[mythtv] No Start with Pulse Audio (Changeset 20310)

[Ed W]

Daniel Kristjansson wrote:
> On Sun, 2009-04-12 at 23:53 +0100, Colin Guthrie wrote:
>   
>> If there is anything I can do here, please let me know. I do quite a bit 
>> of hacking on pulseaudio, so any problem you're having, I can try and 
>> sort things out.
>> I certainly believe that a better solution than just exiting when pulse 
>> is enabled can be found!
>>     
> Colin the problem is very large latencies. This is especially a problem
> with Ubuntu, and especially with the as of yet unreleased version called
> jaunty which makes it difficult for regular users to remove pulseaudio
> and which sets pulseaudio up as providing the ALSA default interface.
>
>   
>> Is there a specific problem with the pulseaudio support patch? If so, 
>> perhaps I can help make it better if I can be told what's wrong. Would 
>> detecting pulse and switching to that output system be acceptable?
>>     
> There is nothing wrong with the patch per say, but the existing patch
> doesn't do anything for the latencies when using pulse audio. Not that
> I would expect it to.
>
> I have started working on a solution basically it will disable
> pulseaudio by default if the audio server is local, but also allow
> you to explicitly enable pulseaudio should you really want it.
>   

Forgive joining in 2 weeks later, but the issue seems to be unrelated to
pulseaudio or myth - the issue seems to be just kernel scheduling being
way off the mark on some kernels. Actually it seems surprising that myth
works at all reliably under these conditions either - if busy wait loops
are being serviced at 1/5 second kind of intervals then likely plenty
more things are going to go wrong than the relatively small problem of
feeding the audio server with a small amount of data?

I'm baffled that people get so excited about audio APIs.  They nearly
all work in roughly the same manner and you just feed them data on a
regular basis - I personally can't see a huge bunch of difference
between OSS, Alsa, Jack, RtAudio - they all work roughly the same from
the point of view of handing off the audio.  They all seem to have
various boring implementation details of setting up the interface (Alsa
being my least favourite from that point of view...), and they all seem
to have a ton a bugs due to implementation problems with the low level
drivers (but that's hardly the specific fault of the upper layers -
trying to get a stable permutation of buffer size and periods in Alsa is
one of the more fun games...)

Anyway, sounds like the issue is change the kernel more than change the
software?

Just to put my stake in the ground, I am a Jack user running without
problem at a mere two periods of 256 samples (it runs fine at 2x 64
samples also), which is around 5ms latency (I also burn a ton of CPU at
the same time using resampling and Brutefir doing some massive
convolution filters).  I had hoped to switch to Pulseaudio at some point
and eliminate Jack since the features of Pulse look pretty nice.  Here's
hoping that support isn't dropped, just that the instructions warn
people about getting the kernel sorted...

Good luck

Ed W
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mythtv.org/pipermail/mythtv-dev/attachments/20090423/55d805cf/attachment.htm>