[pulseaudio-discuss] CPU usage of pulse audio-0.9.5-r4 is at 25-30%

[Lennart Poettering]

On Wed, 27.06.07 12:45, Flemming, Eric (GE Healthcare) (Eric.Flemming at ge.com) wrote:

Hi!

> I upgraded from pulseaudio-0.9.2-r1 to pulseaudio-0.9.5-r4 on my gentoo
> Linux system.  I am playing sound from the gentoo linux system to 3 ALSA
> USB devices.  I can hear sound from the 3 sound cards just fine, but I
> noticed that the CPU usage% (as shown by the top utility) of the pulse
> audio daemon is at around 25-30%.  T
>  
> his much higher than in 0.9.2-r1 which was at around ~10%. Is this high
> of a CPU usage in 0.9.5 expected behavior?   
> Has anyone else seen this?   
> Does anyone know why I might be seeing this higher CPU usage?  Might
> this be a defect?
>  
> Thank you in advance for your help.  I have been trying to fix this for
> awhile and any help you can provide would be greatly appreciated.

This is probably related to the fragment settings that have been
chosen by PA. The smaller the sound card fragment sizes are the higher
will be your CPU load. Its a tradeoff between latency and CPU
load. You can pass different fragment settings to the ALSA or OSS
modules of pa and thus influence the CPU load.

I hope I can convince the ALSA people to allow dynamic fragment size
(or "period size" like they like to call it) changes during runtime
without audio interruption. That would allow us to run with large
fragments during normal runtime and shorten them if a client connects that
requires low latencies. But until that is available in ALSA I fear you
have to configure your fragment sizes manually.

See http://pulseaudio.org/wiki/Modules#module-alsa-sink for more
information.

Also, if you run pa with "-vv" it will log the fragment sizes that are
actually used.

Lennart

-- 
Lennart Poettering                        Red Hat, Inc.
lennart [at] poettering [dot] net         ICQ# 11060553
http://0pointer.net/lennart/           GnuPG 0x1A015CC4