[pulseaudio-discuss] Pulse Audio System Overhead

[keith preston]

I was wondering if anyone had done much of a performance characterization on
the extra overhead imposed by running pulse audio.   Especially in seemly
trivial cases.   I've been evaluating pulse audio and I've found that.

On an ARM based system (no Floating Point)

I am playing a MP3 through gstreamer(mad) to a pulseaudio sink (s16le
44100) to pulseaudio through the native protocol.
In this configuration pulse audio was set up with an alsa output source
(s16le 44100).  This is a direct alsa hookup (no dmix)

In this configuration the gstreamer process takes up ~30% of my CPU and
pulse uses about 20%.

My base case is the same setup without pulse audio.   I am playing through
gstreamer to an alsasink (s16le 44100) directly to the hardware.
In this case the gstreamer process takes ~12% of my cpu.

To me this seems like I am doing something wrong, that is quite a bit of
overheard for essentially what should be transferring and buffering at a
sound server.  I double and triple checked most of my setting.   Are there
any performance detractors that I could be missing?   I started to
investigate volume and noticed that changing the(pulse)  volume didn't
affect CPU utilization.   Is the volume filter always on?   Even in the case
where it should be an identity (100% volume).?  I definitely want to improve
this, I just need a little guidance on where to look.


Keith Preston
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/pulseaudio-discuss/attachments/20070215/77e7fb29/attachment.htm>