[ardour-dev] Ancient history?

[Steve Harris]

On Tue, Nov 02, 2004 at 08:46:37AM +0200, Tommi Sakari Uimonen wrote:
> >>2.6.9.  It'd be nice to see some experiments on that...  I still run
> >>jackd -R -n 3 -p 256 and have ardour 0.9beta19 xrun sometimes even w/
> >>only a few tracks.  I didn't expect that on my dual athlon 2400MP.
> >>
> >>1.  I don't kill every process known to man before running ardour, and
> >>I use windowmaker instead of twm or something, but I bought a dual
> >>proc box so I wouldn't have to do that...
> >
> >You shouldn't have to if your soft-realtimeness is working ok. I run
> >gnome, and dont kill any background processes. Sometimes I even run with a
> >(busy) mysql in the background.
> 
> I found a very confusing behaviour some days ago. I was trying the limits 
> of my jackd system and got as far as -R -n 2 -p 16. Just jackd running, no 
> ardour nor any other jackd client.
> 
> Then I just moved mouse around my kde (focus follows mouse was on) and 
> witnessed xruns during the focus changes.
> 
> The confusing part was that after I reniced most of my running processes 
> to lowest (19) priority, I got fewer xruns when moving mouse across kde.
> 
> I have lowlatency and pre-emptive 2.4.20, M-Audio Audiophile 2496 at irq 9 
> (nothing at irq 8 & 10), ATI card at irq 11, Athlon 1700+
> 
> 
> I thought that renicing shouldn't affect since jackd is running -R

Yeah, but when youre running with that short a period size (400 us,
roughly) even tiny things can have enough effect to cause xruns, eg. a bit
extra cache damage could be enough to make you miss the dealline, so
renicing them may make them run less often (as a side effect), and hence
cause cache damage less often.

I dont know what the current scheduler does with low priority processes,
so I dont know if thats likely.

- Steve