Re: [Xen-ia64-devel] PATCH [resend]: more case handled in mmio.c

[Tristan Gingold <Tristan.Gingold@xxxxxxxx>]

Le Vendredi 14 Avril 2006 16:37, Xu, Anthony a écrit :
> From: Tristan Gingold
>
> >Sent: 2006?4?14? 19:36
> >To: xen-ia64-devel@xxxxxxxxxxxxxxxxxxx; Alex Williamson
> >Subject: [Xen-ia64-devel] PATCH [resend]: more case handled in mmio.c
> >
> >Hi,
> >
> >thanks to Anthony comments, I have rewritten getfpregs.
> >
> >Tested by booting domVTI.
>
> I don't think getfpreg will be executed even domVTI is booting. :-).
Yes, but don't worry I have tested it.

> Currently, every time domain switch, all floating registers are saved.
> So lazy floating registers save/restore algorithm is not implemented in
> xen.
Right.

> It's not ease. :-).
> When domain switch happens, below function is called.
> /* load f32-f127 from FPH */
> static inline void
> ia64_load_fpu (struct ia64_fpreg *fph) {
>       ia64_fph_enable();
>       __ia64_load_fpu(fph);
>       ia64_fph_disable();
> }
> Fph is disabled, and if getfpreg is called, disabled floating
> Register fault may happen.
Correct, but this doesn't happen in the mmio case, because fp_reg_disabled is 
tested before doing the memory access.

> Because all fr are saved at domain switch, seems VMM don't
> Need to disable fph.
Yes.

> My question is,
> Do we need to implement lazy fr algorithm?
Only if performance are improved !
I will add this on my todo list.

> If yes, we need to handle disabled fr fault and not need to save
> all fr at domain switch.
Yes.

> If no, we can simply remove all ia64_fph_enable() and ia64_fph_disable()
> in VMM.
Don't remove ia64_fph_enable, or else we won't be able to save fp regs!

> What's your option?

Tristan.


_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel