{ X86::VSQRTPSYr_Int, X86::VSQRTPSYm_Int, TB_ALIGN_32 },
{ X86::VBROADCASTSSYrr, X86::VBROADCASTSSYrm, TB_NO_REVERSE },
{ X86::VBROADCASTSDYrr, X86::VBROADCASTSDYrm, TB_NO_REVERSE },
+
+ // BMI/BMI2 foldable instructions
+ { X86::RORX32ri, X86::RORX32mi, 0 },
+ { X86::RORX64ri, X86::RORX64mi, 0 },
+ { X86::SARX32rr, X86::SARX32rm, 0 },
+ { X86::SARX64rr, X86::SARX64rm, 0 },
+ { X86::SHRX32rr, X86::SHRX32rm, 0 },
+ { X86::SHRX64rr, X86::SHRX64rm, 0 },
+ { X86::SHLX32rr, X86::SHLX32rm, 0 },
+ { X86::SHLX64rr, X86::SHLX64rm, 0 },
};
for (unsigned i = 0, e = array_lengthof(OpTbl1); i != e; ++i) {
{ X86::VPUNPCKLWDYrr, X86::VPUNPCKLWDYrm, TB_ALIGN_32 },
{ X86::VPXORYrr, X86::VPXORYrm, TB_ALIGN_32 },
// FIXME: add AVX 256-bit foldable instructions
+
+ // FMA4 foldable patterns
+ { X86::VFMADDSS4rr, X86::VFMADDSS4mr, 0 },
+ { X86::VFMADDSD4rr, X86::VFMADDSD4mr, 0 },
+ { X86::VFMADDPS4rr, X86::VFMADDPS4mr, TB_ALIGN_16 },
+ { X86::VFMADDPD4rr, X86::VFMADDPD4mr, TB_ALIGN_16 },
+ { X86::VFMADDPS4rrY, X86::VFMADDPS4mrY, TB_ALIGN_32 },
+ { X86::VFMADDPD4rrY, X86::VFMADDPD4mrY, TB_ALIGN_32 },
+ { X86::VFNMADDSS4rr, X86::VFNMADDSS4mr, 0 },
+ { X86::VFNMADDSD4rr, X86::VFNMADDSD4mr, 0 },
+ { X86::VFNMADDPS4rr, X86::VFNMADDPS4mr, TB_ALIGN_16 },
+ { X86::VFNMADDPD4rr, X86::VFNMADDPD4mr, TB_ALIGN_16 },
+ { X86::VFNMADDPS4rrY, X86::VFNMADDPS4mrY, TB_ALIGN_32 },
+ { X86::VFNMADDPD4rrY, X86::VFNMADDPD4mrY, TB_ALIGN_32 },
+ { X86::VFMSUBSS4rr, X86::VFMSUBSS4mr, 0 },
+ { X86::VFMSUBSD4rr, X86::VFMSUBSD4mr, 0 },
+ { X86::VFMSUBPS4rr, X86::VFMSUBPS4mr, TB_ALIGN_16 },
+ { X86::VFMSUBPD4rr, X86::VFMSUBPD4mr, TB_ALIGN_16 },
+ { X86::VFMSUBPS4rrY, X86::VFMSUBPS4mrY, TB_ALIGN_32 },
+ { X86::VFMSUBPD4rrY, X86::VFMSUBPD4mrY, TB_ALIGN_32 },
+ { X86::VFNMSUBSS4rr, X86::VFNMSUBSS4mr, 0 },
+ { X86::VFNMSUBSD4rr, X86::VFNMSUBSD4mr, 0 },
+ { X86::VFNMSUBPS4rr, X86::VFNMSUBPS4mr, TB_ALIGN_16 },
+ { X86::VFNMSUBPD4rr, X86::VFNMSUBPD4mr, TB_ALIGN_16 },
+ { X86::VFNMSUBPS4rrY, X86::VFNMSUBPS4mrY, TB_ALIGN_32 },
+ { X86::VFNMSUBPD4rrY, X86::VFNMSUBPD4mrY, TB_ALIGN_32 },
+ { X86::VFMADDSUBPS4rr, X86::VFMADDSUBPS4mr, TB_ALIGN_16 },
+ { X86::VFMADDSUBPD4rr, X86::VFMADDSUBPD4mr, TB_ALIGN_16 },
+ { X86::VFMADDSUBPS4rrY, X86::VFMADDSUBPS4mrY, TB_ALIGN_32 },
+ { X86::VFMADDSUBPD4rrY, X86::VFMADDSUBPD4mrY, TB_ALIGN_32 },
+ { X86::VFMSUBADDPS4rr, X86::VFMSUBADDPS4mr, TB_ALIGN_16 },
+ { X86::VFMSUBADDPD4rr, X86::VFMSUBADDPD4mr, TB_ALIGN_16 },
+ { X86::VFMSUBADDPS4rrY, X86::VFMSUBADDPS4mrY, TB_ALIGN_32 },
+ { X86::VFMSUBADDPD4rrY, X86::VFMSUBADDPD4mrY, TB_ALIGN_32 },
+
+ // BMI/BMI2 foldable instructions
+ { X86::MULX32rr, X86::MULX32rm, 0 },
+ { X86::MULX64rr, X86::MULX64rm, 0 },
};
for (unsigned i = 0, e = array_lengthof(OpTbl2); i != e; ++i) {
{ X86::VFMSUBADDPDr132rY, X86::VFMSUBADDPDr132mY, TB_ALIGN_32 },
{ X86::VFMSUBADDPSr213rY, X86::VFMSUBADDPSr213mY, TB_ALIGN_32 },
{ X86::VFMSUBADDPDr213rY, X86::VFMSUBADDPDr213mY, TB_ALIGN_32 },
+
+ // FMA4 foldable patterns
+ { X86::VFMADDSS4rr, X86::VFMADDSS4rm, 0 },
+ { X86::VFMADDSD4rr, X86::VFMADDSD4rm, 0 },
+ { X86::VFMADDPS4rr, X86::VFMADDPS4rm, TB_ALIGN_16 },
+ { X86::VFMADDPD4rr, X86::VFMADDPD4rm, TB_ALIGN_16 },
+ { X86::VFMADDPS4rrY, X86::VFMADDPS4rmY, TB_ALIGN_32 },
+ { X86::VFMADDPD4rrY, X86::VFMADDPD4rmY, TB_ALIGN_32 },
+ { X86::VFNMADDSS4rr, X86::VFNMADDSS4rm, 0 },
+ { X86::VFNMADDSD4rr, X86::VFNMADDSD4rm, 0 },
+ { X86::VFNMADDPS4rr, X86::VFNMADDPS4rm, TB_ALIGN_16 },
+ { X86::VFNMADDPD4rr, X86::VFNMADDPD4rm, TB_ALIGN_16 },
+ { X86::VFNMADDPS4rrY, X86::VFNMADDPS4rmY, TB_ALIGN_32 },
+ { X86::VFNMADDPD4rrY, X86::VFNMADDPD4rmY, TB_ALIGN_32 },
+ { X86::VFMSUBSS4rr, X86::VFMSUBSS4rm, 0 },
+ { X86::VFMSUBSD4rr, X86::VFMSUBSD4rm, 0 },
+ { X86::VFMSUBPS4rr, X86::VFMSUBPS4rm, TB_ALIGN_16 },
+ { X86::VFMSUBPD4rr, X86::VFMSUBPD4rm, TB_ALIGN_16 },
+ { X86::VFMSUBPS4rrY, X86::VFMSUBPS4rmY, TB_ALIGN_32 },
+ { X86::VFMSUBPD4rrY, X86::VFMSUBPD4rmY, TB_ALIGN_32 },
+ { X86::VFNMSUBSS4rr, X86::VFNMSUBSS4rm, 0 },
+ { X86::VFNMSUBSD4rr, X86::VFNMSUBSD4rm, 0 },
+ { X86::VFNMSUBPS4rr, X86::VFNMSUBPS4rm, TB_ALIGN_16 },
+ { X86::VFNMSUBPD4rr, X86::VFNMSUBPD4rm, TB_ALIGN_16 },
+ { X86::VFNMSUBPS4rrY, X86::VFNMSUBPS4rmY, TB_ALIGN_32 },
+ { X86::VFNMSUBPD4rrY, X86::VFNMSUBPD4rmY, TB_ALIGN_32 },
+ { X86::VFMADDSUBPS4rr, X86::VFMADDSUBPS4rm, TB_ALIGN_16 },
+ { X86::VFMADDSUBPD4rr, X86::VFMADDSUBPD4rm, TB_ALIGN_16 },
+ { X86::VFMADDSUBPS4rrY, X86::VFMADDSUBPS4rmY, TB_ALIGN_32 },
+ { X86::VFMADDSUBPD4rrY, X86::VFMADDSUBPD4rmY, TB_ALIGN_32 },
+ { X86::VFMSUBADDPS4rr, X86::VFMSUBADDPS4rm, TB_ALIGN_16 },
+ { X86::VFMSUBADDPD4rr, X86::VFMSUBADDPD4rm, TB_ALIGN_16 },
+ { X86::VFMSUBADDPS4rrY, X86::VFMSUBADDPS4rmY, TB_ALIGN_32 },
+ { X86::VFMSUBADDPD4rrY, X86::VFMSUBADDPD4rmY, TB_ALIGN_32 },
};
for (unsigned i = 0, e = array_lengthof(OpTbl3); i != e; ++i) {
AliasAnalysis *AA) const {
switch (MI->getOpcode()) {
default: break;
- case X86::MOV8rm:
- case X86::MOV16rm:
- case X86::MOV32rm:
- case X86::MOV64rm:
- case X86::LD_Fp64m:
- case X86::MOVSSrm:
- case X86::MOVSDrm:
- case X86::MOVAPSrm:
- case X86::MOVUPSrm:
- case X86::MOVAPDrm:
- case X86::MOVDQArm:
- case X86::VMOVSSrm:
- case X86::VMOVSDrm:
- case X86::VMOVAPSrm:
- case X86::VMOVUPSrm:
- case X86::VMOVAPDrm:
- case X86::VMOVDQArm:
- case X86::VMOVAPSYrm:
- case X86::VMOVUPSYrm:
- case X86::VMOVAPDYrm:
- case X86::VMOVDQAYrm:
- case X86::MMX_MOVD64rm:
- case X86::MMX_MOVQ64rm:
- case X86::FsVMOVAPSrm:
- case X86::FsVMOVAPDrm:
- case X86::FsMOVAPSrm:
- case X86::FsMOVAPDrm: {
- // Loads from constant pools are trivially rematerializable.
- if (MI->getOperand(1).isReg() &&
- MI->getOperand(2).isImm() &&
- MI->getOperand(3).isReg() && MI->getOperand(3).getReg() == 0 &&
- MI->isInvariantLoad(AA)) {
- unsigned BaseReg = MI->getOperand(1).getReg();
- if (BaseReg == 0 || BaseReg == X86::RIP)
- return true;
- // Allow re-materialization of PIC load.
- if (!ReMatPICStubLoad && MI->getOperand(4).isGlobal())
- return false;
- const MachineFunction &MF = *MI->getParent()->getParent();
- const MachineRegisterInfo &MRI = MF.getRegInfo();
- return regIsPICBase(BaseReg, MRI);
- }
- return false;
+ case X86::MOV8rm:
+ case X86::MOV16rm:
+ case X86::MOV32rm:
+ case X86::MOV64rm:
+ case X86::LD_Fp64m:
+ case X86::MOVSSrm:
+ case X86::MOVSDrm:
+ case X86::MOVAPSrm:
+ case X86::MOVUPSrm:
+ case X86::MOVAPDrm:
+ case X86::MOVDQArm:
+ case X86::VMOVSSrm:
+ case X86::VMOVSDrm:
+ case X86::VMOVAPSrm:
+ case X86::VMOVUPSrm:
+ case X86::VMOVAPDrm:
+ case X86::VMOVDQArm:
+ case X86::VMOVAPSYrm:
+ case X86::VMOVUPSYrm:
+ case X86::VMOVAPDYrm:
+ case X86::VMOVDQAYrm:
+ case X86::MMX_MOVD64rm:
+ case X86::MMX_MOVQ64rm:
+ case X86::FsVMOVAPSrm:
+ case X86::FsVMOVAPDrm:
+ case X86::FsMOVAPSrm:
+ case X86::FsMOVAPDrm: {
+ // Loads from constant pools are trivially rematerializable.
+ if (MI->getOperand(1).isReg() &&
+ MI->getOperand(2).isImm() &&
+ MI->getOperand(3).isReg() && MI->getOperand(3).getReg() == 0 &&
+ MI->isInvariantLoad(AA)) {
+ unsigned BaseReg = MI->getOperand(1).getReg();
+ if (BaseReg == 0 || BaseReg == X86::RIP)
+ return true;
+ // Allow re-materialization of PIC load.
+ if (!ReMatPICStubLoad && MI->getOperand(4).isGlobal())
+ return false;
+ const MachineFunction &MF = *MI->getParent()->getParent();
+ const MachineRegisterInfo &MRI = MF.getRegInfo();
+ return regIsPICBase(BaseReg, MRI);
}
+ return false;
+ }
- case X86::LEA32r:
- case X86::LEA64r: {
- if (MI->getOperand(2).isImm() &&
- MI->getOperand(3).isReg() && MI->getOperand(3).getReg() == 0 &&
- !MI->getOperand(4).isReg()) {
- // lea fi#, lea GV, etc. are all rematerializable.
- if (!MI->getOperand(1).isReg())
- return true;
- unsigned BaseReg = MI->getOperand(1).getReg();
- if (BaseReg == 0)
- return true;
- // Allow re-materialization of lea PICBase + x.
- const MachineFunction &MF = *MI->getParent()->getParent();
- const MachineRegisterInfo &MRI = MF.getRegInfo();
- return regIsPICBase(BaseReg, MRI);
- }
- return false;
- }
+ case X86::LEA32r:
+ case X86::LEA64r: {
+ if (MI->getOperand(2).isImm() &&
+ MI->getOperand(3).isReg() && MI->getOperand(3).getReg() == 0 &&
+ !MI->getOperand(4).isReg()) {
+ // lea fi#, lea GV, etc. are all rematerializable.
+ if (!MI->getOperand(1).isReg())
+ return true;
+ unsigned BaseReg = MI->getOperand(1).getReg();
+ if (BaseReg == 0)
+ return true;
+ // Allow re-materialization of lea PICBase + x.
+ const MachineFunction &MF = *MI->getParent()->getParent();
+ const MachineRegisterInfo &MRI = MF.getRegInfo();
+ return regIsPICBase(BaseReg, MRI);
+ }
+ return false;
+ }
}
// All other instructions marked M_REMATERIALIZABLE are always trivially
MachineInstr *MI = MBBI;
MachineFunction &MF = *MI->getParent()->getParent();
// All instructions input are two-addr instructions. Get the known operands.
- unsigned Dest = MI->getOperand(0).getReg();
- unsigned Src = MI->getOperand(1).getReg();
- bool isDead = MI->getOperand(0).isDead();
- bool isKill = MI->getOperand(1).isKill();
+ const MachineOperand &Dest = MI->getOperand(0);
+ const MachineOperand &Src = MI->getOperand(1);
MachineInstr *NewMI = NULL;
// FIXME: 16-bit LEA's are really slow on Athlons, but not bad on P4's. When
unsigned B = MI->getOperand(1).getReg();
unsigned C = MI->getOperand(2).getReg();
if (B != C) return 0;
- unsigned A = MI->getOperand(0).getReg();
unsigned M = MI->getOperand(3).getImm();
NewMI = BuildMI(MF, MI->getDebugLoc(), get(X86::PSHUFDri))
- .addReg(A, RegState::Define | getDeadRegState(isDead))
- .addReg(B, getKillRegState(isKill)).addImm(M);
+ .addOperand(Dest).addOperand(Src).addImm(M);
break;
}
case X86::SHUFPDrri: {
unsigned B = MI->getOperand(1).getReg();
unsigned C = MI->getOperand(2).getReg();
if (B != C) return 0;
- unsigned A = MI->getOperand(0).getReg();
unsigned M = MI->getOperand(3).getImm();
// Convert to PSHUFD mask.
M = ((M & 1) << 1) | ((M & 1) << 3) | ((M & 2) << 4) | ((M & 2) << 6)| 0x44;
NewMI = BuildMI(MF, MI->getDebugLoc(), get(X86::PSHUFDri))
- .addReg(A, RegState::Define | getDeadRegState(isDead))
- .addReg(B, getKillRegState(isKill)).addImm(M);
+ .addOperand(Dest).addOperand(Src).addImm(M);
break;
}
case X86::SHL64ri: {
if (ShAmt == 0 || ShAmt >= 4) return 0;
// LEA can't handle RSP.
- if (TargetRegisterInfo::isVirtualRegister(Src) &&
- !MF.getRegInfo().constrainRegClass(Src, &X86::GR64_NOSPRegClass))
+ if (TargetRegisterInfo::isVirtualRegister(Src.getReg()) &&
+ !MF.getRegInfo().constrainRegClass(Src.getReg(),
+ &X86::GR64_NOSPRegClass))
return 0;
NewMI = BuildMI(MF, MI->getDebugLoc(), get(X86::LEA64r))
- .addReg(Dest, RegState::Define | getDeadRegState(isDead))
- .addReg(0).addImm(1 << ShAmt)
- .addReg(Src, getKillRegState(isKill))
- .addImm(0).addReg(0);
+ .addOperand(Dest)
+ .addReg(0).addImm(1 << ShAmt).addOperand(Src).addImm(0).addReg(0);
break;
}
case X86::SHL32ri: {
if (ShAmt == 0 || ShAmt >= 4) return 0;
// LEA can't handle ESP.
- if (TargetRegisterInfo::isVirtualRegister(Src) &&
- !MF.getRegInfo().constrainRegClass(Src, &X86::GR32_NOSPRegClass))
+ if (TargetRegisterInfo::isVirtualRegister(Src.getReg()) &&
+ !MF.getRegInfo().constrainRegClass(Src.getReg(),
+ &X86::GR32_NOSPRegClass))
return 0;
unsigned Opc = is64Bit ? X86::LEA64_32r : X86::LEA32r;
NewMI = BuildMI(MF, MI->getDebugLoc(), get(Opc))
- .addReg(Dest, RegState::Define | getDeadRegState(isDead))
- .addReg(0).addImm(1 << ShAmt)
- .addReg(Src, getKillRegState(isKill)).addImm(0).addReg(0);
+ .addOperand(Dest)
+ .addReg(0).addImm(1 << ShAmt).addOperand(Src).addImm(0).addReg(0);
break;
}
case X86::SHL16ri: {
if (DisableLEA16)
return is64Bit ? convertToThreeAddressWithLEA(MIOpc, MFI, MBBI, LV) : 0;
NewMI = BuildMI(MF, MI->getDebugLoc(), get(X86::LEA16r))
- .addReg(Dest, RegState::Define | getDeadRegState(isDead))
- .addReg(0).addImm(1 << ShAmt)
- .addReg(Src, getKillRegState(isKill))
- .addImm(0).addReg(0);
+ .addOperand(Dest)
+ .addReg(0).addImm(1 << ShAmt).addOperand(Src).addImm(0).addReg(0);
break;
}
default: {
(const TargetRegisterClass*)&X86::GR32_NOSPRegClass;
// LEA can't handle RSP.
- if (TargetRegisterInfo::isVirtualRegister(Src) &&
- !MF.getRegInfo().constrainRegClass(Src, RC))
+ if (TargetRegisterInfo::isVirtualRegister(Src.getReg()) &&
+ !MF.getRegInfo().constrainRegClass(Src.getReg(), RC))
return 0;
- NewMI = addRegOffset(BuildMI(MF, MI->getDebugLoc(), get(Opc))
- .addReg(Dest, RegState::Define |
- getDeadRegState(isDead)),
- Src, isKill, 1);
+ NewMI = addOffset(BuildMI(MF, MI->getDebugLoc(), get(Opc))
+ .addOperand(Dest).addOperand(Src), 1);
break;
}
case X86::INC16r:
if (DisableLEA16)
return is64Bit ? convertToThreeAddressWithLEA(MIOpc, MFI, MBBI, LV) : 0;
assert(MI->getNumOperands() >= 2 && "Unknown inc instruction!");
- NewMI = addRegOffset(BuildMI(MF, MI->getDebugLoc(), get(X86::LEA16r))
- .addReg(Dest, RegState::Define |
- getDeadRegState(isDead)),
- Src, isKill, 1);
+ NewMI = addOffset(BuildMI(MF, MI->getDebugLoc(), get(X86::LEA16r))
+ .addOperand(Dest).addOperand(Src), 1);
break;
case X86::DEC64r:
case X86::DEC32r:
(const TargetRegisterClass*)&X86::GR64_NOSPRegClass :
(const TargetRegisterClass*)&X86::GR32_NOSPRegClass;
// LEA can't handle RSP.
- if (TargetRegisterInfo::isVirtualRegister(Src) &&
- !MF.getRegInfo().constrainRegClass(Src, RC))
+ if (TargetRegisterInfo::isVirtualRegister(Src.getReg()) &&
+ !MF.getRegInfo().constrainRegClass(Src.getReg(), RC))
return 0;
- NewMI = addRegOffset(BuildMI(MF, MI->getDebugLoc(), get(Opc))
- .addReg(Dest, RegState::Define |
- getDeadRegState(isDead)),
- Src, isKill, -1);
+ NewMI = addOffset(BuildMI(MF, MI->getDebugLoc(), get(Opc))
+ .addOperand(Dest).addOperand(Src), -1);
break;
}
case X86::DEC16r:
if (DisableLEA16)
return is64Bit ? convertToThreeAddressWithLEA(MIOpc, MFI, MBBI, LV) : 0;
assert(MI->getNumOperands() >= 2 && "Unknown dec instruction!");
- NewMI = addRegOffset(BuildMI(MF, MI->getDebugLoc(), get(X86::LEA16r))
- .addReg(Dest, RegState::Define |
- getDeadRegState(isDead)),
- Src, isKill, -1);
+ NewMI = addOffset(BuildMI(MF, MI->getDebugLoc(), get(X86::LEA16r))
+ .addOperand(Dest).addOperand(Src), -1);
break;
case X86::ADD64rr:
case X86::ADD64rr_DB:
return 0;
NewMI = addRegReg(BuildMI(MF, MI->getDebugLoc(), get(Opc))
- .addReg(Dest, RegState::Define |
- getDeadRegState(isDead)),
- Src, isKill, Src2, isKill2);
+ .addOperand(Dest),
+ Src.getReg(), Src.isKill(), Src2, isKill2);
// Preserve undefness of the operands.
bool isUndef = MI->getOperand(1).isUndef();
unsigned Src2 = MI->getOperand(2).getReg();
bool isKill2 = MI->getOperand(2).isKill();
NewMI = addRegReg(BuildMI(MF, MI->getDebugLoc(), get(X86::LEA16r))
- .addReg(Dest, RegState::Define |
- getDeadRegState(isDead)),
- Src, isKill, Src2, isKill2);
+ .addOperand(Dest),
+ Src.getReg(), Src.isKill(), Src2, isKill2);
+
+ // Preserve undefness of the operands.
+ bool isUndef = MI->getOperand(1).isUndef();
+ bool isUndef2 = MI->getOperand(2).isUndef();
+ NewMI->getOperand(1).setIsUndef(isUndef);
+ NewMI->getOperand(3).setIsUndef(isUndef2);
+
if (LV && isKill2)
LV->replaceKillInstruction(Src2, MI, NewMI);
break;
case X86::ADD64ri32_DB:
case X86::ADD64ri8_DB:
assert(MI->getNumOperands() >= 3 && "Unknown add instruction!");
- NewMI = addRegOffset(BuildMI(MF, MI->getDebugLoc(), get(X86::LEA64r))
- .addReg(Dest, RegState::Define |
- getDeadRegState(isDead)),
- Src, isKill, MI->getOperand(2).getImm());
+ NewMI = addOffset(BuildMI(MF, MI->getDebugLoc(), get(X86::LEA64r))
+ .addOperand(Dest).addOperand(Src),
+ MI->getOperand(2).getImm());
break;
case X86::ADD32ri:
case X86::ADD32ri8:
case X86::ADD32ri8_DB: {
assert(MI->getNumOperands() >= 3 && "Unknown add instruction!");
unsigned Opc = is64Bit ? X86::LEA64_32r : X86::LEA32r;
- NewMI = addRegOffset(BuildMI(MF, MI->getDebugLoc(), get(Opc))
- .addReg(Dest, RegState::Define |
- getDeadRegState(isDead)),
- Src, isKill, MI->getOperand(2).getImm());
+ NewMI = addOffset(BuildMI(MF, MI->getDebugLoc(), get(Opc))
+ .addOperand(Dest).addOperand(Src),
+ MI->getOperand(2).getImm());
break;
}
case X86::ADD16ri:
if (DisableLEA16)
return is64Bit ? convertToThreeAddressWithLEA(MIOpc, MFI, MBBI, LV) : 0;
assert(MI->getNumOperands() >= 3 && "Unknown add instruction!");
- NewMI = addRegOffset(BuildMI(MF, MI->getDebugLoc(), get(X86::LEA16r))
- .addReg(Dest, RegState::Define |
- getDeadRegState(isDead)),
- Src, isKill, MI->getOperand(2).getImm());
+ NewMI = addOffset(BuildMI(MF, MI->getDebugLoc(), get(X86::LEA16r))
+ .addOperand(Dest).addOperand(Src),
+ MI->getOperand(2).getImm());
break;
}
}
if (!NewMI) return 0;
if (LV) { // Update live variables
- if (isKill)
- LV->replaceKillInstruction(Src, MI, NewMI);
- if (isDead)
- LV->replaceKillInstruction(Dest, MI, NewMI);
+ if (Src.isKill())
+ LV->replaceKillInstruction(Src.getReg(), MI, NewMI);
+ if (Dest.isDead())
+ LV->replaceKillInstruction(Dest.getReg(), MI, NewMI);
}
MFI->insert(MBBI, NewMI); // Insert the new inst
}
/// getCondFromCmovOpc - return condition code of a CMov opcode.
-static X86::CondCode getCondFromCMovOpc(unsigned Opc) {
+X86::CondCode X86::getCondFromCMovOpc(unsigned Opc) {
switch (Opc) {
default: return X86::COND_INVALID;
case X86::CMOVA16rm: case X86::CMOVA16rr: case X86::CMOVA32rm:
/// whether it has memory operand.
static unsigned getSETFromCond(X86::CondCode CC,
bool HasMemoryOperand) {
- static const unsigned Opc[16][2] = {
+ static const uint16_t Opc[16][2] = {
{ X86::SETAr, X86::SETAm },
{ X86::SETAEr, X86::SETAEm },
{ X86::SETBr, X86::SETBm },
/// register size in bytes, and operand type.
static unsigned getCMovFromCond(X86::CondCode CC, unsigned RegBytes,
bool HasMemoryOperand) {
- static const unsigned Opc[32][3] = {
+ static const uint16_t Opc[32][3] = {
{ X86::CMOVA16rr, X86::CMOVA32rr, X86::CMOVA64rr },
{ X86::CMOVAE16rr, X86::CMOVAE32rr, X86::CMOVAE64rr },
{ X86::CMOVB16rr, X86::CMOVB32rr, X86::CMOVB64rr },
// SrcReg(GR64) -> DestReg(VR64)
if (X86::GR64RegClass.contains(DestReg)) {
- if (X86::VR128RegClass.contains(SrcReg)) {
+ if (X86::VR128RegClass.contains(SrcReg))
// Copy from a VR128 register to a GR64 register.
return HasAVX ? X86::VMOVPQIto64rr : X86::MOVPQIto64rr;
- } else if (X86::VR64RegClass.contains(SrcReg)) {
+ if (X86::VR64RegClass.contains(SrcReg))
// Copy from a VR64 register to a GR64 register.
return X86::MOVSDto64rr;
- }
} else if (X86::GR64RegClass.contains(SrcReg)) {
// Copy from a GR64 register to a VR128 register.
if (X86::VR128RegClass.contains(DestReg))
return HasAVX ? X86::VMOV64toPQIrr : X86::MOV64toPQIrr;
// Copy from a GR64 register to a VR64 register.
- else if (X86::VR64RegClass.contains(DestReg))
+ if (X86::VR64RegClass.contains(DestReg))
return X86::MOV64toSDrr;
}
// SrcReg(GR32) -> DestReg(FR32)
if (X86::GR32RegClass.contains(DestReg) && X86::FR32RegClass.contains(SrcReg))
- // Copy from a FR32 register to a GR32 register.
- return HasAVX ? X86::VMOVSS2DIrr : X86::MOVSS2DIrr;
+ // Copy from a FR32 register to a GR32 register.
+ return HasAVX ? X86::VMOVSS2DIrr : X86::MOVSS2DIrr;
if (X86::FR32RegClass.contains(DestReg) && X86::GR32RegClass.contains(SrcReg))
- // Copy from a GR32 register to a FR32 register.
- return HasAVX ? X86::VMOVDI2SSrr : X86::MOVDI2SSrr;
+ // Copy from a GR32 register to a FR32 register.
+ return HasAVX ? X86::VMOVDI2SSrr : X86::MOVDI2SSrr;
return 0;
}
bool KillSrc) const {
// First deal with the normal symmetric copies.
bool HasAVX = TM.getSubtarget<X86Subtarget>().hasAVX();
- unsigned Opc = 0;
+ unsigned Opc;
if (X86::GR64RegClass.contains(DestReg, SrcReg))
Opc = X86::MOV64rr;
else if (X86::GR32RegClass.contains(DestReg, SrcReg))
BuildMI(MBB, MI, DL, get(X86::PUSHF64));
BuildMI(MBB, MI, DL, get(X86::POP64r), DestReg);
return;
- } else if (X86::GR32RegClass.contains(DestReg)) {
+ }
+ if (X86::GR32RegClass.contains(DestReg)) {
BuildMI(MBB, MI, DL, get(X86::PUSHF32));
BuildMI(MBB, MI, DL, get(X86::POP32r), DestReg);
return;
.addReg(SrcReg, getKillRegState(KillSrc));
BuildMI(MBB, MI, DL, get(X86::POPF64));
return;
- } else if (X86::GR32RegClass.contains(SrcReg)) {
+ }
+ if (X86::GR32RegClass.contains(SrcReg)) {
BuildMI(MBB, MI, DL, get(X86::PUSH32r))
.addReg(SrcReg, getKillRegState(KillSrc));
BuildMI(MBB, MI, DL, get(X86::POPF32));
case X86::SUB8ri: case X86::SUB64rr: case X86::SUB32rr:
case X86::SUB16rr: case X86::SUB8rr: case X86::SUB64rm:
case X86::SUB32rm: case X86::SUB16rm: case X86::SUB8rm:
+ case X86::DEC64r: case X86::DEC32r: case X86::DEC16r: case X86::DEC8r:
+ case X86::DEC64m: case X86::DEC32m: case X86::DEC16m: case X86::DEC8m:
+ case X86::DEC64_32r: case X86::DEC64_16r:
+ case X86::DEC64_32m: case X86::DEC64_16m:
case X86::ADD64ri32: case X86::ADD64ri8: case X86::ADD32ri:
case X86::ADD32ri8: case X86::ADD16ri: case X86::ADD16ri8:
case X86::ADD8ri: case X86::ADD64rr: case X86::ADD32rr:
case X86::ADD16rr: case X86::ADD8rr: case X86::ADD64rm:
case X86::ADD32rm: case X86::ADD16rm: case X86::ADD8rm:
+ case X86::INC64r: case X86::INC32r: case X86::INC16r: case X86::INC8r:
+ case X86::INC64m: case X86::INC32m: case X86::INC16m: case X86::INC8m:
+ case X86::INC64_32r: case X86::INC64_16r:
+ case X86::INC64_32m: case X86::INC64_16m:
case X86::AND64ri32: case X86::AND64ri8: case X86::AND32ri:
case X86::AND32ri8: case X86::AND16ri: case X86::AND16ri8:
case X86::AND8ri: case X86::AND64rr: case X86::AND32rr:
if (OldCC != X86::COND_INVALID)
OpcIsSET = true;
else
- OldCC = getCondFromCMovOpc(Instr.getOpcode());
+ OldCC = X86::getCondFromCMovOpc(Instr.getOpcode());
}
if (OldCC == X86::COND_INVALID) return false;
}
Sub->getParent()->insert(MachineBasicBlock::iterator(Sub), Movr0Inst);
}
- // Make sure Sub instruction defines EFLAGS.
+ // Make sure Sub instruction defines EFLAGS and mark the def live.
+ unsigned LastOperand = Sub->getNumOperands() - 1;
assert(Sub->getNumOperands() >= 2 &&
- Sub->getOperand(Sub->getNumOperands()-1).isReg() &&
- Sub->getOperand(Sub->getNumOperands()-1).getReg() == X86::EFLAGS &&
+ Sub->getOperand(LastOperand).isReg() &&
+ Sub->getOperand(LastOperand).getReg() == X86::EFLAGS &&
"EFLAGS should be the last operand of SUB, ADD, OR, XOR, AND");
- Sub->getOperand(Sub->getNumOperands()-1).setIsDef(true);
+ Sub->getOperand(LastOperand).setIsDef(true);
+ Sub->getOperand(LastOperand).setIsDead(false);
CmpInstr->eraseFromParent();
// Modify the condition code of instructions in OpsToUpdate.
bool X86InstrInfo::expandPostRAPseudo(MachineBasicBlock::iterator MI) const {
bool HasAVX = TM.getSubtarget<X86Subtarget>().hasAVX();
switch (MI->getOpcode()) {
+ case X86::SETB_C8r:
+ return Expand2AddrUndef(MI, get(X86::SBB8rr));
+ case X86::SETB_C16r:
+ return Expand2AddrUndef(MI, get(X86::SBB16rr));
+ case X86::SETB_C32r:
+ return Expand2AddrUndef(MI, get(X86::SBB32rr));
+ case X86::SETB_C64r:
+ return Expand2AddrUndef(MI, get(X86::SBB64rr));
case X86::V_SET0:
case X86::FsFLD0SS:
case X86::FsFLD0SD:
return Expand2AddrUndef(MI, get(HasAVX ? X86::VXORPSrr : X86::XORPSrr));
+ case X86::AVX_SET0:
+ assert(HasAVX && "AVX not supported");
+ return Expand2AddrUndef(MI, get(X86::VXORPSYrr));
+ case X86::V_SETALLONES:
+ return Expand2AddrUndef(MI, get(HasAVX ? X86::VPCMPEQDrr : X86::PCMPEQDrr));
+ case X86::AVX2_SETALLONES:
+ return Expand2AddrUndef(MI, get(X86::VPCMPEQDYrr));
case X86::TEST8ri_NOREX:
MI->setDesc(get(X86::TEST8ri));
return true;
OpcodeTablePtr = &RegOp2MemOpTable2Addr;
isTwoAddrFold = true;
} else if (i == 0) { // If operand 0
- if (MI->getOpcode() == X86::MOV64r0)
- NewMI = MakeM0Inst(*this, X86::MOV64mi32, MOs, MI);
- else if (MI->getOpcode() == X86::MOV32r0)
- NewMI = MakeM0Inst(*this, X86::MOV32mi, MOs, MI);
- else if (MI->getOpcode() == X86::MOV16r0)
- NewMI = MakeM0Inst(*this, X86::MOV16mi, MOs, MI);
- else if (MI->getOpcode() == X86::MOV8r0)
- NewMI = MakeM0Inst(*this, X86::MOV8mi, MOs, MI);
+ unsigned Opc = 0;
+ switch (MI->getOpcode()) {
+ default: break;
+ case X86::MOV64r0: Opc = X86::MOV64mi32; break;
+ case X86::MOV32r0: Opc = X86::MOV32mi; break;
+ case X86::MOV16r0: Opc = X86::MOV16mi; break;
+ case X86::MOV8r0: Opc = X86::MOV8mi; break;
+ }
+ if (Opc)
+ NewMI = MakeM0Inst(*this, Opc, MOs, MI);
if (NewMI)
return NewMI;
// Unless optimizing for size, don't fold to avoid partial
// register update stalls
- if (!MF.getFunction()->hasFnAttr(Attribute::OptimizeForSize) &&
+ if (!MF.getFunction()->getFnAttributes().
+ hasAttribute(Attributes::OptimizeForSize) &&
hasPartialRegUpdate(MI->getOpcode()))
return 0;
// Unless optimizing for size, don't fold to avoid partial
// register update stalls
- if (!MF.getFunction()->hasFnAttr(Attribute::OptimizeForSize) &&
+ if (!MF.getFunction()->getFnAttributes().
+ hasAttribute(Attributes::OptimizeForSize) &&
hasPartialRegUpdate(MI->getOpcode()))
return 0;
Alignment = (*LoadMI->memoperands_begin())->getAlignment();
else
switch (LoadMI->getOpcode()) {
- case X86::AVX_SET0PSY:
- case X86::AVX_SET0PDY:
case X86::AVX2_SETALLONES:
- case X86::AVX2_SET0:
+ case X86::AVX_SET0:
Alignment = 32;
break;
case X86::V_SET0:
case X86::V_SETALLONES:
- case X86::AVX_SETALLONES:
Alignment = 16;
break;
case X86::FsFLD0SD:
switch (LoadMI->getOpcode()) {
case X86::V_SET0:
case X86::V_SETALLONES:
- case X86::AVX_SET0PSY:
- case X86::AVX_SET0PDY:
- case X86::AVX_SETALLONES:
case X86::AVX2_SETALLONES:
- case X86::AVX2_SET0:
+ case X86::AVX_SET0:
case X86::FsFLD0SD:
case X86::FsFLD0SS: {
// Folding a V_SET0 or V_SETALLONES as a load, to ease register pressure.
Ty = Type::getFloatTy(MF.getFunction()->getContext());
else if (Opc == X86::FsFLD0SD)
Ty = Type::getDoubleTy(MF.getFunction()->getContext());
- else if (Opc == X86::AVX_SET0PSY || Opc == X86::AVX_SET0PDY)
- Ty = VectorType::get(Type::getFloatTy(MF.getFunction()->getContext()), 8);
- else if (Opc == X86::AVX2_SETALLONES || Opc == X86::AVX2_SET0)
+ else if (Opc == X86::AVX2_SETALLONES || Opc == X86::AVX_SET0)
Ty = VectorType::get(Type::getInt32Ty(MF.getFunction()->getContext()), 8);
else
Ty = VectorType::get(Type::getInt32Ty(MF.getFunction()->getContext()), 4);
- bool IsAllOnes = (Opc == X86::V_SETALLONES || Opc == X86::AVX_SETALLONES ||
- Opc == X86::AVX2_SETALLONES);
+ bool IsAllOnes = (Opc == X86::V_SETALLONES || Opc == X86::AVX2_SETALLONES);
const Constant *C = IsAllOnes ? Constant::getAllOnesValue(Ty) :
Constant::getNullValue(Ty);
unsigned CPI = MCP.getConstantPoolIndex(C, Alignment);
break;
}
default: {
+ if ((LoadMI->getOpcode() == X86::MOVSSrm ||
+ LoadMI->getOpcode() == X86::VMOVSSrm) &&
+ MF.getRegInfo().getRegClass(LoadMI->getOperand(0).getReg())->getSize()
+ > 4)
+ // These instructions only load 32 bits, we can't fold them if the
+ // destination register is wider than 32 bits (4 bytes).
+ return NULL;
+ if ((LoadMI->getOpcode() == X86::MOVSDrm ||
+ LoadMI->getOpcode() == X86::VMOVSDrm) &&
+ MF.getRegInfo().getRegClass(LoadMI->getOperand(0).getReg())->getSize()
+ > 8)
+ // These instructions only load 64 bits, we can't fold them if the
+ // destination register is wider than 64 bits (8 bytes).
+ return NULL;
+
// Folding a normal load. Just copy the load's address operands.
unsigned NumOps = LoadMI->getDesc().getNumOperands();
for (unsigned i = NumOps - X86::AddrNumOperands; i != NumOps; ++i)
OpcodeTablePtr = &RegOp2MemOpTable1;
} else if (OpNum == 2) {
OpcodeTablePtr = &RegOp2MemOpTable2;
+ } else if (OpNum == 3) {
+ OpcodeTablePtr = &RegOp2MemOpTable3;
}
if (OpcodeTablePtr && OpcodeTablePtr->count(Opc))
projects / oota-llvm.git / blobdiff