lars brinkhoff wrote:
>
> "Kevin P. Lawton" <[EMAIL PROTECTED]> writes:
> > Since we'll have to virtualize ring0 kernel code, it's
> > simple to detect idle code such as:
> > hlt
> > ... or for that matter, maybe even guest specific OS calls that
> > call an idler function. So maybe if we detect when the
> > guest doesn't have anything useful to do and is burning
> > CPU cycles until the next clock tick, we can suspend
> > running the monitor/guest until the next time event
> > is supposed to occur
>
> Yes. Please do that. It was a huge win for Linux/a386.
>
> (http://linux.a386.nocrew.org/)
Might even be able to use the info that the system is
idling to feedback into the page swapping algorithm.
Perhaps we could put a heavier weight on returning
user pages to the host-OS swap elgibility list while
idling.
-Kevin
