Messages in this thread | | | Date | Mon, 20 Aug 2007 21:12:08 +0200 | From | Ingo Molnar <> | Subject | Re: [accounting regression since rc1] scheduler updates |
| |
* Martin Schwidefsky <schwidefsky@de.ibm.com> wrote:
> > If a virtual CPU is idle then i think the "real = steal, virtual = > > 0" way of thinking about idle looks a bit unnatural to me - wouldnt > > it be better to think in terms of "steal = 0, virtual = real" ? > > Basically a virtual CPU can idle at "perfect speed", without the > > host "stealing" any cycles from it. And with that way of thinking, > > if s390 passed in the real-idle-time value to the new callbacks > > below it would all fall into place. Hm? > > How you think about an idle cpu depends on your viewpoint. The source > for the virtual cpu time on s390 is the cpu timer. This timer is > stopped when a virtual cpu looses the physical cpu, so it seems > natural to me to think that real=steal, virtual=0 because the cpu > timer is stopped while the cpu is idle. The other way of thinking > about it is as valid though.
my thinking is this: the structure of "idle time" only matters if it can be observed from "within" a virtual machine - via timers. Are on s390 any of the typical app-visible timers (timer_list, etc.) driven by the virtual tick? [which slows down if a virtual CPU is scheduled away by the host/monitor/hypervisor?] Or is the virtual tick only affecting scheduling/cpu-accounting statistics in essence?
Ingo - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |