[MPlayer-users] mplayer 1.0rc3 - CPU load issues

[Reimar Döffinger]

On Tue, Apr 05, 2011 at 05:20:41PM +0100, John Stirling wrote:
> We've compiled for arm and have an audio device fixed at 48829hz.

Honestly? Make the audio device claim it runs at 48 kHz.
Resampling from 48 kHz to 48829 Hz won't really give great quality
and probably to most the 2% frequency offset will sound better
(for people used to PAL that's almost certain).

> We
> are having some issues with audio quality when we set this option
> -af-adv force=2

You are rather unspecific about what kind of issue exactly.

> . Does the increase in CPU load when switching from force=2 to
> force=1 sound about as expected

Without knowing what kind of device that is it's not really possible
to answer. Well, we probably can't answer either way.
I assume you disabled any frequency scaling if available?
Otherwise your numbers may be completely off.
Either way, since there is no assembler optimization for the
resampling code it can be rather expensive.

> . When the audio device natively supports the sample rate of the
> source content will -af-adv force=2 have a (or much) negative effect
> on audio quality. It seems fairly good to me..

That option doesn't really do anything significant besides selecting
resampler options.

> We'd ideally like to keep CPU load close to what we're seeing with
> [-af-adv force=2] but without the audio quality issues.

Use -af lavcresample and play with the options to see if you find something
that works for you.
-af lavcresample=<freq>:1
might be a good starting point.