llvm dev - Jul 2009 - [LLVMdev] Bug in LiveIntervals? Please Examine

In LiveIntervals::processImplicitDefs() we have this:

      for (MachineRegisterInfo::use_iterator UI = mri_->use_begin(Reg),
             UE = mri_->use_end(); UI != UE; ) {
        MachineOperand &RMO = UI.getOperand();
        MachineInstr *RMI = &*UI;
        ++UI;
        MachineBasicBlock *RMBB = RMI->getParent();
        if (RMBB == MBB)
          continue;
        const TargetRegisterClass* RC = mri_->getRegClass(Reg);
        unsigned NewVReg = mri_->createVirtualRegister(RC);
        MachineInstrBuilder MIB           BuildMI(*RMBB, RMI,
RMI->getDebugLoc(),
                  tii_->get(TargetInstrInfo::IMPLICIT_DEF), NewVReg);
        (*MIB).getOperand(0).setIsUndef();
        RMO.setReg(NewVReg);
        RMO.setIsUndef();
        RMO.setIsKill();
      }

Souldn't there be a call to vrm.grow() after each call to 
mri->createVirtualRegister()?

                            -Dave

On Jul 17, 2009, at 7:57 AM, David Greene wrote:
> In LiveIntervals::processImplicitDefs() we have this:
>
>      for (MachineRegisterInfo::use_iterator UI = mri_->use_begin(Reg),
>             UE = mri_->use_end(); UI != UE; ) {
>        MachineOperand &RMO = UI.getOperand();
>        MachineInstr *RMI = &*UI;
>        ++UI;
>        MachineBasicBlock *RMBB = RMI->getParent();
>        if (RMBB == MBB)
>          continue;
>        const TargetRegisterClass* RC = mri_->getRegClass(Reg);
>        unsigned NewVReg = mri_->createVirtualRegister(RC);
>        MachineInstrBuilder MIB >          BuildMI(*RMBB, RMI,
RMI->getDebugLoc(),
>                  tii_->get(TargetInstrInfo::IMPLICIT_DEF), NewVReg);
>        (*MIB).getOperand(0).setIsUndef();
>        RMO.setReg(NewVReg);
>        RMO.setIsUndef();
>        RMO.setIsKill();
>      }
>
> Souldn't there be a call to vrm.grow() after each call to
> mri->createVirtualRegister()?
No. This function is run during the analysis stage. vrm is actually  
owned by the register allocator which then passed it into  
liveintervals. Is there an actual problem that you run into?

Evan
>
>                            -Dave
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev

On Friday 17 July 2009 14:00, Evan Cheng wrote:
> On Jul 17, 2009, at 7:57 AM, David Greene wrote:
> > In LiveIntervals::processImplicitDefs() we have this:
> >
> >      for (MachineRegisterInfo::use_iterator UI =
mri_->use_begin(Reg),
> >             UE = mri_->use_end(); UI != UE; ) {
> >        MachineOperand &RMO = UI.getOperand();
> >        MachineInstr *RMI = &*UI;
> >        ++UI;
> >        MachineBasicBlock *RMBB = RMI->getParent();
> >        if (RMBB == MBB)
> >          continue;
> >        const TargetRegisterClass* RC = mri_->getRegClass(Reg);
> >        unsigned NewVReg = mri_->createVirtualRegister(RC);
> >        MachineInstrBuilder MIB > >          BuildMI(*RMBB, RMI,
RMI->getDebugLoc(),
> >                  tii_->get(TargetInstrInfo::IMPLICIT_DEF),
NewVReg);
> >        (*MIB).getOperand(0).setIsUndef();
> >        RMO.setReg(NewVReg);
> >        RMO.setIsUndef();
> >        RMO.setIsKill();
> >      }
> >
> > Souldn't there be a call to vrm.grow() after each call to
> > mri->createVirtualRegister()?
>
> No. This function is run during the analysis stage. vrm is actually
> owned by the register allocator which then passed it into
> liveintervals. Is there an actual problem that you run into?
No, I was just examining some bugfixes we've made here.  Looks like
all the places we found missing vrm.grow() have them now.  I just
diodn't know about this one.  Thanks.

                              -Dave