return Count == 0 ? 0 : (uint64_t(1) << (Count - 1) << 1) - 1;
}
+// Reg should be a 32-bit GPR. Return true if it is a high register rather
+// than a low register.
+static bool isHighReg(unsigned int Reg) {
+ if (SystemZ::GRH32BitRegClass.contains(Reg))
+ return true;
+ assert(SystemZ::GR32BitRegClass.contains(Reg) && "Invalid GRX32");
+ return false;
+}
+
SystemZInstrInfo::SystemZInstrInfo(SystemZTargetMachine &tm)
: SystemZGenInstrInfo(SystemZ::ADJCALLSTACKDOWN, SystemZ::ADJCALLSTACKUP),
RI(tm), TM(tm) {
// Set up the two 64-bit registers.
MachineOperand &HighRegOp = EarlierMI->getOperand(0);
MachineOperand &LowRegOp = MI->getOperand(0);
- HighRegOp.setReg(RI.getSubReg(HighRegOp.getReg(), SystemZ::subreg_high));
- LowRegOp.setReg(RI.getSubReg(LowRegOp.getReg(), SystemZ::subreg_low));
+ HighRegOp.setReg(RI.getSubReg(HighRegOp.getReg(), SystemZ::subreg_h64));
+ LowRegOp.setReg(RI.getSubReg(LowRegOp.getReg(), SystemZ::subreg_l64));
// The address in the first (high) instruction is already correct.
// Adjust the offset in the second (low) instruction.
OffsetMO.setImm(Offset);
}
+// MI is an RI-style pseudo instruction. Replace it with LowOpcode
+// if the first operand is a low GR32 and HighOpcode if the first operand
+// is a high GR32. ConvertHigh is true if LowOpcode takes a signed operand
+// and HighOpcode takes an unsigned 32-bit operand. In those cases,
+// MI has the same kind of operand as LowOpcode, so needs to be converted
+// if HighOpcode is used.
+void SystemZInstrInfo::expandRIPseudo(MachineInstr *MI, unsigned LowOpcode,
+ unsigned HighOpcode,
+ bool ConvertHigh) const {
+ unsigned Reg = MI->getOperand(0).getReg();
+ bool IsHigh = isHighReg(Reg);
+ MI->setDesc(get(IsHigh ? HighOpcode : LowOpcode));
+ if (IsHigh && ConvertHigh)
+ MI->getOperand(1).setImm(uint32_t(MI->getOperand(1).getImm()));
+}
+
+// MI is an RXY-style pseudo instruction. Replace it with LowOpcode
+// if the first operand is a low GR32 and HighOpcode if the first operand
+// is a high GR32.
+void SystemZInstrInfo::expandRXYPseudo(MachineInstr *MI, unsigned LowOpcode,
+ unsigned HighOpcode) const {
+ unsigned Reg = MI->getOperand(0).getReg();
+ unsigned Opcode = getOpcodeForOffset(isHighReg(Reg) ? HighOpcode : LowOpcode,
+ MI->getOperand(2).getImm());
+ MI->setDesc(get(Opcode));
+}
+
+// MI is an RR-style pseudo instruction that zero-extends the low Size bits
+// of one GRX32 into another. Replace it with LowOpcode if both operands
+// are low registers, otherwise use RISB[LH]G.
+void SystemZInstrInfo::expandZExtPseudo(MachineInstr *MI, unsigned LowOpcode,
+ unsigned Size) const {
+ emitGRX32Move(*MI->getParent(), MI, MI->getDebugLoc(),
+ MI->getOperand(0).getReg(), MI->getOperand(1).getReg(),
+ LowOpcode, Size, MI->getOperand(1).isKill());
+ MI->eraseFromParent();
+}
+
+// Emit a zero-extending move from 32-bit GPR SrcReg to 32-bit GPR
+// DestReg before MBBI in MBB. Use LowLowOpcode when both DestReg and SrcReg
+// are low registers, otherwise use RISB[LH]G. Size is the number of bits
+// taken from the low end of SrcReg (8 for LLCR, 16 for LLHR and 32 for LR).
+// KillSrc is true if this move is the last use of SrcReg.
+void SystemZInstrInfo::emitGRX32Move(MachineBasicBlock &MBB,
+ MachineBasicBlock::iterator MBBI,
+ DebugLoc DL, unsigned DestReg,
+ unsigned SrcReg, unsigned LowLowOpcode,
+ unsigned Size, bool KillSrc) const {
+ unsigned Opcode;
+ bool DestIsHigh = isHighReg(DestReg);
+ bool SrcIsHigh = isHighReg(SrcReg);
+ if (DestIsHigh && SrcIsHigh)
+ Opcode = SystemZ::RISBHH;
+ else if (DestIsHigh && !SrcIsHigh)
+ Opcode = SystemZ::RISBHL;
+ else if (!DestIsHigh && SrcIsHigh)
+ Opcode = SystemZ::RISBLH;
+ else {
+ BuildMI(MBB, MBBI, DL, get(LowLowOpcode), DestReg)
+ .addReg(SrcReg, getKillRegState(KillSrc));
+ return;
+ }
+ unsigned Rotate = (DestIsHigh != SrcIsHigh ? 32 : 0);
+ BuildMI(MBB, MBBI, DL, get(Opcode), DestReg)
+ .addReg(DestReg, RegState::Undef)
+ .addReg(SrcReg, getKillRegState(KillSrc))
+ .addImm(32 - Size).addImm(128 + 31).addImm(Rotate);
+}
+
// If MI is a simple load or store for a frame object, return the register
// it loads or stores and set FrameIndex to the index of the frame object.
// Return 0 otherwise.
return Count;
}
+bool SystemZInstrInfo::analyzeCompare(const MachineInstr *MI,
+ unsigned &SrcReg, unsigned &SrcReg2,
+ int &Mask, int &Value) const {
+ assert(MI->isCompare() && "Caller should have checked for a comparison");
+
+ if (MI->getNumExplicitOperands() == 2 &&
+ MI->getOperand(0).isReg() &&
+ MI->getOperand(1).isImm()) {
+ SrcReg = MI->getOperand(0).getReg();
+ SrcReg2 = 0;
+ Value = MI->getOperand(1).getImm();
+ Mask = ~0;
+ return true;
+ }
+
+ return false;
+}
+
+// If Reg is a virtual register, return its definition, otherwise return null.
+static MachineInstr *getDef(unsigned Reg,
+ const MachineRegisterInfo *MRI) {
+ if (TargetRegisterInfo::isPhysicalRegister(Reg))
+ return 0;
+ return MRI->getUniqueVRegDef(Reg);
+}
+
+// Return true if MI is a shift of type Opcode by Imm bits.
+static bool isShift(MachineInstr *MI, int Opcode, int64_t Imm) {
+ return (MI->getOpcode() == Opcode &&
+ !MI->getOperand(2).getReg() &&
+ MI->getOperand(3).getImm() == Imm);
+}
+
+// If the destination of MI has no uses, delete it as dead.
+static void eraseIfDead(MachineInstr *MI, const MachineRegisterInfo *MRI) {
+ if (MRI->use_nodbg_empty(MI->getOperand(0).getReg()))
+ MI->eraseFromParent();
+}
+
+// Compare compares SrcReg against zero. Check whether SrcReg contains
+// the result of an IPM sequence whose input CC survives until Compare,
+// and whether Compare is therefore redundant. Delete it and return
+// true if so.
+static bool removeIPMBasedCompare(MachineInstr *Compare, unsigned SrcReg,
+ const MachineRegisterInfo *MRI,
+ const TargetRegisterInfo *TRI) {
+ MachineInstr *LGFR = 0;
+ MachineInstr *RLL = getDef(SrcReg, MRI);
+ if (RLL && RLL->getOpcode() == SystemZ::LGFR) {
+ LGFR = RLL;
+ RLL = getDef(LGFR->getOperand(1).getReg(), MRI);
+ }
+ if (!RLL || !isShift(RLL, SystemZ::RLL, 31))
+ return false;
+
+ MachineInstr *SRL = getDef(RLL->getOperand(1).getReg(), MRI);
+ if (!SRL || !isShift(SRL, SystemZ::SRL, 28))
+ return false;
+
+ MachineInstr *IPM = getDef(SRL->getOperand(1).getReg(), MRI);
+ if (!IPM || IPM->getOpcode() != SystemZ::IPM)
+ return false;
+
+ // Check that there are no assignments to CC between the IPM and Compare,
+ if (IPM->getParent() != Compare->getParent())
+ return false;
+ MachineBasicBlock::iterator MBBI = IPM, MBBE = Compare;
+ for (++MBBI; MBBI != MBBE; ++MBBI) {
+ MachineInstr *MI = MBBI;
+ if (MI->modifiesRegister(SystemZ::CC, TRI))
+ return false;
+ }
+
+ Compare->eraseFromParent();
+ if (LGFR)
+ eraseIfDead(LGFR, MRI);
+ eraseIfDead(RLL, MRI);
+ eraseIfDead(SRL, MRI);
+ eraseIfDead(IPM, MRI);
+
+ return true;
+}
+
+bool
+SystemZInstrInfo::optimizeCompareInstr(MachineInstr *Compare,
+ unsigned SrcReg, unsigned SrcReg2,
+ int Mask, int Value,
+ const MachineRegisterInfo *MRI) const {
+ assert(!SrcReg2 && "Only optimizing constant comparisons so far");
+ bool IsLogical = (Compare->getDesc().TSFlags & SystemZII::IsLogical) != 0;
+ if (Value == 0 &&
+ !IsLogical &&
+ removeIPMBasedCompare(Compare, SrcReg, MRI, TM.getRegisterInfo()))
+ return true;
+ return false;
+}
+
// If Opcode is a move that has a conditional variant, return that variant,
// otherwise return 0.
static unsigned getConditionalMove(unsigned Opcode) {
bool KillSrc) const {
// Split 128-bit GPR moves into two 64-bit moves. This handles ADDR128 too.
if (SystemZ::GR128BitRegClass.contains(DestReg, SrcReg)) {
- copyPhysReg(MBB, MBBI, DL, RI.getSubReg(DestReg, SystemZ::subreg_high),
- RI.getSubReg(SrcReg, SystemZ::subreg_high), KillSrc);
- copyPhysReg(MBB, MBBI, DL, RI.getSubReg(DestReg, SystemZ::subreg_low),
- RI.getSubReg(SrcReg, SystemZ::subreg_low), KillSrc);
+ copyPhysReg(MBB, MBBI, DL, RI.getSubReg(DestReg, SystemZ::subreg_h64),
+ RI.getSubReg(SrcReg, SystemZ::subreg_h64), KillSrc);
+ copyPhysReg(MBB, MBBI, DL, RI.getSubReg(DestReg, SystemZ::subreg_l64),
+ RI.getSubReg(SrcReg, SystemZ::subreg_l64), KillSrc);
+ return;
+ }
+
+ if (SystemZ::GRX32BitRegClass.contains(DestReg, SrcReg)) {
+ emitGRX32Move(MBB, MBBI, DL, DestReg, SrcReg, SystemZ::LR, 32, KillSrc);
return;
}
// Everything else needs only one instruction.
unsigned Opcode;
- if (SystemZ::GR32BitRegClass.contains(DestReg, SrcReg))
- Opcode = SystemZ::LR;
- else if (SystemZ::GR64BitRegClass.contains(DestReg, SrcReg))
+ if (SystemZ::GR64BitRegClass.contains(DestReg, SrcReg))
Opcode = SystemZ::LGR;
else if (SystemZ::FP32BitRegClass.contains(DestReg, SrcReg))
Opcode = SystemZ::LER;
static LogicOp interpretAndImmediate(unsigned Opcode) {
switch (Opcode) {
- case SystemZ::NILL32: return LogicOp(32, 0, 16);
- case SystemZ::NILH32: return LogicOp(32, 16, 16);
- case SystemZ::NILL: return LogicOp(64, 0, 16);
- case SystemZ::NILH: return LogicOp(64, 16, 16);
- case SystemZ::NIHL: return LogicOp(64, 32, 16);
- case SystemZ::NIHH: return LogicOp(64, 48, 16);
- case SystemZ::NILF32: return LogicOp(32, 0, 32);
- case SystemZ::NILF: return LogicOp(64, 0, 32);
- case SystemZ::NIHF: return LogicOp(64, 32, 32);
+ case SystemZ::NILMux: return LogicOp(32, 0, 16);
+ case SystemZ::NIHMux: return LogicOp(32, 16, 16);
+ case SystemZ::NILL64: return LogicOp(64, 0, 16);
+ case SystemZ::NILH64: return LogicOp(64, 16, 16);
+ case SystemZ::NIHL64: return LogicOp(64, 32, 16);
+ case SystemZ::NIHH64: return LogicOp(64, 48, 16);
+ case SystemZ::NIFMux: return LogicOp(32, 0, 32);
+ case SystemZ::NILF64: return LogicOp(64, 0, 32);
+ case SystemZ::NIHF64: return LogicOp(64, 32, 32);
default: return LogicOp();
}
}
// Try to convert an AND into an RISBG-type instruction.
if (LogicOp And = interpretAndImmediate(Opcode)) {
- unsigned NewOpcode;
- if (And.RegSize == 64)
- NewOpcode = SystemZ::RISBG;
- else if (TM.getSubtargetImpl()->hasHighWord())
- NewOpcode = SystemZ::RISBLG32;
- else
- // We can't use RISBG for 32-bit operations because it clobbers the
- // high word of the destination too.
- NewOpcode = 0;
- if (NewOpcode) {
- uint64_t Imm = MI->getOperand(2).getImm() << And.ImmLSB;
- // AND IMMEDIATE leaves the other bits of the register unchanged.
- Imm |= allOnes(And.RegSize) & ~(allOnes(And.ImmSize) << And.ImmLSB);
- unsigned Start, End;
- if (isRxSBGMask(Imm, And.RegSize, Start, End)) {
- if (NewOpcode == SystemZ::RISBLG32) {
- Start &= 31;
- End &= 31;
- }
- MachineOperand &Dest = MI->getOperand(0);
- MachineOperand &Src = MI->getOperand(1);
- MachineInstrBuilder MIB =
- BuildMI(*MBB, MI, MI->getDebugLoc(), get(NewOpcode))
- .addOperand(Dest).addReg(0)
- .addReg(Src.getReg(), getKillRegState(Src.isKill()), Src.getSubReg())
- .addImm(Start).addImm(End + 128).addImm(0);
- return finishConvertToThreeAddress(MI, MIB, LV);
+ uint64_t Imm = MI->getOperand(2).getImm() << And.ImmLSB;
+ // AND IMMEDIATE leaves the other bits of the register unchanged.
+ Imm |= allOnes(And.RegSize) & ~(allOnes(And.ImmSize) << And.ImmLSB);
+ unsigned Start, End;
+ if (isRxSBGMask(Imm, And.RegSize, Start, End)) {
+ unsigned NewOpcode;
+ if (And.RegSize == 64)
+ NewOpcode = SystemZ::RISBG;
+ else {
+ NewOpcode = SystemZ::RISBMux;
+ Start &= 31;
+ End &= 31;
}
+ MachineOperand &Dest = MI->getOperand(0);
+ MachineOperand &Src = MI->getOperand(1);
+ MachineInstrBuilder MIB =
+ BuildMI(*MBB, MI, MI->getDebugLoc(), get(NewOpcode))
+ .addOperand(Dest).addReg(0)
+ .addReg(Src.getReg(), getKillRegState(Src.isKill()), Src.getSubReg())
+ .addImm(Start).addImm(End + 128).addImm(0);
+ return finishConvertToThreeAddress(MI, MIB, LV);
}
}
return 0;
//
// Although MVC is in practice a fast choice in these cases, it is still
// logically a bytewise copy. This means that we cannot use it if the
- // load or store is volatile. It also means that the transformation is
- // not valid in cases where the two memories partially overlap; however,
- // that is not a problem here, because we know that one of the memories
- // is a full frame index.
+ // load or store is volatile. We also wouldn't be able to use MVC if
+ // the two memories partially overlap, but that case cannot occur here,
+ // because we know that one of the memories is a full frame index.
+ //
+ // For performance reasons, we also want to avoid using MVC if the addresses
+ // might be equal. We don't worry about that case here, because spill slot
+ // coloring happens later, and because we have special code to remove
+ // MVCs that turn out to be redundant.
if (OpNum == 0 && MI->hasOneMemOperand()) {
MachineMemOperand *MMO = *MI->memoperands_begin();
if (MMO->getSize() == Size && !MMO->isVolatile()) {
splitMove(MI, SystemZ::STD);
return true;
+ case SystemZ::LBMux:
+ expandRXYPseudo(MI, SystemZ::LB, SystemZ::LBH);
+ return true;
+
+ case SystemZ::LHMux:
+ expandRXYPseudo(MI, SystemZ::LH, SystemZ::LHH);
+ return true;
+
+ case SystemZ::LLCRMux:
+ expandZExtPseudo(MI, SystemZ::LLCR, 8);
+ return true;
+
+ case SystemZ::LLHRMux:
+ expandZExtPseudo(MI, SystemZ::LLHR, 16);
+ return true;
+
+ case SystemZ::LLCMux:
+ expandRXYPseudo(MI, SystemZ::LLC, SystemZ::LLCH);
+ return true;
+
+ case SystemZ::LLHMux:
+ expandRXYPseudo(MI, SystemZ::LLH, SystemZ::LLHH);
+ return true;
+
+ case SystemZ::LMux:
+ expandRXYPseudo(MI, SystemZ::L, SystemZ::LFH);
+ return true;
+
+ case SystemZ::STCMux:
+ expandRXYPseudo(MI, SystemZ::STC, SystemZ::STCH);
+ return true;
+
+ case SystemZ::STHMux:
+ expandRXYPseudo(MI, SystemZ::STH, SystemZ::STHH);
+ return true;
+
+ case SystemZ::STMux:
+ expandRXYPseudo(MI, SystemZ::ST, SystemZ::STFH);
+ return true;
+
+ case SystemZ::LHIMux:
+ expandRIPseudo(MI, SystemZ::LHI, SystemZ::IIHF, true);
+ return true;
+
+ case SystemZ::IIFMux:
+ expandRIPseudo(MI, SystemZ::IILF, SystemZ::IIHF, false);
+ return true;
+
+ case SystemZ::IILMux:
+ expandRIPseudo(MI, SystemZ::IILL, SystemZ::IIHL, false);
+ return true;
+
+ case SystemZ::IIHMux:
+ expandRIPseudo(MI, SystemZ::IILH, SystemZ::IIHH, false);
+ return true;
+
+ case SystemZ::NIFMux:
+ expandRIPseudo(MI, SystemZ::NILF, SystemZ::NIHF, false);
+ return true;
+
+ case SystemZ::NILMux:
+ expandRIPseudo(MI, SystemZ::NILL, SystemZ::NIHL, false);
+ return true;
+
+ case SystemZ::NIHMux:
+ expandRIPseudo(MI, SystemZ::NILH, SystemZ::NIHH, false);
+ return true;
+
+ case SystemZ::OIFMux:
+ expandRIPseudo(MI, SystemZ::OILF, SystemZ::OIHF, false);
+ return true;
+
+ case SystemZ::OILMux:
+ expandRIPseudo(MI, SystemZ::OILL, SystemZ::OIHL, false);
+ return true;
+
+ case SystemZ::OIHMux:
+ expandRIPseudo(MI, SystemZ::OILH, SystemZ::OIHH, false);
+ return true;
+
+ case SystemZ::XIFMux:
+ expandRIPseudo(MI, SystemZ::XILF, SystemZ::XIHF, false);
+ return true;
+
+ case SystemZ::RISBMux: {
+ bool DestIsHigh = isHighReg(MI->getOperand(0).getReg());
+ bool SrcIsHigh = isHighReg(MI->getOperand(2).getReg());
+ if (SrcIsHigh == DestIsHigh)
+ MI->setDesc(get(DestIsHigh ? SystemZ::RISBHH : SystemZ::RISBLL));
+ else {
+ MI->setDesc(get(DestIsHigh ? SystemZ::RISBHL : SystemZ::RISBLH));
+ MI->getOperand(5).setImm(MI->getOperand(5).getImm() ^ 32);
+ }
+ return true;
+ }
+
case SystemZ::ADJDYNALLOC:
splitAdjDynAlloc(MI);
return true;
return SystemZII::Branch(SystemZII::BranchC, SystemZ::CCMASK_ICMP,
MI->getOperand(2).getImm(), &MI->getOperand(3));
+ case SystemZ::CLIJ:
+ case SystemZ::CLRJ:
+ return SystemZII::Branch(SystemZII::BranchCL, SystemZ::CCMASK_ICMP,
+ MI->getOperand(2).getImm(), &MI->getOperand(3));
+
case SystemZ::CGIJ:
case SystemZ::CGRJ:
return SystemZII::Branch(SystemZII::BranchCG, SystemZ::CCMASK_ICMP,
MI->getOperand(2).getImm(), &MI->getOperand(3));
+ case SystemZ::CLGIJ:
+ case SystemZ::CLGRJ:
+ return SystemZII::Branch(SystemZII::BranchCLG, SystemZ::CCMASK_ICMP,
+ MI->getOperand(2).getImm(), &MI->getOperand(3));
+
default:
llvm_unreachable("Unrecognized branch opcode");
}
unsigned &StoreOpcode) const {
if (RC == &SystemZ::GR32BitRegClass || RC == &SystemZ::ADDR32BitRegClass) {
LoadOpcode = SystemZ::L;
- StoreOpcode = SystemZ::ST32;
+ StoreOpcode = SystemZ::ST;
+ } else if (RC == &SystemZ::GRH32BitRegClass) {
+ LoadOpcode = SystemZ::LFH;
+ StoreOpcode = SystemZ::STFH;
+ } else if (RC == &SystemZ::GRX32BitRegClass) {
+ LoadOpcode = SystemZ::LMux;
+ StoreOpcode = SystemZ::STMux;
} else if (RC == &SystemZ::GR64BitRegClass ||
RC == &SystemZ::ADDR64BitRegClass) {
LoadOpcode = SystemZ::LG;
return MI && isInt<8>(MI->getOperand(1).getImm()) ? SystemZ::CIJ : 0;
case SystemZ::CGHI:
return MI && isInt<8>(MI->getOperand(1).getImm()) ? SystemZ::CGIJ : 0;
+ case SystemZ::CLR:
+ return SystemZ::CLRJ;
+ case SystemZ::CLGR:
+ return SystemZ::CLGRJ;
+ case SystemZ::CLFI:
+ return MI && isUInt<8>(MI->getOperand(1).getImm()) ? SystemZ::CLIJ : 0;
+ case SystemZ::CLGFI:
+ return MI && isUInt<8>(MI->getOperand(1).getImm()) ? SystemZ::CLGIJ : 0;
default:
return 0;
}