[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: pulseaudio CPU usage and process name



On Fri, 16.11.07 20:52, Ahmed Kamal (email ahmedkamal googlemail com) wrote:

Hi!

> Hi, I'm enjoying the wonders of pulse audio. However, on my centrino
> 2.0Ghzlaptop, when idle, almost always pulse audio is the most CPU
> intensive task running, even when I am not playing any audio! It
> consumes around 6% of CPU competing with firefox! This seems a bit
> too much to me ?  Also, any idea why the process name appears in top
> as "exe" rather than pulseaudio!  Regards

The process name issue has been fixed already, but I haven't uploaded
a new version of PA into Rawhide yet. Also note that this issue is
alrady listed in rhbz as #345761.

PA closes all devices when idle for more than 1s. Then, CPU
consumption goes to 0%.

Now, there is a certain piece of software which happens to be
horrendously broken and is called "Macromedia Flash". I would guess
that you are using it, aren't you? Now, Flash never closes any
playback streams until you terminate your browser. Due to that PA
never considers the device idle and thus never closes the audio device
and continues to write audio data to it.

Due the closed source nature of Flash this issue is only fixable with
exceptionally ugly hacks in PA or libflashsupport.so, which I am not
willing to do.

If PA consumes 6% CPU this is probably because resampling is needed to
output audio on your device. Resampling is usually the only CPU
intensive code that is run in PA. While traditionally resampling is
done in the various media players if you use PA it is done in
PA. Hence the CPU load will appear as if it was PA's fault, but we
just moved the resampling work from the app to PA, that's all.

Please terminate your browser, and pause/stop all media players and check
the CPU load then.

(Some codepaths in PA are not optimized as much as they could of
course, but from the little information you posted I guessed what I
wrote above. If you feel that my guess is not correct, then please
open a bug in rhbz)

Thanks,

Lennart

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


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]