[Xenomai] Latency test shows weird results

Gilles Chanteperdrix gilles.chanteperdrix at xenomai.org
Fri Apr 29 21:42:12 CEST 2016


On Thu, Apr 28, 2016 at 05:03:59PM +0200, Florian Richer wrote:
> Hello,
> > This is very much likely the issue fixed by this patch:
> > https://git.xenomai.org/xenomai-3.git/commit/?h=stable-3.0.x&id=b4188f68e9ac7c9e9da25cf57d9756f0ccb78f4d
> >
> > You may want to try Xenomai 3.0.2 to see if you still have the
> > problem.
> Xenomai 3.0.2 does indeed correct the "offset" caused by the overrun.
> > Note that if it is the problem, the real problem is that you have
> > latencies close to 100us, which should not happen on such a
> > processor.
> My latencies were probably caused by CONFIG_COMPACTION, which is now 
> disabled; After a full day of latency testing my maximum latencies are 
> now fine, but the min was at around -1000000us (-1s), any idea where 
> that could come from ? (dmesg and .config attached if it helps)
> > To get useful traces with the I-pipe tracer, you should increase the
> > number of "back trace points".
> Thanks for your help !

Has the minimum drifted slowly over time, or is this an isolated
spot?

-- 
					    Gilles.
https://click-hack.org



More information about the Xenomai mailing list