[LAU] Kxstudio RT kernel vs low latency

[James Stone]

On Sun, Jul 27, 2014 at 11:04 AM, Joakim Hernberg <jbh at alchemy.lu> wrote:
> On Sat, 26 Jul 2014 10:33:07 +0200
> Jeremy Jongepier <jeremy at autostatic.com> wrote:
>
>> Just a big thank you for the comprehensive explanation! I'll add this
>> to the System Configuration wiki page on linuxaudio.org if you don't
>> mind.
>
> Not at all, go ahead.  There are a few things I wish I would have
> written differently, but it was a ml post and not an essay :)
>
> The rt testing package is indeed called rt-tests and not rt-tools.
>
> The reason I see max scheduling latencies of about 100us on the -rt
> kernel with cyclictest is most likely due to using the -i100 parameter.
> Setting it lower would likely result in lower max values, but at the
> cost of increasing cpu use.
>
> In fact I think the max scheduling latencies on my system is more
> likely around 40us or so, but what really interests me is to know
> that I don't have max values into the millisecond range.
>

Hi Joakim,

I've just been playing around with this, and can confirm that a
hand-built -rt kernel has lower max sched latencies than a generic
lowlatency kernel in ubuntu on my system (<100us compared to 1500+).
However, I noticed a really weird thing - that when running the test
using sudo cyclictest -m -n -Sp99 -i100 -d0,  the reported DSP load on
qjackctl is reduced (by around 50%), and there are fewer xruns at low
latencies. As soon as I stop cyclictest, the xruns (at a rather low
jack frames/period of 32) come back.. Any idea why this should be???

James