[LAD] JACK and CPU stress testing

Dominique Michel dominique.michel at vtxnet.ch
Sat Mar 2 21:37:53 UTC 2013

Le Sat, 2 Mar 2013 15:47:00 +0000,
Harry van Haaren <harryhaaren at gmail.com> a écrit :

> Dominique: I've not tried your suggestion yet. Perhaps later if I
> have time.

A little update. I succeeded to crash amule on my 4 core system. Same
behaviour: amule is using 100% processor on 1 core, eat all the ram
(8 GB), swap like hell and crash. During swapping, the system was so
slow that it was unusable (latency for a mouse click around ~30 seconds
or so). Also, the only thing that crashed with amule was mplayer and
jackdbus. This was not with a rt kernel, but with the gentto sources
with cgroups configured for rt (CONFIG_RT_GROUP_SCHED) and no other
cgroups policy.

So, jack2 crash easily on my 4 core system with a cgroup enabled kernel
than jack1 on my old 1 core system with a rt kernel. That's too many
variables... so more test is needed in order to know for sure if the
cause of this jack2 crash is the huge swapping or something else.

Someone know the impact of swapping on jack?

I think than to compile hugin is a better test, it will load the system
without swapping but with a very high processor use on all the cores.

"We have the heroes we deserve."

More information about the Linux-audio-dev mailing list