Re: [a-devel] AGNULA/DeMuDi 1.1.1 and 2.6.x kernel
On Thu, 2004-06-03 at 13:03, Free Ekanayaka wrote:
> >>>>> On 02 Jun 2004 21:03:50 -0700, Fernando Pablo Lopez-Lezcano <nando@ccrma.Stanford.EDU> said:
>
> Fernando> On Tue, 2004-05-25 at 17:41, Fernando Pablo
> Fernando> Lopez-Lezcano wrote:
> >> > seems that some people have done some deeper tests ( Fernando
> >> Pablo > Lopez-Lezcano for ex.) which havr shown than the 2.6.x
> >> kernel is not > good enugh yet for audio production. Maybe it
> >> worth to take this into > consideration.
> >>
> >> And then maybe not :-)
> >>
> >> Another data point, I just (finally) booted into FC2 (that
> >> comes native with 2.6.x/ALSA) using 2.4.26-1.ll+ALSA (as built
> >> on FC1). I still get xruns. Much less than with 2.6.x but they
> >> are there nevertheless. So there is something other than 2.6.x
> >> messing things up (I suspect the xorg X server, but have no
> >> hard data to back the claim).
>
> Fernando> So now I know (sort of) what was going on. The culprit
> Fernando> was not the kernel. The culprit was not xorg.
>
> Fernando> The culprit was a change in behavior of pthread_create
> Fernando> (what it is related to I don't know yet, I assume this
> Fernando> change is part of glibc - I just found out a few minutes
> Fernando> ago). That change means that the current jack code,
> Fernando> together with glibc or whatever it is that changed in
> Fernando> FC2, creates threads for the jack clients that are _not_
> Fernando> SCHED_FIFO (but jack itself is still SCHED_FIFO). You
> Fernando> can imagine that that can cause xruns :-)
>
> Fernando> I just did a test with a hacked jack and things are back
> Fernando> to normal (meaning that it should be now possible to
> Fernando> really test 2.6.6 in my environment for low latency
> Fernando> behavior).
>
> Fernando, thanks for this deep analysis of the issue.
>
> I'd like to know whether this bug holds even for Debian
> glibc/2.6.x.
Apparently so, Jack O'Quin mentioned that the hack seemed to make some
problems he had been seeing in jack under Debian/2.6 dissapear. BTW, I
should have been more specific, this happens only with NPTL threading
(not the normal pthread library). More testing is needed, I could reach
a state where jack would pretty much hang the machine... killing jack
returned things to normal (the machine was not hung, just very very very
slow).
> If this is the case I think that A/DeMuDi should really
> stick to 2.4.x until the jack code is fixed upstream.
-- Fernando
Reply to: