[pulseaudio-discuss] question to recording with pa.

Alexey Fisher bug-track at fisher-privat.net
Tue Oct 20 09:50:01 PDT 2009


Am Dienstag, den 20.10.2009, 17:15 +0100 schrieb Colin Guthrie:
> 'Twas brillig, and Alexey Fisher at 20/10/09 16:02 did gyre and gimble:
> > Thank you, the answer was clear. 
> > Now i need to test what is cheaper: to reasample by pa or by aplication.
> 
> Please note that PA supports several different resamplers and the 
> evaluation of "what is cheaper" could easily be invalidated by the user 
> selecting a different resampler.
> 
> I'd advice you to generally let pulse worry about the resampling. Let 
> the appropriate part of the stack deal with it in the way the user likes  :)
> 
> Col

The question i made was a part of my investigation "what making my
netbook on simple tasks so busy?" Currently ubuntu karmic use default
"resample-method = speex-float-1" and it seems to be not optimal for me.
More over, i use gnome-sound-recorder which is not really optimal too.
For example: to record sound to speex, "audio/x-raw-float,rate=44100 ...
speexenc" which make PA resampling stream. After i switched to
"audio/x-raw-int" i reduced CPU usage to 5%. But i don't need
"rate=44100" with speex i need "rate=16000", so we are back on resample.
if i use "resample-method = ffmpeg" it seems to be better, but i don't
wont to change anything in "/etc/pulse/*". So it will be probably better
to use mplayer to record sound with internal resampler.

But currently i care more about cpu usage on record without resampling.

Alexey.




More information about the pulseaudio-discuss mailing list