State of cleanup-audio-jumble

[Toby Smithe]

On Mon, 2007-07-23 at 15:15 +0200, Alexandre Franke wrote:
> Hi,
> 
> There is a spec [1] that was announced to be implemented for Feisty.
> The goal of this spec is to "fix the Linux audio mess once and for
> all". The spec mainly consists in getting rid of esd and use
> pulseaudio instead. As far as I can see, esd is still used in Feisty
> and I wonder if someone still cares about it. Moreover pulseaudio
> seems to be a proposed feature for Fedora 8 [2]. Is there any plan to
> have this in Gutsy? In Gutsy+1?

Yes. I agree that this is a good plan, and I have implemented PulseAudio
support into my default sound card chooser (asoundconf-gtk, which will
become part of asoundconf-ui[0]). However, at this stage, PulseAudio
causes the sound card to be active most of the time, creating interrupts
for the CPU to handle. This means that it can never enter low power
states, which is very important to save battery life. As a result, I
don't think that PulseAudio is ready for mainstream adoption (unless it
was to be excluded on laptops, but this isn't really going to be
feasible).

I believe that the interrupts (polling?) when there is no actual sound
being emitted is a bug in ALSA, and that it is being worked on (I
chatted in #pulseaudio/freenode about it briefly), but I'm not sure of
the specifics and if there is a bug report open (I had a brief look on
ALSA's bug tracker, but didn't find anything). I think that currently
the only way to reduce the interrupt generation is to increase the
fragment size for all ALSA applications (as this cannot be done
dynamically right now), which means that there is a trade-off in
performance.

Please correct me if any of this is incorrect, as a lot of it is just
what I can remember from my conversation, and is probably inaccurate as
a result :)

Toby Smithe

[0] https://launchpad.net/asoundconf-ui