExtValue(),
Op.getValueType()));
return;
case 'K': if (auto *C = dyn_cast<ConstantSDNode>(Op))
if (isInt<16>(C->getSExtValue()))
Ops.push_back(DAG.getTargetConstant(C->getSExtValue(),
Op.getValueType()));
return;
case 'L': if (auto *C = dyn_cast<ConstantSDNode>(Op))
if (isInt<20>(C->getSExtValue()))
Ops.push_back(DAG.getTargetConstant(C->getSExtValue(),
Op.getValueType()));
return;
case 'M': if (auto *C = dyn_cast<ConstantSDNode>(Op))
if (C->getZExtValue() == 0x7fffffff)
Ops.push_back(DAG.getTargetConstant(C->getZExtValue(),
Op.getValueType()));
return;
}
}
TargetLowering::LowerAsmOperandForConstraint(Op, Constraint, Ops, DAG);
}
#include "SystemZGenCallingConv.inc"
bool SystemZTargetLowering::allowTruncateForTailCall(Type *FromType,
Type *ToType) const {
return isTruncateFree(FromType, ToType);
}
bool SystemZTargetLowering::mayBeEmittedAsTailCall(CallInst *CI) const {
if (!CI->isTailCall())
return false;
return true;
}
static SDValue convertLocVTToValVT(SelectionDAG &DAG, SDLoc DL,
CCValAssign &VA, SDValue Chain,
SDValue Value) {
if (VA.getLocInfo() == CCValAssign::SExt)
Value = DAG.getNode(ISD::AssertSext, DL, VA.getLocVT(), Value,
DAG.getValueType(VA.getValVT()));
else if (VA.getLocInfo() == CCValAssign::ZExt)
Value = DAG.getNode(ISD::AssertZext, DL, VA.getLocVT(), Value,
DAG.getValueType(VA.getValVT()));
if (VA.isExtInLoc())
Value = DAG.getNode(ISD::TRUNCATE, DL, VA.getValVT(), Value);
else if (VA.getLocInfo() == CCValAssign::Indirect)
Value = DAG.getLoad(VA.getValVT(), DL, Chain, Value,
MachinePointerInfo(), false, false, false, 0);
else
assert(VA.getLocInfo() == CCValAssign::Full && "Unsupported getLocInfo");
return Value;
}
static SDValue convertValVTToLocVT(SelectionDAG &DAG, SDLoc DL,
CCValAssign &VA, SDValue Value) {
switch (VA.getLocInfo()) {
case CCValAssign::SExt:
return DAG.getNode(ISD::SIGN_EXTEND, DL, VA.getLocVT(), Value);
case CCValAssign::ZExt:
return DAG.getNode(ISD::ZERO_EXTEND, DL, VA.getLocVT(), Value);
case CCValAssign::AExt:
return DAG.getNode(ISD::ANY_EXTEND, DL, VA.getLocVT(), Value);
case CCValAssign::Full:
return Value;
default:
llvm_unreachable("Unhandled getLocInfo()");
}
}
SDValue SystemZTargetLowering::
LowerFormalArguments(SDValue Chain, CallingConv::ID CallConv, bool IsVarArg,
const SmallVectorImpl<ISD::InputArg> &Ins,
SDLoc DL, SelectionDAG &DAG,
SmallVectorImpl<SDValue> &InVals) const {
MachineFunction &MF = DAG.getMachineFunction();
MachineFrameInfo *MFI = MF.getFrameInfo();
MachineRegisterInfo &MRI = MF.getRegInfo();
SystemZMachineFunctionInfo *FuncInfo =
MF.getInfo<SystemZMachineFunctionInfo>();
auto *TFL =
static_cast<const SystemZFrameLowering *>(Subtarget.getFrameLowering());
SmallVector<CCValAssign, 16> ArgLocs;
CCState CCInfo(CallConv, IsVarArg, MF, ArgLocs, *DAG.getContext());
CCInfo.AnalyzeFormalArguments(Ins, CC_SystemZ);
unsigned NumFixedGPRs = 0;
unsigned NumFixedFPRs = 0;
for (unsigned I = 0, E = ArgLocs.size(); I != E; ++I) {
SDValue ArgValue;
CCValAssign &VA = ArgLocs[I];
EVT LocVT = VA.getLocVT();
if (VA.isRegLoc()) {
const TargetRegisterClass *RC;
switch (LocVT.getSimpleVT().SimpleTy) {
default:
llvm_unreachable("Unexpected argument type");
case MVT::i32:
NumFixedGPRs += 1;
RC = &SystemZ::GR32BitRegClass;
break;
case MVT::i64:
NumFixedGPRs += 1;
RC = &SystemZ::GR64BitRegClass;
break;
case MVT::f32:
NumFixedFPRs += 1;
RC = &SystemZ::FP32BitRegClass;
break;
case MVT::f64:
NumFixedFPRs += 1;
RC = &SystemZ::FP64BitRegClass;
break;
}
unsigned VReg = MRI.createVirtualRegister(RC);
MRI.addLiveIn(VA.getLocReg(), VReg);
ArgValue = DAG.getCopyFromReg(Chain, DL, VReg, LocVT);
} else {
assert(VA.isMemLoc() && "Argument not register or memory");
int FI = MFI->CreateFixedObject(LocVT.getSizeInBits() / 8,
VA.getLocMemOffset(), true);
EVT PtrVT = getPointerTy();
SDValue FIN = DAG.getFrameIndex(FI, PtrVT);
if (VA.getLocVT() == MVT::i32 || VA.getLocVT() == MVT::f32)
FIN = DAG.getNode(ISD::ADD, DL, PtrVT, FIN, DAG.getIntPtrConstant(4));
ArgValue = DAG.getLoad(LocVT, DL, Chain, FIN,
MachinePointerInfo::getFixedStack(FI),
false, false, false, 0);
}
InVals.push_back(convertLocVTToValVT(DAG, DL, VA, Chain, ArgValue));
}
if (IsVarArg) {
FuncInfo->setVarArgsFirstGPR(NumFixedGPRs);
FuncInfo->setVarArgsFirstFPR(NumFixedFPRs);
int64_t StackSize = CCInfo.getNextStackOffset();
FuncInfo->setVarArgsFrameIndex(MFI->CreateFixedObject(1, StackSize, true));
int64_t RegSaveOffset = TFL->getOffsetOfLocalArea();
unsigned RegSaveIndex = MFI->CreateFixedObject(1, RegSaveOffset, true);
FuncInfo->setRegSaveFrameIndex(RegSaveIndex);
if (NumFixedFPRs < SystemZ::NumArgFPRs) {
SDValue MemOps[SystemZ::NumArgFPRs];
for (unsigned I = NumFixedFPRs; I < SystemZ::NumArgFPRs; ++I) {
unsigned Offset = TFL->getRegSpillOffset(SystemZ::ArgFPRs[I]);
int FI = MFI->CreateFixedObject(8, RegSaveOffset + Offset, true);
SDValue FIN = DAG.getFrameIndex(FI, getPointerTy());
unsigned VReg = MF.addLiveIn(SystemZ::ArgFPRs[I],
&SystemZ::FP64BitRegClass);
SDValue ArgValue = DAG.getCopyFromReg(Chain, DL, VReg, MVT::f64);
MemOps[I] = DAG.getStore(ArgValue.getValue(1), DL, ArgValue, FIN,
MachinePointerInfo::getFixedStack(FI),
false, false, 0);
}
Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other,
makeArrayRef(&MemOps[NumFixedFPRs],
SystemZ::NumArgFPRs-NumFixedFPRs));
}
}
return Chain;
}
static bool canUseSiblingCall(const CCState &ArgCCInfo,
SmallVectorImpl<CCValAssign> &ArgLocs) {
for (unsigned I = 0, E = ArgLocs.size(); I != E; ++I) {
CCValAssign &VA = ArgLocs[I];
if (VA.getLocInfo() == CCValAssign::Indirect)
return false;
if (!VA.isRegLoc())
return false;
unsigned Reg = VA.getLocReg();
if (Reg == SystemZ::R6H || Reg == SystemZ::R6L || Reg == SystemZ::R6D)
return false;
}
return true;
}
SDValue
SystemZTargetLowering::LowerCall(CallLoweringInfo &CLI,
SmallVectorImpl<SDValue> &InVals) const {
SelectionDAG &DAG = CLI.DAG;
SDLoc &DL = CLI.DL;
SmallVectorImpl<ISD::OutputArg> &Outs = CLI.Outs;
SmallVectorImpl<SDValue> &OutVals = CLI.OutVals;
SmallVectorImpl<ISD::InputArg> &Ins = CLI.Ins;
SDValue Chain = CLI.Chain;
SDValue Callee = CLI.Callee;
bool &IsTailCall = CLI.IsTailCall;
CallingConv::ID CallConv = CLI.CallConv;
bool IsVarArg = CLI.IsVarArg;
MachineFunction &MF = DAG.getMachineFunction();
EVT PtrVT = getPointerTy();
SmallVector<CCValAssign, 16> ArgLocs;
CCState ArgCCInfo(CallConv, IsVarArg, MF, ArgLocs, *DAG.getContext());
ArgCCInfo.AnalyzeCallOperands(Outs, CC_SystemZ);
if (IsTailCall && !canUseSiblingCall(ArgCCInfo, ArgLocs))
IsTailCall = false;
unsigned NumBytes = ArgCCInfo.getNextStackOffset();
if (!IsTailCall)
Chain = DAG.getCALLSEQ_START(Chain, DAG.getConstant(NumBytes, PtrVT, true),
DL);
SmallVector<std::pair<unsigned, SDValue>, 9> RegsToPass;
SmallVector<SDValue, 8> MemOpChains;
SDValue StackPtr;
for (unsigned I = 0, E = ArgLocs.size(); I != E; ++I) {
CCValAssign &VA = ArgLocs[I];
SDValue ArgValue = OutVals[I];
if (VA.getLocInfo() == CCValAssign::Indirect) {
SDValue SpillSlot = DAG.CreateStackTemporary(VA.getValVT());
int FI = cast<FrameIndexSDNode>(SpillSlot)->getIndex();
MemOpChains.push_back(DAG.getStore(Chain, DL, ArgValue, SpillSlot,
MachinePointerInfo::getFixedStack(FI),
false, false, 0));
ArgValue = SpillSlot;
} else
ArgValue = convertValVTToLocVT(DAG, DL, VA, ArgValue);
if (VA.isRegLoc())
RegsToPass.push_back(std::make_pair(VA.getLocReg(), ArgValue));
else {
assert(VA.isMemLoc() && "Argument not register or memory");
if (!StackPtr.getNode())
StackPtr = DAG.getCopyFromReg(Chain, DL, SystemZ::R15D, PtrVT);
unsigned Offset = SystemZMC::CallFrameSize + VA.getLocMemOffset();
if (VA.getLocVT() == MVT::i32 || VA.getLocVT() == MVT::f32)
Offset += 4;
SDValue Address = DAG.getNode(ISD::ADD, DL, PtrVT, StackPtr,
DAG.getIntPtrConstant(Offset));
MemOpChains.push_back(DAG.getStore(Chain, DL, ArgValue, Address,
MachinePointerInfo(),
false, false, 0));
}
}
if (!MemOpChains.empty())
Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other, MemOpChains);
SDValue Glue;
if (auto *G = dyn_cast<GlobalAddressSDNode>(Callee)) {
Callee = DAG.getTargetGlobalAddress(G->getGlobal(), DL, PtrVT);
Callee = DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Callee);
} else if (auto *E = dyn_cast<ExternalSymbolSDNode>(Callee)) {
Callee = DAG.getTargetExternalSymbol(E->getSymbol(), PtrVT);
Callee = DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Callee);
} else if (IsTailCall) {
Chain = DAG.getCopyToReg(Chain, DL, SystemZ::R1D, Callee, Glue);
Glue = Chain.getValue(1);
Callee = DAG.getRegister(SystemZ::R1D, Callee.getValueType());
}
for (unsigned I = 0, E = RegsToPass.size(); I != E; ++I) {
Chain = DAG.getCopyToReg(Chain, DL, RegsToPass[I].first,
RegsToPass[I].second, Glue);
Glue = Chain.getValue(1);
}
SmallVector<SDValue, 8> Ops;
Ops.push_back(Chain);
Ops.push_back(Callee);
for (unsigned I = 0, E = RegsToPass.size(); I != E; ++I)
Ops.push_back(DAG.getRegister(RegsToPass[I].first,
RegsToPass[I].second.getValueType()));
const TargetRegisterInfo *TRI = Subtarget.getRegisterInfo();
const uint32_t *Mask = TRI->getCallPreservedMask(MF, CallConv);
assert(Mask && "Missing call preserved mask for calling convention");
Ops.push_back(DAG.getRegisterMask(Mask));
if (Glue.getNode())
Ops.push_back(Glue);
SDVTList NodeTys = DAG.getVTList(MVT::Other, MVT::Glue);
if (IsTailCall)
return DAG.getNode(SystemZISD::SIBCALL, DL, NodeTys, Ops);
Chain = DAG.getNode(SystemZISD::CALL, DL, NodeTys, Ops);
Glue = Chain.getValue(1);
Chain = DAG.getCALLSEQ_END(Chain,
DAG.getConstant(NumBytes, PtrVT, true),
DAG.getConstant(0, PtrVT, true),
Glue, DL);
Glue = Chain.getValue(1);
SmallVector<CCValAssign, 16> RetLocs;
CCState RetCCInfo(CallConv, IsVarArg, MF, RetLocs, *DAG.getContext());
RetCCInfo.AnalyzeCallResult(Ins, RetCC_SystemZ);
for (unsigned I = 0, E = RetLocs.size(); I != E; ++I) {
CCValAssign &VA = RetLocs[I];
SDValue RetValue = DAG.getCopyFromReg(Chain, DL, VA.getLocReg(),
VA.getLocVT(), Glue);
Chain = RetValue.getValue(1);
Glue = RetValue.getValue(2);
InVals.push_back(convertLocVTToValVT(DAG, DL, VA, Chain, RetValue));
}
return Chain;
}
SDValue
SystemZTargetLowering::LowerReturn(SDValue Chain,
CallingConv::ID CallConv, bool IsVarArg,
const SmallVectorImpl<ISD::OutputArg> &Outs,
const SmallVectorImpl<SDValue> &OutVals,
SDLoc DL, SelectionDAG &DAG) const {
MachineFunction &MF = DAG.getMachineFunction();
SmallVector<CCValAssign, 16> RetLocs;
CCState RetCCInfo(CallConv, IsVarArg, MF, RetLocs, *DAG.getContext());
RetCCInfo.AnalyzeReturn(Outs, RetCC_SystemZ);
if (RetLocs.empty())
return DAG.getNode(SystemZISD::RET_FLAG, DL, MVT::Other, Chain);
SDValue Glue;
SmallVector<SDValue, 4> RetOps;
RetOps.push_back(Chain);
for (unsigned I = 0, E = RetLocs.size(); I != E; ++I) {
CCValAssign &VA = RetLocs[I];
SDValue RetValue = OutVals[I];
assert(VA.isRegLoc() && "Can only return in registers!");
RetValue = convertValVTToLocVT(DAG, DL, VA, RetValue);
unsigned Reg = VA.getLocReg();
Chain = DAG.getCopyToReg(Chain, DL, Reg, RetValue, Glue);
Glue = Chain.getValue(1);
RetOps.push_back(DAG.getRegister(Reg, VA.getLocVT()));
}
RetOps[0] = Chain;
if (Glue.getNode())
RetOps.push_back(Glue);
return DAG.getNode(SystemZISD::RET_FLAG, DL, MVT::Other, RetOps);
}
SDValue SystemZTargetLowering::
prepareVolatileOrAtomicLoad(SDValue Chain, SDLoc DL, SelectionDAG &DAG) const {
return DAG.getNode(SystemZISD::SERIALIZE, DL, MVT::Other, Chain);
}
static unsigned CCMaskForCondCode(ISD::CondCode CC) {
#define CONV(X) \
case ISD::SET##X: return SystemZ::CCMASK_CMP_##X; \
case ISD::SETO##X: return SystemZ::CCMASK_CMP_##X; \
case ISD::SETU##X: return SystemZ::CCMASK_CMP_UO | SystemZ::CCMASK_CMP_##X
switch (CC) {
default:
llvm_unreachable("Invalid integer condition!");
CONV(EQ);
CONV(NE);
CONV(GT);
CONV(GE);
CONV(LT);
CONV(LE);
case ISD::SETO: return SystemZ::CCMASK_CMP_O;
case ISD::SETUO: return SystemZ::CCMASK_CMP_UO;
}
#undef CONV
}
static IPMConversion getIPMConversion(unsigned CCValid, unsigned CCMask) {
if (CCMask == (CCValid & (SystemZ::CCMASK_1 | SystemZ::CCMASK_3)))
return IPMConversion(0, 0, SystemZ::IPM_CC);
if (CCMask == (CCValid & (SystemZ::CCMASK_2 | SystemZ::CCMASK_3)))
return IPMConversion(0, 0, SystemZ::IPM_CC + 1);
uint64_t TopBit = uint64_t(1) << 31;
if (CCMask == (CCValid & SystemZ::CCMASK_0))
return IPMConversion(0, -(1 << SystemZ::IPM_CC), 31);
if (CCMask == (CCValid & (SystemZ::CCMASK_0 | SystemZ::CCMASK_1)))
return IPMConversion(0, -(2 << SystemZ::IPM_CC), 31);
if (CCMask == (CCValid & (SystemZ::CCMASK_0
| SystemZ::CCMASK_1
| SystemZ::CCMASK_2)))
return IPMConversion(0, -(3 << SystemZ::IPM_CC), 31);
if (CCMask == (CCValid & SystemZ::CCMASK_3))
return IPMConversion(0, TopBit - (3 << SystemZ::IPM_CC), 31);
if (CCMask == (CCValid & (SystemZ::CCMASK_1
| SystemZ::CCMASK_2
| SystemZ::CCMASK_3)))
return IPMConversion(0, TopBit - (1 << SystemZ::IPM_CC), 31);
if (CCMask == (CCValid & (SystemZ::CCMASK_0 | SystemZ::CCMASK_2)))
return IPMConversion(-1, 0, SystemZ::IPM_CC);
if (CCMask == (CCValid & (SystemZ::CCMASK_1 | SystemZ::CCMASK_2)))
return IPMConversion(0, 1 << SystemZ::IPM_CC, SystemZ::IPM_CC + 1);
if (CCMask == (CCValid & (SystemZ::CCMASK_0 | SystemZ::CCMASK_3)))
return IPMConversion(0, -(1 << SystemZ::IPM_CC), SystemZ::IPM_CC + 1);
if (CCMask == (CCValid & SystemZ::CCMASK_1))
return IPMConversion(1 << SystemZ::IPM_CC, -(1 << SystemZ::IPM_CC), 31);
if (CCMask == (CCValid & SystemZ::CCMASK_2))
return IPMConversion(1 << SystemZ::IPM_CC,
TopBit - (3 << SystemZ::IPM_CC), 31);
if (CCMask == (CCValid & (SystemZ::CCMASK_0
| SystemZ::CCMASK_1
| SystemZ::CCMASK_3)))
return IPMConversion(1 << SystemZ::IPM_CC, -(3 << SystemZ::IPM_CC), 31);
if (CCMask == (CCValid & (SystemZ::CCMASK_0
| SystemZ::CCMASK_2
| SystemZ::CCMASK_3)))
return IPMConversion(1 << SystemZ::IPM_CC,
TopBit - (1 << SystemZ::IPM_CC), 31);
llvm_unreachable("Unexpected CC combination");
}
static void adjustZeroCmp(SelectionDAG &DAG, Comparison &C) {
if (C.ICmpType == SystemZICMP::UnsignedOnly)
return;
auto *ConstOp1 = dyn_cast<ConstantSDNode>(C.Op1.getNode());
if (!ConstOp1)
return;
int64_t Value = ConstOp1->getSExtValue();
if ((Value == -1 && C.CCMask == SystemZ::CCMASK_CMP_GT) ||
(Value == -1 && C.CCMask == SystemZ::CCMASK_CMP_LE) ||
(Value == 1 && C.CCMask == SystemZ::CCMASK_CMP_LT) ||
(Value == 1 && C.CCMask == SystemZ::CCMASK_CMP_GE)) {
C.CCMask ^= SystemZ::CCMASK_CMP_EQ;
C.Op1 = DAG.getConstant(0, C.Op1.getValueType());
}
}
static void adjustSubwordCmp(SelectionDAG &DAG, Comparison &C) {
if (!C.Op0.hasOneUse() ||
C.Op0.getOpcode() != ISD::LOAD ||
C.Op1.getOpcode() != ISD::Constant)
return;
auto *Load = cast<LoadSDNode>(C.Op0);
unsigned NumBits = Load->getMemoryVT().getStoreSizeInBits();
if (NumBits != 8 && NumBits != 16)
return;
auto *ConstOp1 = cast<ConstantSDNode>(C.Op1);
uint64_t Value = ConstOp1->getZExtValue();
uint64_t Mask = (1 << NumBits) - 1;
if (Load->getExtensionType() == ISD::SEXTLOAD) {
int64_t SignedValue = ConstOp1->getSExtValue();
if (uint64_t(SignedValue) + (uint64_t(1) << (NumBits - 1)) > Mask)
return;
if (C.ICmpType != SystemZICMP::SignedOnly) {
Value &= Mask;
} else if (NumBits == 8) {
if (Value == 0 && C.CCMask == SystemZ::CCMASK_CMP_LT)
Value = 127, C.CCMask = SystemZ::CCMASK_CMP_GT;
else if (Value == 0 && C.CCMask == SystemZ::CCMASK_CMP_GE)
Value = 128, C.CCMask = SystemZ::CCMASK_CMP_LT;
else
return;
C.ICmpType = SystemZICMP::UnsignedOnly;
}
} else if (Load->getExtensionType() == ISD::ZEXTLOAD) {
if (Value > Mask)
return;
assert(C.ICmpType == SystemZICMP::Any &&
"Signedness shouldn't matter here.");
} else
return;
ISD::LoadExtType ExtType = (C.ICmpType == SystemZICMP::SignedOnly ?
ISD::SEXTLOAD :
ISD::ZEXTLOAD);
if (C.Op0.getValueType() != MVT::i32 ||
Load->getExtensionType() != ExtType)
C.Op0 = DAG.getExtLoad(ExtType, SDLoc(Load), MVT::i32,
Load->getChain(), Load->getBasePtr(),
Load->getPointerInfo(), Load->getMemoryVT(),
Load->isVolatile(), Load->isNonTemporal(),
Load->isInvariant(), Load->getAlignment());
if (C.Op1.getValueType() != MVT::i32 ||
Value != ConstOp1->getZExtValue())
C.Op1 = DAG.getConstant(Value, MVT::i32);
}
static bool isNaturalMemoryOperand(SDValue Op, unsigned ICmpType) {
auto *Load = dyn_cast<LoadSDNode>(Op.getNode());
if (Load) {
if (Load->getMemoryVT() == MVT::i8)
return false;
switch (Load->getExtensionType()) {
case ISD::NON_EXTLOAD:
return true;
case ISD::SEXTLOAD:
return ICmpType != SystemZICMP::UnsignedOnly;
case ISD::ZEXTLOAD:
return ICmpType != SystemZICMP::SignedOnly;
default:
break;
}
}
return false;
}
static bool shouldSwapCmpOperands(const Comparison &C) {
if (C.Op0.getValueType() == MVT::f128)
return false;
if (isa<ConstantFPSDNode>(C.Op1))
return false;
auto *ConstOp1 = dyn_cast<ConstantSDNode>(C.Op1);
if (ConstOp1 && ConstOp1->getZExtValue() == 0)
return false;
if (isNaturalMemoryOperand(C.Op1, C.ICmpType) && C.Op1.hasOneUse())
return false;
if (isNaturalMemoryOperand(C.Op0, C.ICmpType) && C.Op0.hasOneUse()) {
if (!ConstOp1)
return true;
if (C.ICmpType != SystemZICMP::SignedOnly &&
isUInt<16>(ConstOp1->getZExtValue()))
return false;
if (C.ICmpType != SystemZICMP::UnsignedOnly &&
isInt<16>(ConstOp1->getSExtValue()))
return false;
return true;
}
unsigned Opcode0 = C.Op0.getOpcode();
if (C.ICmpType != SystemZICMP::UnsignedOnly && Opcode0 == ISD::SIGN_EXTEND)
return true;
if (C.ICmpType != SystemZICMP::SignedOnly && Opcode0 == ISD::ZERO_EXTEND)
return true;
if (C.ICmpType != SystemZICMP::SignedOnly &&
Opcode0 == ISD::AND &&
C.Op0.getOperand(1).getOpcode() == ISD::Constant &&
cast<ConstantSDNode>(C.Op0.getOperand(1))->getZExtValue() == 0xffffffff)
return true;
return false;
}
static unsigned reverseCCMask(unsigned CCMask) {
return ((CCMask & SystemZ::CCMASK_CMP_EQ) |
(CCMask & SystemZ::CCMASK_CMP_GT ? SystemZ::CCMASK_CMP_LT : 0) |
(CCMask & SystemZ::CCMASK_CMP_LT ? SystemZ::CCMASK_CMP_GT : 0) |
(CCMask & SystemZ::CCMASK_CMP_UO));
}
static void adjustForSubtraction(SelectionDAG &DAG, Comparison &C) {
if (C.CCMask == SystemZ::CCMASK_CMP_EQ ||
C.CCMask == SystemZ::CCMASK_CMP_NE) {
for (auto I = C.Op0->use_begin(), E = C.Op0->use_end(); I != E; ++I) {
SDNode *N = *I;
if (N->getOpcode() == ISD::SUB &&
((N->getOperand(0) == C.Op0 && N->getOperand(1) == C.Op1) ||
(N->getOperand(0) == C.Op1 && N->getOperand(1) == C.Op0))) {
C.Op0 = SDValue(N, 0);
C.Op1 = DAG.getConstant(0, N->getValueType(0));
return;
}
}
}
}
static void adjustForFNeg(Comparison &C) {
auto *C1 = dyn_cast<ConstantFPSDNode>(C.Op1);
if (C1 && C1->isZero()) {
for (auto I = C.Op0->use_begin(), E = C.Op0->use_end(); I != E; ++I) {
SDNode *N = *I;
if (N->getOpcode() == ISD::FNEG) {
C.Op0 = SDValue(N, 0);
C.CCMask = reverseCCMask(C.CCMask);
return;
}
}
}
}
static void adjustForLTGFR(Comparison &C) {
if (C.Op0.getOpcode() == ISD::SHL &&
C.Op0.getValueType() == MVT::i64 &&
C.Op1.getOpcode() == ISD::Constant &&
cast<ConstantSDNode>(C.Op1)->getZExtValue() == 0) {
auto *C1 = dyn_cast<ConstantSDNode>(C.Op0.getOperand(1));
if (C1 && C1->getZExtValue() == 32) {
SDValue ShlOp0 = C.Op0.getOperand(0);
for (auto I = ShlOp0->use_begin(), E = ShlOp0->use_end(); I != E; ++I) {
SDNode *N = *I;
if (N->getOpcode() == ISD::SIGN_EXTEND_INREG &&
cast<VTSDNode>(N->getOperand(1))->getVT() == MVT::i32) {
C.Op0 = SDValue(N, 0);
return;
}
}
}
}
}
static void adjustICmpTruncate(SelectionDAG &DAG, Comparison &C) {
if (C.Op0.getOpcode() == ISD::TRUNCATE &&
C.Op0.getOperand(0).getOpcode() == ISD::LOAD &&
C.Op1.getOpcode() == ISD::Constant &&
cast<ConstantSDNode>(C.Op1)->getZExtValue() == 0) {
auto *L = cast<LoadSDNode>(C.Op0.getOperand(0));
if (L->getMemoryVT().getStoreSizeInBits()
<= C.Op0.getValueType().getSizeInBits()) {
unsigned Type = L->getExtensionType();
if ((Type == ISD::ZEXTLOAD && C.ICmpType != SystemZICMP::SignedOnly) ||
(Type == ISD::SEXTLOAD && C.ICmpType != SystemZICMP::UnsignedOnly)) {
C.Op0 = C.Op0.getOperand(0);
C.Op1 = DAG.getConstant(0, C.Op0.getValueType());
}
}
}
}
static bool isSimpleShift(SDValue N, unsigned &ShiftVal) {
auto *Shift = dyn_cast<ConstantSDNode>(N.getOperand(1));
if (!Shift)
return false;
uint64_t Amount = Shift->getZExtValue();
if (Amount >= N.getValueType().getSizeInBits())
return false;
ShiftVal = Amount;
return true;
}
static unsigned getTestUnderMaskCond(unsigned BitSize, unsigned CCMask,
uint64_t Mask, uint64_t CmpVal,
unsigned ICmpType) {
assert(Mask != 0 && "ANDs with zero should have been removed by now");
if (!SystemZ::isImmLL(Mask) && !SystemZ::isImmLH(Mask) &&
!SystemZ::isImmHL(Mask) && !SystemZ::isImmHH(Mask))
return 0;
unsigned HighShift = 63 - countLeadingZeros(Mask);
uint64_t High = uint64_t(1) << HighShift;
uint64_t Low = uint64_t(1) << countTrailingZeros(Mask);
bool EffectivelyUnsigned = (ICmpType != SystemZICMP::SignedOnly);
if (CmpVal == 0) {
if (CCMask == SystemZ::CCMASK_CMP_EQ)
return SystemZ::CCMASK_TM_ALL_0;
if (CCMask == SystemZ::CCMASK_CMP_NE)
return SystemZ::CCMASK_TM_SOME_1;
}
if (EffectivelyUnsigned && CmpVal <= Low) {
if (CCMask == SystemZ::CCMASK_CMP_LT)
return SystemZ::CCMASK_TM_ALL_0;
if (CCMask == SystemZ::CCMASK_CMP_GE)
return SystemZ::CCMASK_TM_SOME_1;
}
if (EffectivelyUnsigned && CmpVal < Low) {
if (CCMask == SystemZ::CCMASK_CMP_LE)
return SystemZ::CCMASK_TM_ALL_0;
if (CCMask == SystemZ::CCMASK_CMP_GT)
return SystemZ::CCMASK_TM_SOME_1;
}
if (CmpVal == Mask) {
if (CCMask == SystemZ::CCMASK_CMP_EQ)
return SystemZ::CCMASK_TM_ALL_1;
if (CCMask == SystemZ::CCMASK_CMP_NE)
return SystemZ::CCMASK_TM_SOME_0;
}
if (EffectivelyUnsigned && CmpVal >= Mask - Low && CmpVal < Mask) {
if (CCMask == SystemZ::CCMASK_CMP_GT)
return SystemZ::CCMASK_TM_ALL_1;
if (CCMask == SystemZ::CCMASK_CMP_LE)
return SystemZ::CCMASK_TM_SOME_0;
}
if (EffectivelyUnsigned && CmpVal > Mask - Low && CmpVal <= Mask) {
if (CCMask == SystemZ::CCMASK_CMP_GE)
return SystemZ::CCMASK_TM_ALL_1;
if (CCMask == SystemZ::CCMASK_CMP_LT)
return SystemZ::CCMASK_TM_SOME_0;
}
if (EffectivelyUnsigned && CmpVal >= Mask - High && CmpVal < High) {
if (CCMask == SystemZ::CCMASK_CMP_LE)
return SystemZ::CCMASK_TM_MSB_0;
if (CCMask == SystemZ::CCMASK_CMP_GT)
return SystemZ::CCMASK_TM_MSB_1;
}
if (EffectivelyUnsigned && CmpVal > Mask - High && CmpVal <= High) {
if (CCMask == SystemZ::CCMASK_CMP_LT)
return SystemZ::CCMASK_TM_MSB_0;
if (CCMask == SystemZ::CCMASK_CMP_GE)
return SystemZ::CCMASK_TM_MSB_1;
}
if (Mask == Low + High) {
if (CCMask == SystemZ::CCMASK_CMP_EQ && CmpVal == Low)
return SystemZ::CCMASK_TM_MIXED_MSB_0;
if (CCMask == SystemZ::CCMASK_CMP_NE && CmpVal == Low)
return SystemZ::CCMASK_TM_MIXED_MSB_0 ^ SystemZ::CCMASK_ANY;
if (CCMask == SystemZ::CCMASK_CMP_EQ && CmpVal == High)
return SystemZ::CCMASK_TM_MIXED_MSB_1;
if (CCMask == SystemZ::CCMASK_CMP_NE && CmpVal == High)
return SystemZ::CCMASK_TM_MIXED_MSB_1 ^ SystemZ::CCMASK_ANY;
}
return 0;
}
static void adjustForTestUnderMask(SelectionDAG &DAG, Comparison &C) {
auto *ConstOp1 = dyn_cast<ConstantSDNode>(C.Op1);
if (!ConstOp1)
return;
uint64_t CmpVal = ConstOp1->getZExtValue();
Comparison NewC(C);
uint64_t MaskVal;
ConstantSDNode *Mask = nullptr;
if (C.Op0.getOpcode() == ISD::AND) {
NewC.Op0 = C.Op0.getOperand(0);
NewC.Op1 = C.Op0.getOperand(1);
Mask = dyn_cast<ConstantSDNode>(NewC.Op1);
if (!Mask)
return;
MaskVal = Mask->getZExtValue();
} else {
if (NewC.Op0.getValueType() != MVT::i64 ||
NewC.CCMask == SystemZ::CCMASK_CMP_EQ ||
NewC.CCMask == SystemZ::CCMASK_CMP_NE ||
NewC.ICmpType == SystemZICMP::SignedOnly)
return;
if (NewC.CCMask == SystemZ::CCMASK_CMP_LE ||
NewC.CCMask == SystemZ::CCMASK_CMP_GT) {
if (CmpVal == uint64_t(-1))
return;
CmpVal += 1;
NewC.CCMask ^= SystemZ::CCMASK_CMP_EQ;
}
MaskVal = -(CmpVal & -CmpVal);
NewC.ICmpType = SystemZICMP::UnsignedOnly;
}
unsigned BitSize = NewC.Op0.getValueType().getSizeInBits();
unsigned NewCCMask, ShiftVal;
if (NewC.ICmpType != SystemZICMP::SignedOnly &&
NewC.Op0.getOpcode() == ISD::SHL &&
isSimpleShift(NewC.Op0, ShiftVal) &&
(NewCCMask = getTestUnderMaskCond(BitSize, NewC.CCMask,
MaskVal >> ShiftVal,
CmpVal >> ShiftVal,
SystemZICMP::Any))) {
NewC.Op0 = NewC.Op0.getOperand(0);
MaskVal >>= ShiftVal;
} else if (NewC.ICmpType != SystemZICMP::SignedOnly &&
NewC.Op0.getOpcode() == ISD::SRL &&
isSimpleShift(NewC.Op0, ShiftVal) &&
(NewCCMask = getTestUnderMaskCond(BitSize, NewC.CCMask,
MaskVal << ShiftVal,
CmpVal << ShiftVal,
SystemZICMP::UnsignedOnly))) {
NewC.Op0 = NewC.Op0.getOperand(0);
MaskVal <<= ShiftVal;
} else {
NewCCMask = getTestUnderMaskCond(BitSize, NewC.CCMask, MaskVal, CmpVal,
NewC.ICmpType);
if (!NewCCMask)
return;
}
C.Opcode = SystemZISD::TM;
C.Op0 = NewC.Op0;
if (Mask && Mask->getZExtValue() == MaskVal)
C.Op1 = SDValue(Mask, 0);
else
C.Op1 = DAG.getConstant(MaskVal, C.Op0.getValueType());
C.CCValid = SystemZ::CCMASK_TM;
C.CCMask = NewCCMask;
}
static Comparison getCmp(SelectionDAG &DAG, SDValue CmpOp0, SDValue CmpOp1,
ISD::CondCode Cond) {
Comparison C(CmpOp0, CmpOp1);
C.CCMask = CCMaskForCondCode(Cond);
if (C.Op0.getValueType().isFloatingPoint()) {
C.CCValid = SystemZ::CCMASK_FCMP;
C.Opcode = SystemZISD::FCMP;
adjustForFNeg(C);
} else {
C.CCValid = SystemZ::CCMASK_ICMP;
C.Opcode = SystemZISD::ICMP;
if (C.CCMask == SystemZ::CCMASK_CMP_EQ ||
C.CCMask == SystemZ::CCMASK_CMP_NE ||
(DAG.SignBitIsZero(C.Op0) && DAG.SignBitIsZero(C.Op1)))
C.ICmpType = SystemZICMP::Any;
else if (C.CCMask & SystemZ::CCMASK_CMP_UO)
C.ICmpType = SystemZICMP::UnsignedOnly;
else
C.ICmpType = SystemZICMP::SignedOnly;
C.CCMask &= ~SystemZ::CCMASK_CMP_UO;
adjustZeroCmp(DAG, C);
adjustSubwordCmp(DAG, C);
adjustForSubtraction(DAG, C);
adjustForLTGFR(C);
adjustICmpTruncate(DAG, C);
}
if (shouldSwapCmpOperands(C)) {
std::swap(C.Op0, C.Op1);
C.CCMask = reverseCCMask(C.CCMask);
}
adjustForTestUnderMask(DAG, C);
return C;
}
static SDValue emitCmp(SelectionDAG &DAG, SDLoc DL, Comparison &C) {
if (C.Opcode == SystemZISD::ICMP)
return DAG.getNode(SystemZISD::ICMP, DL, MVT::Glue, C.Op0, C.Op1,
DAG.getConstant(C.ICmpType, MVT::i32));
if (C.Opcode == SystemZISD::TM) {
bool RegisterOnly = (bool(C.CCMask & SystemZ::CCMASK_TM_MIXED_MSB_0) !=
bool(C.CCMask & SystemZ::CCMASK_TM_MIXED_MSB_1));
return DAG.getNode(SystemZISD::TM, DL, MVT::Glue, C.Op0, C.Op1,
DAG.getConstant(RegisterOnly, MVT::i32));
}
return DAG.getNode(C.Opcode, DL, MVT::Glue, C.Op0, C.Op1);
}
static void lowerMUL_LOHI32(SelectionDAG &DAG, SDLoc DL,
unsigned Extend, SDValue Op0, SDValue Op1,
SDValue &Hi, SDValue &Lo) {
Op0 = DAG.getNode(Extend, DL, MVT::i64, Op0);
Op1 = DAG.getNode(Extend, DL, MVT::i64, Op1);
SDValue Mul = DAG.getNode(ISD::MUL, DL, MVT::i64, Op0, Op1);
Hi = DAG.getNode(ISD::SRL, DL, MVT::i64, Mul, DAG.getConstant(32, MVT::i64));
Hi = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, Hi);
Lo = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, Mul);
}
static void lowerGR128Binary(SelectionDAG &DAG, SDLoc DL, EVT VT,
unsigned Extend, unsigned Opcode,
SDValue Op0, SDValue Op1,
SDValue &Even, SDValue &Odd) {
SDNode *In128 = DAG.getMachineNode(Extend, DL, MVT::Untyped, Op0);
SDValue Result = DAG.getNode(Opcode, DL, MVT::Untyped,
SDValue(In128, 0), Op1);
bool Is32Bit = is32Bit(VT);
Even = DAG.getTargetExtractSubreg(SystemZ::even128(Is32Bit), DL, VT, Result);
Odd = DAG.getTargetExtractSubreg(SystemZ::odd128(Is32Bit), DL, VT, Result);
}
static SDValue emitSETCC(SelectionDAG &DAG, SDLoc DL, SDValue Glue,
unsigned CCValid, unsigned CCMask) {
IPMConversion Conversion = getIPMConversion(CCValid, CCMask);
SDValue Result = DAG.getNode(SystemZISD::IPM, DL, MVT::i32, Glue);
if (Conversion.XORValue)
Result = DAG.getNode(ISD::XOR, DL, MVT::i32, Result,
DAG.getConstant(Conversion.XORValue, MVT::i32));
if (Conversion.AddValue)
Result = DAG.getNode(ISD::ADD, DL, MVT::i32, Result,
DAG.getConstant(Conversion.AddValue, MVT::i32));
Result = DAG.getNode(ISD::SRL, DL, MVT::i32, Result,
DAG.getConstant(Conversion.Bit, MVT::i32));
if (Conversion.Bit != 31)
Result = DAG.getNode(ISD::AND, DL, MVT::i32, Result,
DAG.getConstant(1, MVT::i32));
return Result;
}
SDValue SystemZTargetLowering::lowerSETCC(SDValue Op,
SelectionDAG &DAG) const {
SDValue CmpOp0 = Op.getOperand(0);
SDValue CmpOp1 = Op.getOperand(1);
ISD::CondCode CC = cast<CondCodeSDNode>(Op.getOperand(2))->get();
SDLoc DL(Op);
Comparison C(getCmp(DAG, CmpOp0, CmpOp1, CC));
SDValue Glue = emitCmp(DAG, DL, C);
return emitSETCC(DAG, DL, Glue, C.CCValid, C.CCMask);
}
SDValue SystemZTargetLowering::lowerBR_CC(SDValue Op, SelectionDAG &DAG) const {
SDValue Chain = Op.getOperand(0);
ISD::CondCode CC = cast<CondCodeSDNode>(Op.getOperand(1))->get();
SDValue CmpOp0 = Op.getOperand(2);
SDValue CmpOp1 = Op.getOperand(3);
SDValue Dest = Op.getOperand(4);
SDLoc DL(Op);
Comparison C(getCmp(DAG, CmpOp0, CmpOp1, CC));
SDValue Glue = emitCmp(DAG, DL, C);
return DAG.getNode(SystemZISD::BR_CCMASK, DL, Op.getValueType(),
Chain, DAG.getConstant(C.CCValid, MVT::i32),
DAG.getConstant(C.CCMask, MVT::i32), Dest, Glue);
}
static bool isAbsolute(SDValue CmpOp, SDValue Pos, SDValue Neg) {
return (Neg.getOpcode() == ISD::SUB &&
Neg.getOperand(0).getOpcode() == ISD::Constant &&
cast<ConstantSDNode>(Neg.getOperand(0))->getZExtValue() == 0 &&
Neg.getOperand(1) == Pos &&
(Pos == CmpOp ||
(Pos.getOpcode() == ISD::SIGN_EXTEND &&
Pos.getOperand(0) == CmpOp)));
}
static SDValue getAbsolute(SelectionDAG &DAG, SDLoc DL, SDValue Op,
bool IsNegative) {
Op = DAG.getNode(SystemZISD::IABS, DL, Op.getValueType(), Op);
if (IsNegative)
Op = DAG.getNode(ISD::SUB, DL, Op.getValueType(),
DAG.getConstant(0, Op.getValueType()), Op);
return Op;
}
SDValue SystemZTargetLowering::lowerSELECT_CC(SDValue Op,
SelectionDAG &DAG) const {
SDValue CmpOp0 = Op.getOperand(0);
SDValue CmpOp1 = Op.getOperand(1);
SDValue TrueOp = Op.getOperand(2);
SDValue FalseOp = Op.getOperand(3);
ISD::CondCode CC = cast<CondCodeSDNode>(Op.getOperand(4))->get();
SDLoc DL(Op);
Comparison C(getCmp(DAG, CmpOp0, CmpOp1, CC));
if (C.Opcode == SystemZISD::ICMP &&
C.CCMask != SystemZ::CCMASK_CMP_EQ &&
C.CCMask != SystemZ::CCMASK_CMP_NE &&
C.Op1.getOpcode() == ISD::Constant &&
cast<ConstantSDNode>(C.Op1)->getZExtValue() == 0) {
if (isAbsolute(C.Op0, TrueOp, FalseOp))
return getAbsolute(DAG, DL, TrueOp, C.CCMask & SystemZ::CCMASK_CMP_LT);
if (isAbsolute(C.Op0, FalseOp, TrueOp))
return getAbsolute(DAG, DL, FalseOp, C.CCMask & SystemZ::CCMASK_CMP_GT);
}
SDValue Glue = emitCmp(DAG, DL, C);
auto *TrueC = dyn_cast<ConstantSDNode>(TrueOp);
auto *FalseC = dyn_cast<ConstantSDNode>(FalseOp);
if (TrueC && FalseC) {
int64_t TrueVal = TrueC->getSExtValue();
int64_t FalseVal = FalseC->getSExtValue();
if ((TrueVal == -1 && FalseVal == 0) || (TrueVal == 0 && FalseVal == -1)) {
if (TrueVal == 0)
C.CCMask ^= C.CCValid;
SDValue Result = emitSETCC(DAG, DL, Glue, C.CCValid, C.CCMask);
EVT VT = Op.getValueType();
if (!is32Bit(VT))
Result = DAG.getNode(ISD::ANY_EXTEND, DL, VT, Result);
SDValue ShAmt = DAG.getConstant(VT.getSizeInBits() - 1, MVT::i32);
SDValue Shl = DAG.getNode(ISD::SHL, DL, VT, Result, ShAmt);
return DAG.getNode(ISD::SRA, DL, VT, Shl, ShAmt);
}
}
SmallVector<SDValue, 5> Ops;
Ops.push_back(TrueOp);
Ops.push_back(FalseOp);
Ops.push_back(DAG.getConstant(C.CCValid, MVT::i32));
Ops.push_back(DAG.getConstant(C.CCMask, MVT::i32));
Ops.push_back(Glue);
SDVTList VTs = DAG.getVTList(Op.getValueType(), MVT::Glue);
return DAG.getNode(SystemZISD::SELECT_CCMASK, DL, VTs, Ops);
}
SDValue SystemZTargetLowering::lowerGlobalAddress(GlobalAddressSDNode *Node,
SelectionDAG &DAG) const {
SDLoc DL(Node);
const GlobalValue *GV = Node->getGlobal();
int64_t Offset = Node->getOffset();
EVT PtrVT = getPointerTy();
Reloc::Model RM = DAG.getTarget().getRelocationModel();
CodeModel::Model CM = DAG.getTarget().getCodeModel();
SDValue Result;
if (Subtarget.isPC32DBLSymbol(GV, RM, CM)) {
uint64_t Anchor = Offset & ~uint64_t(0xfff);
Result = DAG.getTargetGlobalAddress(GV, DL, PtrVT, Anchor);
Result = DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Result);
Offset -= Anchor;
if (Offset != 0 && (Offset & 1) == 0) {
SDValue Full = DAG.getTargetGlobalAddress(GV, DL, PtrVT, Anchor + Offset);
Result = DAG.getNode(SystemZISD::PCREL_OFFSET, DL, PtrVT, Full, Result);
Offset = 0;
}
} else {
Result = DAG.getTargetGlobalAddress(GV, DL, PtrVT, 0, SystemZII::MO_GOT);
Result = DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Result);
Result = DAG.getLoad(PtrVT, DL, DAG.getEntryNode(), Result,
MachinePointerInfo::getGOT(), false, false, false, 0);
}
if (Offset != 0)
Result = DAG.getNode(ISD::ADD, DL, PtrVT, Result,
DAG.getConstant(Offset, PtrVT));
return Result;
}
SDValue SystemZTargetLowering::lowerTLSGetOffset(GlobalAddressSDNode *Node,
SelectionDAG &DAG,
unsigned Opcode,
SDValue GOTOffset) const {
SDLoc DL(Node);
EVT PtrVT = getPointerTy();
SDValue Chain = DAG.getEntryNode();
SDValue Glue;
SDValue GOT = DAG.getGLOBAL_OFFSET_TABLE(PtrVT);
Chain = DAG.getCopyToReg(Chain, DL, SystemZ::R12D, GOT, Glue);
Glue = Chain.getValue(1);
Chain = DAG.getCopyToReg(Chain, DL, SystemZ::R2D, GOTOffset, Glue);
Glue = Chain.getValue(1);
SmallVector<SDValue, 8> Ops;
Ops.push_back(Chain);
Ops.push_back(DAG.getTargetGlobalAddress(Node->getGlobal(), DL,
Node->getValueType(0),
0, 0));
Ops.push_back(DAG.getRegister(SystemZ::R2D, PtrVT));
Ops.push_back(DAG.getRegister(SystemZ::R12D, PtrVT));
const TargetRegisterInfo *TRI = Subtarget.getRegisterInfo();
const uint32_t *Mask =
TRI->getCallPreservedMask(DAG.getMachineFunction(), CallingConv::C);
assert(Mask && "Missing call preserved mask for calling convention");
Ops.push_back(DAG.getRegisterMask(Mask));
Ops.push_back(Glue);
SDVTList NodeTys = DAG.getVTList(MVT::Other, MVT::Glue);
Chain = DAG.getNode(Opcode, DL, NodeTys, Ops);
Glue = Chain.getValue(1);
return DAG.getCopyFromReg(Chain, DL, SystemZ::R2D, PtrVT, Glue);
}
SDValue SystemZTargetLowering::lowerGlobalTLSAddress(GlobalAddressSDNode *Node,
SelectionDAG &DAG) const {
SDLoc DL(Node);
const GlobalValue *GV = Node->getGlobal();
EVT PtrVT = getPointerTy();
TLSModel::Model model = DAG.getTarget().getTLSModel(GV);
SDValue TPHi = DAG.getNode(SystemZISD::EXTRACT_ACCESS, DL, MVT::i32,
DAG.getConstant(0, MVT::i32));
TPHi = DAG.getNode(ISD::ANY_EXTEND, DL, PtrVT, TPHi);
SDValue TPLo = DAG.getNode(SystemZISD::EXTRACT_ACCESS, DL, MVT::i32,
DAG.getConstant(1, MVT::i32));
TPLo = DAG.getNode(ISD::ZERO_EXTEND, DL, PtrVT, TPLo);
SDValue TPHiShifted = DAG.getNode(ISD::SHL, DL, PtrVT, TPHi,
DAG.getConstant(32, PtrVT));
SDValue TP = DAG.getNode(ISD::OR, DL, PtrVT, TPHiShifted, TPLo);
SDValue Offset;
switch (model) {
case TLSModel::GeneralDynamic: {
SystemZConstantPoolValue *CPV =
SystemZConstantPoolValue::Create(GV, SystemZCP::TLSGD);
Offset = DAG.getConstantPool(CPV, PtrVT, 8);
Offset = DAG.getLoad(PtrVT, DL, DAG.getEntryNode(),
Offset, MachinePointerInfo::getConstantPool(),
false, false, false, 0);
Offset = lowerTLSGetOffset(Node, DAG, SystemZISD::TLS_GDCALL, Offset);
break;
}
case TLSModel::LocalDynamic: {
SystemZConstantPoolValue *CPV =
SystemZConstantPoolValue::Create(GV, SystemZCP::TLSLDM);
Offset = DAG.getConstantPool(CPV, PtrVT, 8);
Offset = DAG.getLoad(PtrVT, DL, DAG.getEntryNode(),
Offset, MachinePointerInfo::getConstantPool(),
false, false, false, 0);
Offset = lowerTLSGetOffset(Node, DAG, SystemZISD::TLS_LDCALL, Offset);
SystemZMachineFunctionInfo* MFI =
DAG.getMachineFunction().getInfo<SystemZMachineFunctionInfo>();
MFI->incNumLocalDynamicTLSAccesses();
CPV = SystemZConstantPoolValue::Create(GV, SystemZCP::DTPOFF);
SDValue DTPOffset = DAG.getConstantPool(CPV, PtrVT, 8);
DTPOffset = DAG.getLoad(PtrVT, DL, DAG.getEntryNode(),
DTPOffset, MachinePointerInfo::getConstantPool(),
false, false, false, 0);
Offset = DAG.getNode(ISD::ADD, DL, PtrVT, Offset, DTPOffset);
break;
}
case TLSModel::InitialExec: {
Offset = DAG.getTargetGlobalAddress(GV, DL, PtrVT, 0,
SystemZII::MO_INDNTPOFF);
Offset = DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Offset);
Offset = DAG.getLoad(PtrVT, DL, DAG.getEntryNode(),
Offset, MachinePointerInfo::getGOT(),
false, false, false, 0);
break;
}
case TLSModel::LocalExec: {
SystemZConstantPoolValue *CPV =
SystemZConstantPoolValue::Create(GV, SystemZCP::NTPOFF);
Offset = DAG.getConstantPool(CPV, PtrVT, 8);
Offset = DAG.getLoad(PtrVT, DL, DAG.getEntryNode(),
Offset, MachinePointerInfo::getConstantPool(),
false, false, false, 0);
break;
}
}
return DAG.getNode(ISD::ADD, DL, PtrVT, TP, Offset);
}
SDValue SystemZTargetLowering::lowerBlockAddress(BlockAddressSDNode *Node,
SelectionDAG &DAG) const {
SDLoc DL(Node);
const BlockAddress *BA = Node->getBlockAddress();
int64_t Offset = Node->getOffset();
EVT PtrVT = getPointerTy();
SDValue Result = DAG.getTargetBlockAddress(BA, PtrVT, Offset);
Result = DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Result);
return Result;
}
SDValue SystemZTargetLowering::lowerJumpTable(JumpTableSDNode *JT,
SelectionDAG &DAG) const {
SDLoc DL(JT);
EVT PtrVT = getPointerTy();
SDValue Result = DAG.getTargetJumpTable(JT->getIndex(), PtrVT);
return DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Result);
}
SDValue SystemZTargetLowering::lowerConstantPool(ConstantPoolSDNode *CP,
SelectionDAG &DAG) const {
SDLoc DL(CP);
EVT PtrVT = getPointerTy();
SDValue Result;
if (CP->isMachineConstantPoolEntry())
Result = DAG.getTargetConstantPool(CP->getMachineCPVal(), PtrVT,
CP->getAlignment());
else
Result = DAG.getTargetConstantPool(CP->getConstVal(), PtrVT,
CP->getAlignment(), CP->getOffset());
return DAG.getNode(SystemZISD::PCREL_WRAPPER, DL, PtrVT, Result);
}
SDValue SystemZTargetLowering::lowerBITCAST(SDValue Op,
SelectionDAG &DAG) const {
SDLoc DL(Op);
SDValue In = Op.getOperand(0);
EVT InVT = In.getValueType();
EVT ResVT = Op.getValueType();
if (InVT == MVT::i32 && ResVT == MVT::f32) {
SDValue In64;
if (Subtarget.hasHighWord()) {
SDNode *U64 = DAG.getMachineNode(TargetOpcode::IMPLICIT_DEF, DL,
MVT::i64);
In64 = DAG.getTargetInsertSubreg(SystemZ::subreg_h32, DL,
MVT::i64, SDValue(U64, 0), In);
} else {
In64 = DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, In);
In64 = DAG.getNode(ISD::SHL, DL, MVT::i64, In64,
DAG.getConstant(32, MVT::i64));
}
SDValue Out64 = DAG.getNode(ISD::BITCAST, DL, MVT::f64, In64);
return DAG.getTargetExtractSubreg(SystemZ::subreg_h32,
DL, MVT::f32, Out64);
}
if (InVT == MVT::f32 && ResVT == MVT::i32) {
SDNode *U64 = DAG.getMachineNode(TargetOpcode::IMPLICIT_DEF, DL, MVT::f64);
SDValue In64 = DAG.getTargetInsertSubreg(SystemZ::subreg_h32, DL,
MVT::f64, SDValue(U64, 0), In);
SDValue Out64 = DAG.getNode(ISD::BITCAST, DL, MVT::i64, In64);
if (Subtarget.hasHighWord())
return DAG.getTargetExtractSubreg(SystemZ::subreg_h32, DL,
MVT::i32, Out64);
SDValue Shift = DAG.getNode(ISD::SRL, DL, MVT::i64, Out64,
DAG.getConstant(32, MVT::i64));
return DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, Shift);
}
llvm_unreachable("Unexpected bitcast combination");
}
SDValue SystemZTargetLowering::lowerVASTART(SDValue Op,
SelectionDAG &DAG) const {
MachineFunction &MF = DAG.getMachineFunction();
SystemZMachineFunctionInfo *FuncInfo =
MF.getInfo<SystemZMachineFunctionInfo>();
EVT PtrVT = getPointerTy();
SDValue Chain = Op.getOperand(0);
SDValue Addr = Op.getOperand(1);
const Value *SV = cast<SrcValueSDNode>(Op.getOperand(2))->getValue();
SDLoc DL(Op);
const unsigned NumFields = 4;
SDValue Fields[NumFields] = {
DAG.getConstant(FuncInfo->getVarArgsFirstGPR(), PtrVT),
DAG.getConstant(FuncInfo->getVarArgsFirstFPR(), PtrVT),
DAG.getFrameIndex(FuncInfo->getVarArgsFrameIndex(), PtrVT),
DAG.getFrameIndex(FuncInfo->getRegSaveFrameIndex(), PtrVT)
};
SDValue MemOps[NumFields];
unsigned Offset = 0;
for (unsigned I = 0; I < NumFields; ++I) {
SDValue FieldAddr = Addr;
if (Offset != 0)
FieldAddr = DAG.getNode(ISD::ADD, DL, PtrVT, FieldAddr,
DAG.getIntPtrConstant(Offset));
MemOps[I] = DAG.getStore(Chain, DL, Fields[I], FieldAddr,
MachinePointerInfo(SV, Offset),
false, false, 0);
Offset += 8;
}
return DAG.getNode(ISD::TokenFactor, DL, MVT::Other, MemOps);
}
SDValue SystemZTargetLowering::lowerVACOPY(SDValue Op,
SelectionDAG &DAG) const {
SDValue Chain = Op.getOperand(0);
SDValue DstPtr = Op.getOperand(1);
SDValue SrcPtr = Op.getOperand(2);
const Value *DstSV = cast<SrcValueSDNode>(Op.getOperand(3))->getValue();
const Value *SrcSV = cast<SrcValueSDNode>(Op.getOperand(4))->getValue();
SDLoc DL(Op);
return DAG.getMemcpy(Chain, DL, DstPtr, SrcPtr, DAG.getIntPtrConstant(32),
8, false, false,
MachinePointerInfo(DstSV), MachinePointerInfo(SrcSV));
}
SDValue SystemZTargetLowering::
lowerDYNAMIC_STACKALLOC(SDValue Op, SelectionDAG &DAG) const {
SDValue Chain = Op.getOperand(0);
SDValue Size = Op.getOperand(1);
SDLoc DL(Op);
unsigned SPReg = getStackPointerRegisterToSaveRestore();
SDValue OldSP = DAG.getCopyFromReg(Chain, DL, SPReg, MVT::i64);
SDValue NewSP = DAG.getNode(ISD::SUB, DL, MVT::i64, OldSP, Size);
Chain = DAG.getCopyToReg(Chain, DL, SPReg, NewSP);
SDValue ArgAdjust = DAG.getNode(SystemZISD::ADJDYNALLOC, DL, MVT::i64);
SDValue Result = DAG.getNode(ISD::ADD, DL, MVT::i64, NewSP, ArgAdjust);
SDValue Ops[2] = { Result, Chain };
return DAG.getMergeValues(Ops, DL);
}
SDValue SystemZTargetLowering::lowerSMUL_LOHI(SDValue Op,
SelectionDAG &DAG) const {
EVT VT = Op.getValueType();
SDLoc DL(Op);
SDValue Ops[2];
if (is32Bit(VT))
lowerMUL_LOHI32(DAG, DL, ISD::SIGN_EXTEND, Op.getOperand(0),
Op.getOperand(1), Ops[1], Ops[0]);
else {
SDValue C63 = DAG.getConstant(63, MVT::i64);
SDValue LL = Op.getOperand(0);
SDValue RL = Op.getOperand(1);
SDValue LH = DAG.getNode(ISD::SRA, DL, VT, LL, C63);
SDValue RH = DAG.getNode(ISD::SRA, DL, VT, RL, C63);
lowerGR128Binary(DAG, DL, VT, SystemZ::AEXT128_64, SystemZISD::UMUL_LOHI64,
LL, RL, Ops[1], Ops[0]);
SDValue NegLLTimesRH = DAG.getNode(ISD::AND, DL, VT, LL, RH);
SDValue NegLHTimesRL = DAG.getNode(ISD::AND, DL, VT, LH, RL);
SDValue NegSum = DAG.getNode(ISD::ADD, DL, VT, NegLLTimesRH, NegLHTimesRL);
Ops[1] = DAG.getNode(ISD::SUB, DL, VT, Ops[1], NegSum);
}
return DAG.getMergeValues(Ops, DL);
}
SDValue SystemZTargetLowering::lowerUMUL_LOHI(SDValue Op,
SelectionDAG &DAG) const {
EVT VT = Op.getValueType();
SDLoc DL(Op);
SDValue Ops[2];
if (is32Bit(VT))
lowerMUL_LOHI32(DAG, DL, ISD::ZERO_EXTEND, Op.getOperand(0),
Op.getOperand(1), Ops[1], Ops[0]);
else
lowerGR128Binary(DAG, DL, VT, SystemZ::AEXT128_64, SystemZISD::UMUL_LOHI64,
Op.getOperand(0), Op.getOperand(1), Ops[1], Ops[0]);
return DAG.getMergeValues(Ops, DL);
}
SDValue SystemZTargetLowering::lowerSDIVREM(SDValue Op,
SelectionDAG &DAG) const {
SDValue Op0 = Op.getOperand(0);
SDValue Op1 = Op.getOperand(1);
EVT VT = Op.getValueType();
SDLoc DL(Op);
unsigned Opcode;
if (is32Bit(VT)) {
Op0 = DAG.getNode(ISD::SIGN_EXTEND, DL, MVT::i64, Op0);
Opcode = SystemZISD::SDIVREM32;
} else if (DAG.ComputeNumSignBits(Op1) > 32) {
Op1 = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, Op1);
Opcode = SystemZISD::SDIVREM32;
} else
Opcode = SystemZISD::SDIVREM64;
SDValue Ops[2];
lowerGR128Binary(DAG, DL, VT, SystemZ::AEXT128_64, Opcode,
Op0, Op1, Ops[1], Ops[0]);
return DAG.getMergeValues(Ops, DL);
}
SDValue SystemZTargetLowering::lowerUDIVREM(SDValue Op,
SelectionDAG &DAG) const {
EVT VT = Op.getValueType();
SDLoc DL(Op);
SDValue Ops[2];
if (is32Bit(VT))
lowerGR128Binary(DAG, DL, VT, SystemZ::ZEXT128_32, SystemZISD::UDIVREM32,
Op.getOperand(0), Op.getOperand(1), Ops[1], Ops[0]);
else
lowerGR128Binary(DAG, DL, VT, SystemZ::ZEXT128_64, SystemZISD::UDIVREM64,
Op.getOperand(0), Op.getOperand(1), Ops[1], Ops[0]);
return DAG.getMergeValues(Ops, DL);
}
SDValue SystemZTargetLowering::lowerOR(SDValue Op, SelectionDAG &DAG) const {
assert(Op.getValueType() == MVT::i64 && "Should be 64-bit operation");
SDValue Ops[] = { Op.getOperand(0), Op.getOperand(1) };
APInt KnownZero[2], KnownOne[2];
DAG.computeKnownBits(Ops[0], KnownZero[0], KnownOne[0]);
DAG.computeKnownBits(Ops[1], KnownZero[1], KnownOne[1]);
uint64_t Masks[] = { KnownZero[0].getZExtValue(),
KnownZero[1].getZExtValue() };
unsigned High, Low;
if ((Masks[0] >> 32) == 0xffffffff && uint32_t(Masks[1]) == 0xffffffff)
High = 1, Low = 0;
else if ((Masks[1] >> 32) == 0xffffffff && uint32_t(Masks[0]) == 0xffffffff)
High = 0, Low = 1;
else
return Op;
SDValue LowOp = Ops[Low];
SDValue HighOp = Ops[High];
if (HighOp.getOpcode() == ISD::Constant)
return Op;
if (LowOp.getOpcode() == ISD::Constant) {
int64_t Value = int32_t(cast<ConstantSDNode>(LowOp)->getZExtValue());
if (!isInt<16>(Value))
return Op;
}
if (HighOp.getOpcode() == ISD::AND &&
HighOp.getOperand(1).getOpcode() == ISD::Constant) {
SDValue HighOp0 = HighOp.getOperand(0);
uint64_t Mask = cast<ConstantSDNode>(HighOp.getOperand(1))->getZExtValue();
if (DAG.MaskedValueIsZero(HighOp0, APInt(64, ~(Mask | 0xffffffff))))
HighOp = HighOp0;
}
SDLoc DL(Op);
SDValue Low32 = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, LowOp);
return DAG.getTargetInsertSubreg(SystemZ::subreg_l32, DL,
MVT::i64, HighOp, Low32);
}
SDValue SystemZTargetLowering::lowerATOMIC_LOAD(SDValue Op,
SelectionDAG &DAG) const {
auto *Node = cast<AtomicSDNode>(Op.getNode());
return DAG.getExtLoad(ISD::EXTLOAD, SDLoc(Op), Op.getValueType(),
Node->getChain(), Node->getBasePtr(),
Node->getMemoryVT(), Node->getMemOperand());
}
SDValue SystemZTargetLowering::lowerATOMIC_STORE(SDValue Op,
SelectionDAG &DAG) const {
auto *Node = cast<AtomicSDNode>(Op.getNode());
SDValue Chain = DAG.getTruncStore(Node->getChain(), SDLoc(Op), Node->getVal(),
Node->getBasePtr(), Node->getMemoryVT(),
Node->getMemOperand());
return SDValue(DAG.getMachineNode(SystemZ::Serialize, SDLoc(Op), MVT::Other,
Chain), 0);
}
SDValue SystemZTargetLowering::lowerATOMIC_LOAD_OP(SDValue Op,
SelectionDAG &DAG,
unsigned Opcode) const {
auto *Node = cast<AtomicSDNode>(Op.getNode());
EVT NarrowVT = Node->getMemoryVT();
EVT WideVT = MVT::i32;
if (NarrowVT == WideVT)
return Op;
int64_t BitSize = NarrowVT.getSizeInBits();
SDValue ChainIn = Node->getChain();
SDValue Addr = Node->getBasePtr();
SDValue Src2 = Node->getVal();
MachineMemOperand *MMO = Node->getMemOperand();
SDLoc DL(Node);
EVT PtrVT = Addr.getValueType();
if (Opcode == SystemZISD::ATOMIC_LOADW_SUB)
if (auto *Const = dyn_cast<ConstantSDNode>(Src2)) {
Opcode = SystemZISD::ATOMIC_LOADW_ADD;
Src2 = DAG.getConstant(-Const->getSExtValue(), Src2.getValueType());
}
SDValue AlignedAddr = DAG.getNode(ISD::AND, DL, PtrVT, Addr,
DAG.getConstant(-4, PtrVT));
SDValue BitShift = DAG.getNode(ISD::SHL, DL, PtrVT, Addr,
DAG.getConstant(3, PtrVT));
BitShift = DAG.getNode(ISD::TRUNCATE, DL, WideVT, BitShift);
SDValue NegBitShift = DAG.getNode(ISD::SUB, DL, WideVT,
DAG.getConstant(0, WideVT), BitShift);
if (Opcode != SystemZISD::ATOMIC_SWAPW)
Src2 = DAG.getNode(ISD::SHL, DL, WideVT, Src2,
DAG.getConstant(32 - BitSize, WideVT));
if (Opcode == SystemZISD::ATOMIC_LOADW_AND ||
Opcode == SystemZISD::ATOMIC_LOADW_NAND)
Src2 = DAG.getNode(ISD::OR, DL, WideVT, Src2,
DAG.getConstant(uint32_t(-1) >> BitSize, WideVT));
SDVTList VTList = DAG.getVTList(WideVT, MVT::Other);
SDValue Ops[] = { ChainIn, AlignedAddr, Src2, BitShift, NegBitShift,
DAG.getConstant(BitSize, WideVT) };
SDValue AtomicOp = DAG.getMemIntrinsicNode(Opcode, DL, VTList, Ops,
NarrowVT, MMO);
SDValue ResultShift = DAG.getNode(ISD::ADD, DL, WideVT, BitShift,
DAG.getConstant(BitSize, WideVT));
SDValue Result = DAG.getNode(ISD::ROTL, DL, WideVT, AtomicOp, ResultShift);
SDValue RetOps[2] = { Result, AtomicOp.getValue(1) };
return DAG.getMergeValues(RetOps, DL);
}
SDValue SystemZTargetLowering::lowerATOMIC_LOAD_SUB(SDValue Op,
SelectionDAG &DAG) const {
auto *Node = cast<AtomicSDNode>(Op.getNode());
EVT MemVT = Node->getMemoryVT();
if (MemVT == MVT::i32 || MemVT == MVT::i64) {
assert(Op.getValueType() == MemVT && "Mismatched VTs");
SDValue Src2 = Node->getVal();
SDValue NegSrc2;
SDLoc DL(Src2);
if (auto *Op2 = dyn_cast<ConstantSDNode>(Src2)) {
int64_t Value = (-Op2->getAPIntValue()).getSExtValue();
if (isInt<32>(Value) || Subtarget.hasInterlockedAccess1())
NegSrc2 = DAG.getConstant(Value, MemVT);
} else if (Subtarget.hasInterlockedAccess1())
NegSrc2 = DAG.getNode(ISD::SUB, DL, MemVT, DAG.getConstant(0, MemVT),
Src2);
if (NegSrc2.getNode())
return DAG.getAtomic(ISD::ATOMIC_LOAD_ADD, DL, MemVT,
Node->getChain(), Node->getBasePtr(), NegSrc2,
Node->getMemOperand(), Node->getOrdering(),
Node->getSynchScope());
return Op;
}
return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_SUB);
}
SDValue SystemZTargetLowering::lowerATOMIC_CMP_SWAP(SDValue Op,
SelectionDAG &DAG) const {
auto *Node = cast<AtomicSDNode>(Op.getNode());
EVT NarrowVT = Node->getMemoryVT();
EVT WideVT = MVT::i32;
if (NarrowVT == WideVT)
return Op;
int64_t BitSize = NarrowVT.getSizeInBits();
SDValue ChainIn = Node->getOperand(0);
SDValue Addr = Node->getOperand(1);
SDValue CmpVal = Node->getOperand(2);
SDValue SwapVal = Node->getOperand(3);
MachineMemOperand *MMO = Node->getMemOperand();
SDLoc DL(Node);
EVT PtrVT = Addr.getValueType();
SDValue AlignedAddr = DAG.getNode(ISD::AND, DL, PtrVT, Addr,
DAG.getConstant(-4, PtrVT));
SDValue BitShift = DAG.getNode(ISD::SHL, DL, PtrVT, Addr,
DAG.getConstant(3, PtrVT));
BitShift = DAG.getNode(ISD::TRUNCATE, DL, WideVT, BitShift);
SDValue NegBitShift = DAG.getNode(ISD::SUB, DL, WideVT,
DAG.getConstant(0, WideVT), BitShift);
SDVTList VTList = DAG.getVTList(WideVT, MVT::Other);
SDValue Ops[] = { ChainIn, AlignedAddr, CmpVal, SwapVal, BitShift,
NegBitShift, DAG.getConstant(BitSize, WideVT) };
SDValue AtomicOp = DAG.getMemIntrinsicNode(SystemZISD::ATOMIC_CMP_SWAPW, DL,
VTList, Ops, NarrowVT, MMO);
return AtomicOp;
}
SDValue SystemZTargetLowering::lowerSTACKSAVE(SDValue Op,
SelectionDAG &DAG) const {
MachineFunction &MF = DAG.getMachineFunction();
MF.getInfo<SystemZMachineFunctionInfo>()->setManipulatesSP(true);
return DAG.getCopyFromReg(Op.getOperand(0), SDLoc(Op),
SystemZ::R15D, Op.getValueType());
}
SDValue SystemZTargetLowering::lowerSTACKRESTORE(SDValue Op,
SelectionDAG &DAG) const {
MachineFunction &MF = DAG.getMachineFunction();
MF.getInfo<SystemZMachineFunctionInfo>()->setManipulatesSP(true);
return DAG.getCopyToReg(Op.getOperand(0), SDLoc(Op),
SystemZ::R15D, Op.getOperand(1));
}
SDValue SystemZTargetLowering::lowerPREFETCH(SDValue Op,
SelectionDAG &DAG) const {
bool IsData = cast<ConstantSDNode>(Op.getOperand(4))->getZExtValue();
if (!IsData)
return Op.getOperand(0);
bool IsWrite = cast<ConstantSDNode>(Op.getOperand(2))->getZExtValue();
unsigned Code = IsWrite ? SystemZ::PFD_WRITE : SystemZ::PFD_READ;
auto *Node = cast<MemIntrinsicSDNode>(Op.getNode());
SDValue Ops[] = {
Op.getOperand(0),
DAG.getConstant(Code, MVT::i32),
Op.getOperand(1)
};
return DAG.getMemIntrinsicNode(SystemZISD::PREFETCH, SDLoc(Op),
Node->getVTList(), Ops,
Node->getMemoryVT(), Node->getMemOperand());
}
SDValue SystemZTargetLowering::LowerOperation(SDValue Op,
SelectionDAG &DAG) const {
switch (Op.getOpcode()) {
case ISD::BR_CC:
return lowerBR_CC(Op, DAG);
case ISD::SELECT_CC:
return lowerSELECT_CC(Op, DAG);
case ISD::SETCC:
return lowerSETCC(Op, DAG);
case ISD::GlobalAddress:
return lowerGlobalAddress(cast<GlobalAddressSDNode>(Op), DAG);
case ISD::GlobalTLSAddress:
return lowerGlobalTLSAddress(cast<GlobalAddressSDNode>(Op), DAG);
case ISD::BlockAddress:
return lowerBlockAddress(cast<BlockAddressSDNode>(Op), DAG);
case ISD::JumpTable:
return lowerJumpTable(cast<JumpTableSDNode>(Op), DAG);
case ISD::ConstantPool:
return lowerConstantPool(cast<ConstantPoolSDNode>(Op), DAG);
case ISD::BITCAST:
return lowerBITCAST(Op, DAG);
case ISD::VASTART:
return lowerVASTART(Op, DAG);
case ISD::VACOPY:
return lowerVACOPY(Op, DAG);
case ISD::DYNAMIC_STACKALLOC:
return lowerDYNAMIC_STACKALLOC(Op, DAG);
case ISD::SMUL_LOHI:
return lowerSMUL_LOHI(Op, DAG);
case ISD::UMUL_LOHI:
return lowerUMUL_LOHI(Op, DAG);
case ISD::SDIVREM:
return lowerSDIVREM(Op, DAG);
case ISD::UDIVREM:
return lowerUDIVREM(Op, DAG);
case ISD::OR:
return lowerOR(Op, DAG);
case ISD::ATOMIC_SWAP:
return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_SWAPW);
case ISD::ATOMIC_STORE:
return lowerATOMIC_STORE(Op, DAG);
case ISD::ATOMIC_LOAD:
return lowerATOMIC_LOAD(Op, DAG);
case ISD::ATOMIC_LOAD_ADD:
return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_ADD);
case ISD::ATOMIC_LOAD_SUB:
return lowerATOMIC_LOAD_SUB(Op, DAG);
case ISD::ATOMIC_LOAD_AND:
return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_AND);
case ISD::ATOMIC_LOAD_OR:
return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_OR);
case ISD::ATOMIC_LOAD_XOR:
return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_XOR);
case ISD::ATOMIC_LOAD_NAND:
return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_NAND);
case ISD::ATOMIC_LOAD_MIN:
return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_MIN);
case ISD::ATOMIC_LOAD_MAX:
return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_MAX);
case ISD::ATOMIC_LOAD_UMIN:
return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_UMIN);
case ISD::ATOMIC_LOAD_UMAX:
return lowerATOMIC_LOAD_OP(Op, DAG, SystemZISD::ATOMIC_LOADW_UMAX);
case ISD::ATOMIC_CMP_SWAP:
return lowerATOMIC_CMP_SWAP(Op, DAG);
case ISD::STACKSAVE:
return lowerSTACKSAVE(Op, DAG);
case ISD::STACKRESTORE:
return lowerSTACKRESTORE(Op, DAG);
case ISD::PREFETCH:
return lowerPREFETCH(Op, DAG);
default:
llvm_unreachable("Unexpected node to lower");
}
}
const char *SystemZTargetLowering::getTargetNodeName(unsigned Opcode) const {
#define OPCODE(NAME) case SystemZISD::NAME: return "SystemZISD::" #NAME
switch (Opcode) {
OPCODE(RET_FLAG);
OPCODE(CALL);
OPCODE(SIBCALL);
OPCODE(PCREL_WRAPPER);
OPCODE(PCREL_OFFSET);
OPCODE(IABS);
OPCODE(ICMP);
OPCODE(FCMP);
OPCODE(TM);
OPCODE(BR_CCMASK);
OPCODE(SELECT_CCMASK);
OPCODE(ADJDYNALLOC);
OPCODE(EXTRACT_ACCESS);
OPCODE(UMUL_LOHI64);
OPCODE(SDIVREM64);
OPCODE(UDIVREM32);
OPCODE(UDIVREM64);
OPCODE(MVC);
OPCODE(MVC_LOOP);
OPCODE(NC);
OPCODE(NC_LOOP);
OPCODE(OC);
OPCODE(OC_LOOP);
OPCODE(XC);
OPCODE(XC_LOOP);
OPCODE(CLC);
OPCODE(CLC_LOOP);
OPCODE(STRCMP);
OPCODE(STPCPY);
OPCODE(SEARCH_STRING);
OPCODE(IPM);
OPCODE(SERIALIZE);
OPCODE(ATOMIC_SWAPW);
OPCODE(ATOMIC_LOADW_ADD);
OPCODE(ATOMIC_LOADW_SUB);
OPCODE(ATOMIC_LOADW_AND);
OPCODE(ATOMIC_LOADW_OR);
OPCODE(ATOMIC_LOADW_XOR);
OPCODE(ATOMIC_LOADW_NAND);
OPCODE(ATOMIC_LOADW_MIN);
OPCODE(ATOMIC_LOADW_MAX);
OPCODE(ATOMIC_LOADW_UMIN);
OPCODE(ATOMIC_LOADW_UMAX);
OPCODE(ATOMIC_CMP_SWAPW);
OPCODE(PREFETCH);
}
return nullptr;
#undef OPCODE
}
SDValue SystemZTargetLowering::PerformDAGCombine(SDNode *N,
DAGCombinerInfo &DCI) const {
SelectionDAG &DAG = DCI.DAG;
unsigned Opcode = N->getOpcode();
if (Opcode == ISD::SIGN_EXTEND) {
SDValue N0 = N->getOperand(0);
EVT VT = N->getValueType(0);
if (N0.hasOneUse() && N0.getOpcode() == ISD::SRA) {
auto *SraAmt = dyn_cast<ConstantSDNode>(N0.getOperand(1));
SDValue Inner = N0.getOperand(0);
if (SraAmt && Inner.hasOneUse() && Inner.getOpcode() == ISD::SHL) {
if (auto *ShlAmt = dyn_cast<ConstantSDNode>(Inner.getOperand(1))) {
unsigned Extra = (VT.getSizeInBits() -
N0.getValueType().getSizeInBits());
unsigned NewShlAmt = ShlAmt->getZExtValue() + Extra;
unsigned NewSraAmt = SraAmt->getZExtValue() + Extra;
EVT ShiftVT = N0.getOperand(1).getValueType();
SDValue Ext = DAG.getNode(ISD::ANY_EXTEND, SDLoc(Inner), VT,
Inner.getOperand(0));
SDValue Shl = DAG.getNode(ISD::SHL, SDLoc(Inner), VT, Ext,
DAG.getConstant(NewShlAmt, ShiftVT));
return DAG.getNode(ISD::SRA, SDLoc(N0), VT, Shl,
DAG.getConstant(NewSraAmt, ShiftVT));
}
}
}
}
return SDValue();
}
static MachineBasicBlock *emitBlockAfter(MachineBasicBlock *MBB) {
MachineFunction &MF = *MBB->getParent();
MachineBasicBlock *NewMBB = MF.CreateMachineBasicBlock(MBB->getBasicBlock());
MF.insert(std::next(MachineFunction::iterator(MBB)), NewMBB);
return NewMBB;
}
static MachineBasicBlock *splitBlockAfter(MachineInstr *MI,
MachineBasicBlock *MBB) {
MachineBasicBlock *NewMBB = emitBlockAfter(MBB);
NewMBB->splice(NewMBB->begin(), MBB,
std::next(MachineBasicBlock::iterator(MI)), MBB->end());
NewMBB->transferSuccessorsAndUpdatePHIs(MBB);
return NewMBB;
}
static MachineBasicBlock *splitBlockBefore(MachineInstr *MI,
MachineBasicBlock *MBB) {
MachineBasicBlock *NewMBB = emitBlockAfter(MBB);
NewMBB->splice(NewMBB->begin(), MBB, MI, MBB->end());
NewMBB->transferSuccessorsAndUpdatePHIs(MBB);
return NewMBB;
}
static unsigned forceReg(MachineInstr *MI, MachineOperand &Base,
const SystemZInstrInfo *TII) {
if (Base.isReg())
return Base.getReg();
MachineBasicBlock *MBB = MI->getParent();
MachineFunction &MF = *MBB->getParent();
MachineRegisterInfo &MRI = MF.getRegInfo();
unsigned Reg = MRI.createVirtualRegister(&SystemZ::ADDR64BitRegClass);
BuildMI(*MBB, MI, MI->getDebugLoc(), TII->get(SystemZ::LA), Reg)
.addOperand(Base).addImm(0).addReg(0);
return Reg;
}
MachineBasicBlock *
SystemZTargetLowering::emitSelect(MachineInstr *MI,
MachineBasicBlock *MBB) const {
const SystemZInstrInfo *TII =
static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
unsigned DestReg = MI->getOperand(0).getReg();
unsigned TrueReg = MI->getOperand(1).getReg();
unsigned FalseReg = MI->getOperand(2).getReg();
unsigned CCValid = MI->getOperand(3).getImm();
unsigned CCMask = MI->getOperand(4).getImm();
DebugLoc DL = MI->getDebugLoc();
MachineBasicBlock *StartMBB = MBB;
MachineBasicBlock *JoinMBB = splitBlockBefore(MI, MBB);
MachineBasicBlock *FalseMBB = emitBlockAfter(StartMBB);
MBB = StartMBB;
BuildMI(MBB, DL, TII->get(SystemZ::BRC))
.addImm(CCValid).addImm(CCMask).addMBB(JoinMBB);
MBB->addSuccessor(JoinMBB);
MBB->addSuccessor(FalseMBB);
MBB = FalseMBB;
MBB->addSuccessor(JoinMBB);
MBB = JoinMBB;
BuildMI(*MBB, MI, DL, TII->get(SystemZ::PHI), DestReg)
.addReg(TrueReg).addMBB(StartMBB)
.addReg(FalseReg).addMBB(FalseMBB);
MI->eraseFromParent();
return JoinMBB;
}
MachineBasicBlock *
SystemZTargetLowering::emitCondStore(MachineInstr *MI,
MachineBasicBlock *MBB,
unsigned StoreOpcode, unsigned STOCOpcode,
bool Invert) const {
const SystemZInstrInfo *TII =
static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
unsigned SrcReg = MI->getOperand(0).getReg();
MachineOperand Base = MI->getOperand(1);
int64_t Disp = MI->getOperand(2).getImm();
unsigned IndexReg = MI->getOperand(3).getReg();
unsigned CCValid = MI->getOperand(4).getImm();
unsigned CCMask = MI->getOperand(5).getImm();
DebugLoc DL = MI->getDebugLoc();
StoreOpcode = TII->getOpcodeForOffset(StoreOpcode, Disp);
if (STOCOpcode && !IndexReg && Subtarget.hasLoadStoreOnCond()) {
if (Invert)
CCMask ^= CCValid;
BuildMI(*MBB, MI, DL, TII->get(STOCOpcode))
.addReg(SrcReg).addOperand(Base).addImm(Disp)
.addImm(CCValid).addImm(CCMask);
MI->eraseFromParent();
return MBB;
}
if (!Invert)
CCMask ^= CCValid;
MachineBasicBlock *StartMBB = MBB;
MachineBasicBlock *JoinMBB = splitBlockBefore(MI, MBB);
MachineBasicBlock *FalseMBB = emitBlockAfter(StartMBB);
MBB = StartMBB;
BuildMI(MBB, DL, TII->get(SystemZ::BRC))
.addImm(CCValid).addImm(CCMask).addMBB(JoinMBB);
MBB->addSuccessor(JoinMBB);
MBB->addSuccessor(FalseMBB);
MBB = FalseMBB;
BuildMI(MBB, DL, TII->get(StoreOpcode))
.addReg(SrcReg).addOperand(Base).addImm(Disp).addReg(IndexReg);
MBB->addSuccessor(JoinMBB);
MI->eraseFromParent();
return JoinMBB;
}
MachineBasicBlock *
SystemZTargetLowering::emitAtomicLoadBinary(MachineInstr *MI,
MachineBasicBlock *MBB,
unsigned BinOpcode,
unsigned BitSize,
bool Invert) const {
MachineFunction &MF = *MBB->getParent();
const SystemZInstrInfo *TII =
static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
MachineRegisterInfo &MRI = MF.getRegInfo();
bool IsSubWord = (BitSize < 32);
unsigned Dest = MI->getOperand(0).getReg();
MachineOperand Base = earlyUseOperand(MI->getOperand(1));
int64_t Disp = MI->getOperand(2).getImm();
MachineOperand Src2 = earlyUseOperand(MI->getOperand(3));
unsigned BitShift = (IsSubWord ? MI->getOperand(4).getReg() : 0);
unsigned NegBitShift = (IsSubWord ? MI->getOperand(5).getReg() : 0);
DebugLoc DL = MI->getDebugLoc();
if (IsSubWord)
BitSize = MI->getOperand(6).getImm();
const TargetRegisterClass *RC = (BitSize <= 32 ?
&SystemZ::GR32BitRegClass :
&SystemZ::GR64BitRegClass);
unsigned LOpcode = BitSize <= 32 ? SystemZ::L : SystemZ::LG;
unsigned CSOpcode = BitSize <= 32 ? SystemZ::CS : SystemZ::CSG;
LOpcode = TII->getOpcodeForOffset(LOpcode, Disp);
CSOpcode = TII->getOpcodeForOffset(CSOpcode, Disp);
assert(LOpcode && CSOpcode && "Displacement out of range");
unsigned OrigVal = MRI.createVirtualRegister(RC);
unsigned OldVal = MRI.createVirtualRegister(RC);
unsigned NewVal = (BinOpcode || IsSubWord ?
MRI.createVirtualRegister(RC) : Src2.getReg());
unsigned RotatedOldVal = (IsSubWord ? MRI.createVirtualRegister(RC) : OldVal);
unsigned RotatedNewVal = (IsSubWord ? MRI.createVirtualRegister(RC) : NewVal);
MachineBasicBlock *StartMBB = MBB;
MachineBasicBlock *DoneMBB = splitBlockBefore(MI, MBB);
MachineBasicBlock *LoopMBB = emitBlockAfter(StartMBB);
MBB = StartMBB;
BuildMI(MBB, DL, TII->get(LOpcode), OrigVal)
.addOperand(Base).addImm(Disp).addReg(0);
MBB->addSuccessor(LoopMBB);
MBB = LoopMBB;
BuildMI(MBB, DL, TII->get(SystemZ::PHI), OldVal)
.addReg(OrigVal).addMBB(StartMBB)
.addReg(Dest).addMBB(LoopMBB);
if (IsSubWord)
BuildMI(MBB, DL, TII->get(SystemZ::RLL), RotatedOldVal)
.addReg(OldVal).addReg(BitShift).addImm(0);
if (Invert) {
unsigned Tmp = MRI.createVirtualRegister(RC);
BuildMI(MBB, DL, TII->get(BinOpcode), Tmp)
.addReg(RotatedOldVal).addOperand(Src2);
if (BitSize <= 32)
BuildMI(MBB, DL, TII->get(SystemZ::XILF), RotatedNewVal)
.addReg(Tmp).addImm(-1U << (32 - BitSize));
else {
unsigned Tmp2 = MRI.createVirtualRegister(RC);
BuildMI(MBB, DL, TII->get(SystemZ::LCGR), Tmp2).addReg(Tmp);
BuildMI(MBB, DL, TII->get(SystemZ::AGHI), RotatedNewVal)
.addReg(Tmp2).addImm(-1);
}
} else if (BinOpcode)
BuildMI(MBB, DL, TII->get(BinOpcode), RotatedNewVal)
.addReg(RotatedOldVal).addOperand(Src2);
else if (IsSubWord)
BuildMI(MBB, DL, TII->get(SystemZ::RISBG32), RotatedNewVal)
.addReg(RotatedOldVal).addReg(Src2.getReg())
.addImm(32).addImm(31 + BitSize).addImm(32 - BitSize);
if (IsSubWord)
BuildMI(MBB, DL, TII->get(SystemZ::RLL), NewVal)
.addReg(RotatedNewVal).addReg(NegBitShift).addImm(0);
BuildMI(MBB, DL, TII->get(CSOpcode), Dest)
.addReg(OldVal).addReg(NewVal).addOperand(Base).addImm(Disp);
BuildMI(MBB, DL, TII->get(SystemZ::BRC))
.addImm(SystemZ::CCMASK_CS).addImm(SystemZ::CCMASK_CS_NE).addMBB(LoopMBB);
MBB->addSuccessor(LoopMBB);
MBB->addSuccessor(DoneMBB);
MI->eraseFromParent();
return DoneMBB;
}
MachineBasicBlock *
SystemZTargetLowering::emitAtomicLoadMinMax(MachineInstr *MI,
MachineBasicBlock *MBB,
unsigned CompareOpcode,
unsigned KeepOldMask,
unsigned BitSize) const {
MachineFunction &MF = *MBB->getParent();
const SystemZInstrInfo *TII =
static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
MachineRegisterInfo &MRI = MF.getRegInfo();
bool IsSubWord = (BitSize < 32);
unsigned Dest = MI->getOperand(0).getReg();
MachineOperand Base = earlyUseOperand(MI->getOperand(1));
int64_t Disp = MI->getOperand(2).getImm();
unsigned Src2 = MI->getOperand(3).getReg();
unsigned BitShift = (IsSubWord ? MI->getOperand(4).getReg() : 0);
unsigned NegBitShift = (IsSubWord ? MI->getOperand(5).getReg() : 0);
DebugLoc DL = MI->getDebugLoc();
if (IsSubWord)
BitSize = MI->getOperand(6).getImm();
const TargetRegisterClass *RC = (BitSize <= 32 ?
&SystemZ::GR32BitRegClass :
&SystemZ::GR64BitRegClass);
unsigned LOpcode = BitSize <= 32 ? SystemZ::L : SystemZ::LG;
unsigned CSOpcode = BitSize <= 32 ? SystemZ::CS : SystemZ::CSG;
LOpcode = TII->getOpcodeForOffset(LOpcode, Disp);
CSOpcode = TII->getOpcodeForOffset(CSOpcode, Disp);
assert(LOpcode && CSOpcode && "Displacement out of range");
unsigned OrigVal = MRI.createVirtualRegister(RC);
unsigned OldVal = MRI.createVirtualRegister(RC);
unsigned NewVal = MRI.createVirtualRegister(RC);
unsigned RotatedOldVal = (IsSubWord ? MRI.createVirtualRegister(RC) : OldVal);
unsigned RotatedAltVal = (IsSubWord ? MRI.createVirtualRegister(RC) : Src2);
unsigned RotatedNewVal = (IsSubWord ? MRI.createVirtualRegister(RC) : NewVal);
MachineBasicBlock *StartMBB = MBB;
MachineBasicBlock *DoneMBB = splitBlockBefore(MI, MBB);
MachineBasicBlock *LoopMBB = emitBlockAfter(StartMBB);
MachineBasicBlock *UseAltMBB = emitBlockAfter(LoopMBB);
MachineBasicBlock *UpdateMBB = emitBlockAfter(UseAltMBB);
MBB = StartMBB;
BuildMI(MBB, DL, TII->get(LOpcode), OrigVal)
.addOperand(Base).addImm(Disp).addReg(0);
MBB->addSuccessor(LoopMBB);
MBB = LoopMBB;
BuildMI(MBB, DL, TII->get(SystemZ::PHI), OldVal)
.addReg(OrigVal).addMBB(StartMBB)
.addReg(Dest).addMBB(UpdateMBB);
if (IsSubWord)
BuildMI(MBB, DL, TII->get(SystemZ::RLL), RotatedOldVal)
.addReg(OldVal).addReg(BitShift).addImm(0);
BuildMI(MBB, DL, TII->get(CompareOpcode))
.addReg(RotatedOldVal).addReg(Src2);
BuildMI(MBB, DL, TII->get(SystemZ::BRC))
.addImm(SystemZ::CCMASK_ICMP).addImm(KeepOldMask).addMBB(UpdateMBB);
MBB->addSuccessor(UpdateMBB);
MBB->addSuccessor(UseAltMBB);
MBB = UseAltMBB;
if (IsSubWord)
BuildMI(MBB, DL, TII->get(SystemZ::RISBG32), RotatedAltVal)
.addReg(RotatedOldVal).addReg(Src2)
.addImm(32).addImm(31 + BitSize).addImm(0);
MBB->addSuccessor(UpdateMBB);
MBB = UpdateMBB;
BuildMI(MBB, DL, TII->get(SystemZ::PHI), RotatedNewVal)
.addReg(RotatedOldVal).addMBB(LoopMBB)
.addReg(RotatedAltVal).addMBB(UseAltMBB);
if (IsSubWord)
BuildMI(MBB, DL, TII->get(SystemZ::RLL), NewVal)
.addReg(RotatedNewVal).addReg(NegBitShift).addImm(0);
BuildMI(MBB, DL, TII->get(CSOpcode), Dest)
.addReg(OldVal).addReg(NewVal).addOperand(Base).addImm(Disp);
BuildMI(MBB, DL, TII->get(SystemZ::BRC))
.addImm(SystemZ::CCMASK_CS).addImm(SystemZ::CCMASK_CS_NE).addMBB(LoopMBB);
MBB->addSuccessor(LoopMBB);
MBB->addSuccessor(DoneMBB);
MI->eraseFromParent();
return DoneMBB;
}
MachineBasicBlock *
SystemZTargetLowering::emitAtomicCmpSwapW(MachineInstr *MI,
MachineBasicBlock *MBB) const {
MachineFunction &MF = *MBB->getParent();
const SystemZInstrInfo *TII =
static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
MachineRegisterInfo &MRI = MF.getRegInfo();
unsigned Dest = MI->getOperand(0).getReg();
MachineOperand Base = earlyUseOperand(MI->getOperand(1));
int64_t Disp = MI->getOperand(2).getImm();
unsigned OrigCmpVal = MI->getOperand(3).getReg();
unsigned OrigSwapVal = MI->getOperand(4).getReg();
unsigned BitShift = MI->getOperand(5).getReg();
unsigned NegBitShift = MI->getOperand(6).getReg();
int64_t BitSize = MI->getOperand(7).getImm();
DebugLoc DL = MI->getDebugLoc();
const TargetRegisterClass *RC = &SystemZ::GR32BitRegClass;
unsigned LOpcode = TII->getOpcodeForOffset(SystemZ::L, Disp);
unsigned CSOpcode = TII->getOpcodeForOffset(SystemZ::CS, Disp);
assert(LOpcode && CSOpcode && "Displacement out of range");
unsigned OrigOldVal = MRI.createVirtualRegister(RC);
unsigned OldVal = MRI.createVirtualRegister(RC);
unsigned CmpVal = MRI.createVirtualRegister(RC);
unsigned SwapVal = MRI.createVirtualRegister(RC);
unsigned StoreVal = MRI.createVirtualRegister(RC);
unsigned RetryOldVal = MRI.createVirtualRegister(RC);
unsigned RetryCmpVal = MRI.createVirtualRegister(RC);
unsigned RetrySwapVal = MRI.createVirtualRegister(RC);
MachineBasicBlock *StartMBB = MBB;
MachineBasicBlock *DoneMBB = splitBlockBefore(MI, MBB);
MachineBasicBlock *LoopMBB = emitBlockAfter(StartMBB);
MachineBasicBlock *SetMBB = emitBlockAfter(LoopMBB);
MBB = StartMBB;
BuildMI(MBB, DL, TII->get(LOpcode), OrigOldVal)
.addOperand(Base).addImm(Disp).addReg(0);
MBB->addSuccessor(LoopMBB);
MBB = LoopMBB;
BuildMI(MBB, DL, TII->get(SystemZ::PHI), OldVal)
.addReg(OrigOldVal).addMBB(StartMBB)
.addReg(RetryOldVal).addMBB(SetMBB);
BuildMI(MBB, DL, TII->get(SystemZ::PHI), CmpVal)
.addReg(OrigCmpVal).addMBB(StartMBB)
.addReg(RetryCmpVal).addMBB(SetMBB);
BuildMI(MBB, DL, TII->get(SystemZ::PHI), SwapVal)
.addReg(OrigSwapVal).addMBB(StartMBB)
.addReg(RetrySwapVal).addMBB(SetMBB);
BuildMI(MBB, DL, TII->get(SystemZ::RLL), Dest)
.addReg(OldVal).addReg(BitShift).addImm(BitSize);
BuildMI(MBB, DL, TII->get(SystemZ::RISBG32), RetryCmpVal)
.addReg(CmpVal).addReg(Dest).addImm(32).addImm(63 - BitSize).addImm(0);
BuildMI(MBB, DL, TII->get(SystemZ::CR))
.addReg(Dest).addReg(RetryCmpVal);
BuildMI(MBB, DL, TII->get(SystemZ::BRC))
.addImm(SystemZ::CCMASK_ICMP)
.addImm(SystemZ::CCMASK_CMP_NE).addMBB(DoneMBB);
MBB->addSuccessor(DoneMBB);
MBB->addSuccessor(SetMBB);
MBB = SetMBB;
BuildMI(MBB, DL, TII->get(SystemZ::RISBG32), RetrySwapVal)
.addReg(SwapVal).addReg(Dest).addImm(32).addImm(63 - BitSize).addImm(0);
BuildMI(MBB, DL, TII->get(SystemZ::RLL), StoreVal)
.addReg(RetrySwapVal).addReg(NegBitShift).addImm(-BitSize);
BuildMI(MBB, DL, TII->get(CSOpcode), RetryOldVal)
.addReg(OldVal).addReg(StoreVal).addOperand(Base).addImm(Disp);
BuildMI(MBB, DL, TII->get(SystemZ::BRC))
.addImm(SystemZ::CCMASK_CS).addImm(SystemZ::CCMASK_CS_NE).addMBB(LoopMBB);
MBB->addSuccessor(LoopMBB);
MBB->addSuccessor(DoneMBB);
MI->eraseFromParent();
return DoneMBB;
}
MachineBasicBlock *
SystemZTargetLowering::emitExt128(MachineInstr *MI,
MachineBasicBlock *MBB,
bool ClearEven, unsigned SubReg) const {
MachineFunction &MF = *MBB->getParent();
const SystemZInstrInfo *TII =
static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
MachineRegisterInfo &MRI = MF.getRegInfo();
DebugLoc DL = MI->getDebugLoc();
unsigned Dest = MI->getOperand(0).getReg();
unsigned Src = MI->getOperand(1).getReg();
unsigned In128 = MRI.createVirtualRegister(&SystemZ::GR128BitRegClass);
BuildMI(*MBB, MI, DL, TII->get(TargetOpcode::IMPLICIT_DEF), In128);
if (ClearEven) {
unsigned NewIn128 = MRI.createVirtualRegister(&SystemZ::GR128BitRegClass);
unsigned Zero64 = MRI.createVirtualRegister(&SystemZ::GR64BitRegClass);
BuildMI(*MBB, MI, DL, TII->get(SystemZ::LLILL), Zero64)
.addImm(0);
BuildMI(*MBB, MI, DL, TII->get(TargetOpcode::INSERT_SUBREG), NewIn128)
.addReg(In128).addReg(Zero64).addImm(SystemZ::subreg_h64);
In128 = NewIn128;
}
BuildMI(*MBB, MI, DL, TII->get(TargetOpcode::INSERT_SUBREG), Dest)
.addReg(In128).addReg(Src).addImm(SubReg);
MI->eraseFromParent();
return MBB;
}
MachineBasicBlock *
SystemZTargetLowering::emitMemMemWrapper(MachineInstr *MI,
MachineBasicBlock *MBB,
unsigned Opcode) const {
MachineFunction &MF = *MBB->getParent();
const SystemZInstrInfo *TII =
static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
MachineRegisterInfo &MRI = MF.getRegInfo();
DebugLoc DL = MI->getDebugLoc();
MachineOperand DestBase = earlyUseOperand(MI->getOperand(0));
uint64_t DestDisp = MI->getOperand(1).getImm();
MachineOperand SrcBase = earlyUseOperand(MI->getOperand(2));
uint64_t SrcDisp = MI->getOperand(3).getImm();
uint64_t Length = MI->getOperand(4).getImm();
MachineBasicBlock *EndMBB = (Length > 256 && Opcode == SystemZ::CLC ?
splitBlockAfter(MI, MBB) : nullptr);
if (MI->getNumExplicitOperands() > 5) {
bool HaveSingleBase = DestBase.isIdenticalTo(SrcBase);
uint64_t StartCountReg = MI->getOperand(5).getReg();
uint64_t StartSrcReg = forceReg(MI, SrcBase, TII);
uint64_t StartDestReg = (HaveSingleBase ? StartSrcReg :
forceReg(MI, DestBase, TII));
const TargetRegisterClass *RC = &SystemZ::ADDR64BitRegClass;
uint64_t ThisSrcReg = MRI.createVirtualRegister(RC);
uint64_t ThisDestReg = (HaveSingleBase ? ThisSrcReg :
MRI.createVirtualRegister(RC));
uint64_t NextSrcReg = MRI.createVirtualRegister(RC);
uint64_t NextDestReg = (HaveSingleBase ? NextSrcReg :
MRI.createVirtualRegister(RC));
RC = &SystemZ::GR64BitRegClass;
uint64_t ThisCountReg = MRI.createVirtualRegister(RC);
uint64_t NextCountReg = MRI.createVirtualRegister(RC);
MachineBasicBlock *StartMBB = MBB;
MachineBasicBlock *DoneMBB = splitBlockBefore(MI, MBB);
MachineBasicBlock *LoopMBB = emitBlockAfter(StartMBB);
MachineBasicBlock *NextMBB = (EndMBB ? emitBlockAfter(LoopMBB) : LoopMBB);
MBB->addSuccessor(LoopMBB);
MBB = LoopMBB;
BuildMI(MBB, DL, TII->get(SystemZ::PHI), ThisDestReg)
.addReg(StartDestReg).addMBB(StartMBB)
.addReg(NextDestReg).addMBB(NextMBB);
if (!HaveSingleBase)
BuildMI(MBB, DL, TII->get(SystemZ::PHI), ThisSrcReg)
.addReg(StartSrcReg).addMBB(StartMBB)
.addReg(NextSrcReg).addMBB(NextMBB);
BuildMI(MBB, DL, TII->get(SystemZ::PHI), ThisCountReg)
.addReg(StartCountReg).addMBB(StartMBB)
.addReg(NextCountReg).addMBB(NextMBB);
if (Opcode == SystemZ::MVC)
BuildMI(MBB, DL, TII->get(SystemZ::PFD))
.addImm(SystemZ::PFD_WRITE)
.addReg(ThisDestReg).addImm(DestDisp + 768).addReg(0);
BuildMI(MBB, DL, TII->get(Opcode))
.addReg(ThisDestReg).addImm(DestDisp).addImm(256)
.addReg(ThisSrcReg).addImm(SrcDisp);
if (EndMBB) {
BuildMI(MBB, DL, TII->get(SystemZ::BRC))
.addImm(SystemZ::CCMASK_ICMP).addImm(SystemZ::CCMASK_CMP_NE)
.addMBB(EndMBB);
MBB->addSuccessor(EndMBB);
MBB->addSuccessor(NextMBB);
}
MBB = NextMBB;
BuildMI(MBB, DL, TII->get(SystemZ::LA), NextDestReg)
.addReg(ThisDestReg).addImm(256).addReg(0);
if (!HaveSingleBase)
BuildMI(MBB, DL, TII->get(SystemZ::LA), NextSrcReg)
.addReg(ThisSrcReg).addImm(256).addReg(0);
BuildMI(MBB, DL, TII->get(SystemZ::AGHI), NextCountReg)
.addReg(ThisCountReg).addImm(-1);
BuildMI(MBB, DL, TII->get(SystemZ::CGHI))
.addReg(NextCountReg).addImm(0);
BuildMI(MBB, DL, TII->get(SystemZ::BRC))
.addImm(SystemZ::CCMASK_ICMP).addImm(SystemZ::CCMASK_CMP_NE)
.addMBB(LoopMBB);
MBB->addSuccessor(LoopMBB);
MBB->addSuccessor(DoneMBB);
DestBase = MachineOperand::CreateReg(NextDestReg, false);
SrcBase = MachineOperand::CreateReg(NextSrcReg, false);
Length &= 255;
MBB = DoneMBB;
}
while (Length > 0) {
uint64_t ThisLength = std::min(Length, uint64_t(256));
if (!isUInt<12>(DestDisp)) {
unsigned Reg = MRI.createVirtualRegister(&SystemZ::ADDR64BitRegClass);
BuildMI(*MBB, MI, MI->getDebugLoc(), TII->get(SystemZ::LAY), Reg)
.addOperand(DestBase).addImm(DestDisp).addReg(0);
DestBase = MachineOperand::CreateReg(Reg, false);
DestDisp = 0;
}
if (!isUInt<12>(SrcDisp)) {
unsigned Reg = MRI.createVirtualRegister(&SystemZ::ADDR64BitRegClass);
BuildMI(*MBB, MI, MI->getDebugLoc(), TII->get(SystemZ::LAY), Reg)
.addOperand(SrcBase).addImm(SrcDisp).addReg(0);
SrcBase = MachineOperand::CreateReg(Reg, false);
SrcDisp = 0;
}
BuildMI(*MBB, MI, DL, TII->get(Opcode))
.addOperand(DestBase).addImm(DestDisp).addImm(ThisLength)
.addOperand(SrcBase).addImm(SrcDisp);
DestDisp += ThisLength;
SrcDisp += ThisLength;
Length -= ThisLength;
if (EndMBB && Length > 0) {
MachineBasicBlock *NextMBB = splitBlockBefore(MI, MBB);
BuildMI(MBB, DL, TII->get(SystemZ::BRC))
.addImm(SystemZ::CCMASK_ICMP).addImm(SystemZ::CCMASK_CMP_NE)
.addMBB(EndMBB);
MBB->addSuccessor(EndMBB);
MBB->addSuccessor(NextMBB);
MBB = NextMBB;
}
}
if (EndMBB) {
MBB->addSuccessor(EndMBB);
MBB = EndMBB;
MBB->addLiveIn(SystemZ::CC);
}
MI->eraseFromParent();
return MBB;
}
MachineBasicBlock *
SystemZTargetLowering::emitStringWrapper(MachineInstr *MI,
MachineBasicBlock *MBB,
unsigned Opcode) const {
MachineFunction &MF = *MBB->getParent();
const SystemZInstrInfo *TII =
static_cast<const SystemZInstrInfo *>(Subtarget.getInstrInfo());
MachineRegisterInfo &MRI = MF.getRegInfo();
DebugLoc DL = MI->getDebugLoc();
uint64_t End1Reg = MI->getOperand(0).getReg();
uint64_t Start1Reg = MI->getOperand(1).getReg();
uint64_t Start2Reg = MI->getOperand(2).getReg();
uint64_t CharReg = MI->getOperand(3).getReg();
const TargetRegisterClass *RC = &SystemZ::GR64BitRegClass;
uint64_t This1Reg = MRI.createVirtualRegister(RC);
uint64_t This2Reg = MRI.createVirtualRegister(RC);
uint64_t End2Reg = MRI.createVirtualRegister(RC);
MachineBasicBlock *StartMBB = MBB;
MachineBasicBlock *DoneMBB = splitBlockBefore(MI, MBB);
MachineBasicBlock *LoopMBB = emitBlockAfter(StartMBB);
MBB->addSuccessor(LoopMBB);
MBB = LoopMBB;
BuildMI(MBB, DL, TII->get(SystemZ::PHI), This1Reg)
.addReg(Start1Reg).addMBB(StartMBB)
.addReg(End1Reg).addMBB(LoopMBB);
BuildMI(MBB, DL, TII->get(SystemZ::PHI), This2Reg)
.addReg(Start2Reg).addMBB(StartMBB)
.addReg(End2Reg).addMBB(LoopMBB);
BuildMI(MBB, DL, TII->get(TargetOpcode::COPY), SystemZ::R0L).addReg(CharReg);
BuildMI(MBB, DL, TII->get(Opcode))
.addReg(End1Reg, RegState::Define).addReg(End2Reg, RegState::Define)
.addReg(This1Reg).addReg(This2Reg);
BuildMI(MBB, DL, TII->get(SystemZ::BRC))
.addImm(SystemZ::CCMASK_ANY).addImm(SystemZ::CCMASK_3).addMBB(LoopMBB);
MBB->addSuccessor(LoopMBB);
MBB->addSuccessor(DoneMBB);
DoneMBB->addLiveIn(SystemZ::CC);
MI->eraseFromParent();
return DoneMBB;
}
MachineBasicBlock *SystemZTargetLowering::
EmitInstrWithCustomInserter(MachineInstr *MI, MachineBasicBlock *MBB) const {
switch (MI->getOpcode()) {
case SystemZ::Select32Mux:
case SystemZ::Select32:
case SystemZ::SelectF32:
case SystemZ::Select64:
case SystemZ::SelectF64:
case SystemZ::SelectF128:
return emitSelect(MI, MBB);
case SystemZ::CondStore8Mux:
return emitCondStore(MI, MBB, SystemZ::STCMux, 0, false);
case SystemZ::CondStore8MuxInv:
return emitCondStore(MI, MBB, SystemZ::STCMux, 0, true);
case SystemZ::CondStore16Mux:
return emitCondStore(MI, MBB, SystemZ::STHMux, 0, false);
case SystemZ::CondStore16MuxInv:
return emitCondStore(MI, MBB, SystemZ::STHMux, 0, true);
case SystemZ::CondStore8:
return emitCondStore(MI, MBB, SystemZ::STC, 0, false);
case SystemZ::CondStore8Inv:
return emitCondStore(MI, MBB, SystemZ::STC, 0, true);
case SystemZ::CondStore16:
return emitCondStore(MI, MBB, SystemZ::STH, 0, false);
case SystemZ::CondStore16Inv:
return emitCondStore(MI, MBB, SystemZ::STH, 0, true);
case SystemZ::CondStore32:
return emitCondStore(MI, MBB, SystemZ::ST, SystemZ::STOC, false);
case SystemZ::CondStore32Inv:
return emitCondStore(MI, MBB, SystemZ::ST, SystemZ::STOC, true);
case SystemZ::CondStore64:
return emitCondStore(MI, MBB, SystemZ::STG, SystemZ::STOCG, false);
case SystemZ::CondStore64Inv:
return emitCondStore(MI, MBB, SystemZ::STG, SystemZ::STOCG, true);
case SystemZ::CondStoreF32:
return emitCondStore(MI, MBB, SystemZ::STE, 0, false);
case SystemZ::CondStoreF32Inv:
return emitCondStore(MI, MBB, SystemZ::STE, 0, true);
case SystemZ::CondStoreF64:
return emitCondStore(MI, MBB, SystemZ::STD, 0, false);
case SystemZ::CondStoreF64Inv:
return emitCondStore(MI, MBB, SystemZ::STD, 0, true);
case SystemZ::AEXT128_64:
return emitExt128(MI, MBB, false, SystemZ::subreg_l64);
case SystemZ::ZEXT128_32:
return emitExt128(MI, MBB, true, SystemZ::subreg_l32);
case SystemZ::ZEXT128_64:
return emitExt128(MI, MBB, true, SystemZ::subreg_l64);
case SystemZ::ATOMIC_SWAPW:
return emitAtomicLoadBinary(MI, MBB, 0, 0);
case SystemZ::ATOMIC_SWAP_32:
return emitAtomicLoadBinary(MI, MBB, 0, 32);
case SystemZ::ATOMIC_SWAP_64:
return emitAtomicLoadBinary(MI, MBB, 0, 64);
case SystemZ::ATOMIC_LOADW_AR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::AR, 0);
case SystemZ::ATOMIC_LOADW_AFI:
return emitAtomicLoadBinary(MI, MBB, SystemZ::AFI, 0);
case SystemZ::ATOMIC_LOAD_AR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::AR, 32);
case SystemZ::ATOMIC_LOAD_AHI:
return emitAtomicLoadBinary(MI, MBB, SystemZ::AHI, 32);
case SystemZ::ATOMIC_LOAD_AFI:
return emitAtomicLoadBinary(MI, MBB, SystemZ::AFI, 32);
case SystemZ::ATOMIC_LOAD_AGR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::AGR, 64);
case SystemZ::ATOMIC_LOAD_AGHI:
return emitAtomicLoadBinary(MI, MBB, SystemZ::AGHI, 64);
case SystemZ::ATOMIC_LOAD_AGFI:
return emitAtomicLoadBinary(MI, MBB, SystemZ::AGFI, 64);
case SystemZ::ATOMIC_LOADW_SR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::SR, 0);
case SystemZ::ATOMIC_LOAD_SR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::SR, 32);
case SystemZ::ATOMIC_LOAD_SGR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::SGR, 64);
case SystemZ::ATOMIC_LOADW_NR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NR, 0);
case SystemZ::ATOMIC_LOADW_NILH:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NILH, 0);
case SystemZ::ATOMIC_LOAD_NR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NR, 32);
case SystemZ::ATOMIC_LOAD_NILL:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NILL, 32);
case SystemZ::ATOMIC_LOAD_NILH:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NILH, 32);
case SystemZ::ATOMIC_LOAD_NILF:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NILF, 32);
case SystemZ::ATOMIC_LOAD_NGR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NGR, 64);
case SystemZ::ATOMIC_LOAD_NILL64:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NILL64, 64);
case SystemZ::ATOMIC_LOAD_NILH64:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NILH64, 64);
case SystemZ::ATOMIC_LOAD_NIHL64:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NIHL64, 64);
case SystemZ::ATOMIC_LOAD_NIHH64:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NIHH64, 64);
case SystemZ::ATOMIC_LOAD_NILF64:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NILF64, 64);
case SystemZ::ATOMIC_LOAD_NIHF64:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NIHF64, 64);
case SystemZ::ATOMIC_LOADW_OR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::OR, 0);
case SystemZ::ATOMIC_LOADW_OILH:
return emitAtomicLoadBinary(MI, MBB, SystemZ::OILH, 0);
case SystemZ::ATOMIC_LOAD_OR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::OR, 32);
case SystemZ::ATOMIC_LOAD_OILL:
return emitAtomicLoadBinary(MI, MBB, SystemZ::OILL, 32);
case SystemZ::ATOMIC_LOAD_OILH:
return emitAtomicLoadBinary(MI, MBB, SystemZ::OILH, 32);
case SystemZ::ATOMIC_LOAD_OILF:
return emitAtomicLoadBinary(MI, MBB, SystemZ::OILF, 32);
case SystemZ::ATOMIC_LOAD_OGR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::OGR, 64);
case SystemZ::ATOMIC_LOAD_OILL64:
return emitAtomicLoadBinary(MI, MBB, SystemZ::OILL64, 64);
case SystemZ::ATOMIC_LOAD_OILH64:
return emitAtomicLoadBinary(MI, MBB, SystemZ::OILH64, 64);
case SystemZ::ATOMIC_LOAD_OIHL64:
return emitAtomicLoadBinary(MI, MBB, SystemZ::OIHL64, 64);
case SystemZ::ATOMIC_LOAD_OIHH64:
return emitAtomicLoadBinary(MI, MBB, SystemZ::OIHH64, 64);
case SystemZ::ATOMIC_LOAD_OILF64:
return emitAtomicLoadBinary(MI, MBB, SystemZ::OILF64, 64);
case SystemZ::ATOMIC_LOAD_OIHF64:
return emitAtomicLoadBinary(MI, MBB, SystemZ::OIHF64, 64);
case SystemZ::ATOMIC_LOADW_XR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::XR, 0);
case SystemZ::ATOMIC_LOADW_XILF:
return emitAtomicLoadBinary(MI, MBB, SystemZ::XILF, 0);
case SystemZ::ATOMIC_LOAD_XR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::XR, 32);
case SystemZ::ATOMIC_LOAD_XILF:
return emitAtomicLoadBinary(MI, MBB, SystemZ::XILF, 32);
case SystemZ::ATOMIC_LOAD_XGR:
return emitAtomicLoadBinary(MI, MBB, SystemZ::XGR, 64);
case SystemZ::ATOMIC_LOAD_XILF64:
return emitAtomicLoadBinary(MI, MBB, SystemZ::XILF64, 64);
case SystemZ::ATOMIC_LOAD_XIHF64:
return emitAtomicLoadBinary(MI, MBB, SystemZ::XIHF64, 64);
case SystemZ::ATOMIC_LOADW_NRi:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NR, 0, true);
case SystemZ::ATOMIC_LOADW_NILHi:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NILH, 0, true);
case SystemZ::ATOMIC_LOAD_NRi:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NR, 32, true);
case SystemZ::ATOMIC_LOAD_NILLi:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NILL, 32, true);
case SystemZ::ATOMIC_LOAD_NILHi:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NILH, 32, true);
case SystemZ::ATOMIC_LOAD_NILFi:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NILF, 32, true);
case SystemZ::ATOMIC_LOAD_NGRi:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NGR, 64, true);
case SystemZ::ATOMIC_LOAD_NILL64i:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NILL64, 64, true);
case SystemZ::ATOMIC_LOAD_NILH64i:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NILH64, 64, true);
case SystemZ::ATOMIC_LOAD_NIHL64i:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NIHL64, 64, true);
case SystemZ::ATOMIC_LOAD_NIHH64i:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NIHH64, 64, true);
case SystemZ::ATOMIC_LOAD_NILF64i:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NILF64, 64, true);
case SystemZ::ATOMIC_LOAD_NIHF64i:
return emitAtomicLoadBinary(MI, MBB, SystemZ::NIHF64, 64, true);
case SystemZ::ATOMIC_LOADW_MIN:
return emitAtomicLoadMinMax(MI, MBB, SystemZ::CR,
SystemZ::CCMASK_CMP_LE, 0);
case SystemZ::ATOMIC_LOAD_MIN_32:
return emitAtomicLoadMinMax(MI, MBB, SystemZ::CR,
SystemZ::CCMASK_CMP_LE, 32);
case SystemZ::ATOMIC_LOAD_MIN_64:
return emitAtomicLoadMinMax(MI, MBB, SystemZ::CGR,
SystemZ::CCMASK_CMP_LE, 64);
case SystemZ::ATOMIC_LOADW_MAX:
return emitAtomicLoadMinMax(MI, MBB, SystemZ::CR,
SystemZ::CCMASK_CMP_GE, 0);
case SystemZ::ATOMIC_LOAD_MAX_32:
return emitAtomicLoadMinMax(MI, MBB, SystemZ::CR,
SystemZ::CCMASK_CMP_GE, 32);
case SystemZ::ATOMIC_LOAD_MAX_64:
return emitAtomicLoadMinMax(MI, MBB, SystemZ::CGR,
SystemZ::CCMASK_CMP_GE, 64);
case SystemZ::ATOMIC_LOADW_UMIN:
return emitAtomicLoadMinMax(MI, MBB, SystemZ::CLR,
SystemZ::CCMASK_CMP_LE, 0);
case SystemZ::ATOMIC_LOAD_UMIN_32:
return emitAtomicLoadMinMax(MI, MBB, SystemZ::CLR,
SystemZ::CCMASK_CMP_LE, 32);
case SystemZ::ATOMIC_LOAD_UMIN_64:
return emitAtomicLoadMinMax(MI, MBB, SystemZ::CLGR,
SystemZ::CCMASK_CMP_LE, 64);
case SystemZ::ATOMIC_LOADW_UMAX:
return emitAtomicLoadMinMax(MI, MBB, SystemZ::CLR,
SystemZ::CCMASK_CMP_GE, 0);
case SystemZ::ATOMIC_LOAD_UMAX_32:
return emitAtomicLoadMinMax(MI, MBB, SystemZ::CLR,
SystemZ::CCMASK_CMP_GE, 32);
case SystemZ::ATOMIC_LOAD_UMAX_64:
return emitAtomicLoadMinMax(MI, MBB, SystemZ::CLGR,
SystemZ::CCMASK_CMP_GE, 64);
case SystemZ::ATOMIC_CMP_SWAPW:
return emitAtomicCmpSwapW(MI, MBB);
case SystemZ::MVCSequence:
case SystemZ::MVCLoop:
return emitMemMemWrapper(MI, MBB, SystemZ::MVC);
case SystemZ::NCSequence:
case SystemZ::NCLoop:
return emitMemMemWrapper(MI, MBB, SystemZ::NC);
case SystemZ::OCSequence:
case SystemZ::OCLoop:
return emitMemMemWrapper(MI, MBB, SystemZ::OC);
case SystemZ::XCSequence:
case SystemZ::XCLoop:
return emitMemMemWrapper(MI, MBB, SystemZ::XC);
case SystemZ::CLCSequence:
case SystemZ::CLCLoop:
return emitMemMemWrapper(MI, MBB, SystemZ::CLC);
case SystemZ::CLSTLoop:
return emitStringWrapper(MI, MBB, SystemZ::CLST);
case SystemZ::MVSTLoop:
return emitStringWrapper(MI, MBB, SystemZ::MVST);
case SystemZ::SRSTLoop:
return emitStringWrapper(MI, MBB, SystemZ::SRST);
default:
llvm_unreachable("Unexpected instr type to insert");
}
}