:: Re: [DNG] 3.4.9 rtai kernel?
Página superior
Eliminar este mensaje
Responder a este mensaje
Autor: Arnt Karlsen
Fecha:  
A: dng
Asunto: Re: [DNG] 3.4.9 rtai kernel?
On Sat, 5 Jan 2019 14:07:21 +0100, Dr. wrote in message
<201901051407.21071.dr.klepp@???>:

> Am Samstag, 5. Januar 2019 schrieb Arnt Karlsen:
> > On Sat, 5 Jan 2019 04:13:22 -0500, Steve wrote in message
> > <20190105041322.13e04671@???>:
> >
> > > Hi all,
> > >
> > > Someone asked whether you can run a 3.4.9 rtai kernel on Devuan.
> > > So I'm asking here.
> >
> > ..why an old a 3.4.9 rtai kernel and not e.g.
> > linux-image-4.18.0-0.bpo.3-rt-amd64, now at version 4.18.20-2~bpo9+1
> > with PREEMPT, RT etc?
> >
> >
>
> PREEMPT-RT gives absolutely horrible latency figures on older
> hardware and amd64 (this is from a T60): RTAI ~ 2000ns, RT ~ 150000ns
> on the same hardware, with latency spikes now and then of up to 5
> seconds (again, depending on individual hardware, even seamingly
> identical systems do not give consistent results). The latest
> linuxcnc-patched PREEMPT still gives ~ 20000ns on the same hardware,
> that's ~ 10 times worse than RTAI. Additionall, when running 2
> realtime threads, RTAI gives a reliable latency over different
> baseperiods. On Linuxcnc that means that latency of the servothread
> at 1000000ns and the basthread of ZZZ ns (default 25000, but usually
> you adjust it to be just 20% longer than the measured latency) are
> independent. With PREEMPT that's not the case: Latency spikes
> erratic, depending on the ratio servoperiode to baseperiode and the
> airpressure in the vally of the kings. when baseperiode 25000 gives a
> latency of 10000 then on RTAI you could savely run a baseperiode of
> 12000. On PREEMPT you most likely will lock your system doing the
> same. Even worse, enlarging the baseperiode to 30000 does not
> guarnatee that the latency stays the same or gets better, just on the
> contrary: it can spike up to 100000, and a baseperiode of 50000 might
> give again 10000 latency. This behavour is most irritating and points
> to a well hidden bug just waiting to suck your blood and time or ruin
> your parts. If you get a RTAI kernel working for your system, you
> will definitly want to stick with it. With PREEMPT you are on the
> "interesting" side of live.


..there is _some_ improvement on the PREEMPT-RT side:
http://linuxgizmos.com/real-time-linux-explained/

--
..med vennlig hilsen = with Kind Regards from Arnt Karlsen
...with a number of polar bear hunters in his ancestry...
Scenarios always come in sets of three:
best case, worst case, and just in case.