[pulseaudio-discuss] Distribution kernels and glitch-free (Packagers, read this!)

[Sean McNamara]

Hi,

On Thu, Mar 12, 2009 at 8:37 PM, Mark Greenwood <fatgerman at ntlworld.com> wrote:
> On Thursday 12 March 2009 14:38:18 Sean McNamara wrote:
>> Hi,
>>
>> On Tue, Feb 24, 2009 at 3:36 PM, Lennart Poettering
>> <lennart at poettering.net> wrote:
>> > On Tue, 24.02.09 12:20, Jan Claeys (lists at janc.be) wrote:
>> >
>> >>
>> >> Op maandag 23-02-2009 om 15:18 uur [tijdzone -0500], schreef Sean
>> >> McNamara:
>> >> > * Once upon a time (old releases, e.g. 7.04 and below?) there was the
>> >> > -lowlatency kernel flavor. This one went all the way and gave us
>> >> > CONFIG_HZ=1000 and a fully preemptible kernel (not only voluntary, but
>> >> > forced preemption). I'm not sure of the rationale for discontinuing
>> >> > this kernel flavor, but it would be silly to say it's only because of
>> >> > mirror disk space or something; their disk space consumption has gone
>> >> > way up since then.
>> >>
>> >> Maybe have a look at the -rt kernel?  :)
>> >>
>> >> (There is a fourth kernel too: -virtual, which is optimized for running
>> >> in a virtual machine, where the number of hardware to support is much
>> >> lower than in real machines.)
>> >
>> > I cannot say I believe in the -rt kernel. We don't need *that*
>> > reliable latencies for PA. And given that -rt in real live breaks more
>> > things than it fixes I don't think it is really worth the effort if
>> > all you want to do is run PA.
>> >
>> > The vanilla kernel is mostly fine for doing multimedia work, even for
>> > audio production. There's no need to bother with -rt.
>>
>> Indeed! And with the Ubuntu -generic kernel, 0.9.15-test* is terribly
>> glitchy with glitch-free.  By rebuilding the same kernel sources using
>> CONFIG_HZ=1000 and full preemption, rather than CONFIG_HZ=250 and
>> voluntary preemption, I get *zero* dropouts under a normal desktop
>> load with the same userspace.
>>
>> And the Intel GEM/UXA stuff is fairly CPU intensive at times,
>> especially scrolling Firefox. That used to glitch like crazy. A simple
>> tuning of kernel parameters erased this problem.
>>
>> Now I can run PA 0.9.15-test without touching the default
>> configuration at all, and it's very, very good in terms of (lack of)
>> glitchiness.
>
> I've noticed that the latest ubuntu update (on jaunty) has disabled glitch-free mode (tsched=0 passed to module-hal-detect). Are you sure that isn't the cause of the lack of glitches? It certainly is for me, wrong though it sounds.

I am sure that this is not the cause. tsched=0 is not passed to
module-hal-detect for me, and there's nothing relevant in GConf
either.

Besides, I recompiled PulseAudio from vanilla sources. Anything that
might've been in the Ubuntu configs has been overwritten by the "make
install" command.

Sean


>
> Mark
>
>>
>> I wish I could come up with some quantitative data that demonstrates
>> clearly how the same kernel sources and userspace stack have wildly
>> different user experience results with PA. It would be nice to make a
>> case to the Ubuntu kernel folks to re-introduce the "lowlatency"
>> kernel flavor.
>>
>> BTW, Lennart, last time I ran Fedora I didn't see a "lowlatency"
>> flavor (or similar) either. Is your default kernel fully preemptible,
>> or do you have an alternate kernel flavor that is? If not -- it may be
>> worth considering. At least on my system, full preemption (and
>> possibly to a lesser degree, CONFIG_HZ=1000) does just the right
>> things to get latency down to make glitch-free happy.
>>
>> Thanks,
>>
>> Sean
>>
>> >
>> > Let's keep things in perspective. PA is not a super-ultra-low-latency
>> > sound server. It's a desktop sound server ... that is all. All I am
>> > asking for is for latencies not as bad as 210 ms!
>> >
>> > Lennart
>> >
>> > --
>> > Lennart Poettering                        Red Hat, Inc.
>> > lennart [at] poettering [dot] net         ICQ# 11060553
>> > http://0pointer.net/lennart/           GnuPG 0x1A015CC4
>> > _______________________________________________
>> > pulseaudio-discuss mailing list
>> > pulseaudio-discuss at mail.0pointer.de
>> > https://tango.0pointer.de/mailman/listinfo/pulseaudio-discuss
>> >
>> _______________________________________________
>> pulseaudio-discuss mailing list
>> pulseaudio-discuss at mail.0pointer.de
>> https://tango.0pointer.de/mailman/listinfo/pulseaudio-discuss
>>
> _______________________________________________
> pulseaudio-discuss mailing list
> pulseaudio-discuss at mail.0pointer.de
> https://tango.0pointer.de/mailman/listinfo/pulseaudio-discuss
>