Re: [PATCH v3 1/8] ARM: KVM: Initial skeleton to compile KVM support

[Avi Kivity <avi@redhat.com>]

On 06/05/2011 07:03 PM, Christoffer Dall wrote:
> >>  +
> >>  +       /* Banked FIQ registers r8-r12 */
> >>  +       if (reg_num>= 8&&    reg_num<= 12) {
> >>  +               if (mode == MODE_FIQ) {
> >>  +                       reg_idx = reg_num - 8; /* 0=r8, ..., 4=r12 */
> >>  +                       return&(regs->fiq_regs[reg_idx]);
> >>  +               } else
> >>  +                       return&(regs->usr_regs[reg_num]);
> >>  +       }
> >
> >  You could have a static 2D array indexed by mode and register number,
> >  returning an offsetof() into the vcpu structure.
>
> You think it's simpler or faster? I don't quite see the incentive.
> It's not going to be called a whole lot given the Virt. Extensions.

Yes (those are mostly unpredictable branches), and clearer as well.  But 
I agree it's a rare-enough path that doesn't need optimizing, certainly 
not prematurely.

-- 
error compiling committee.c: too many arguments to function