[MPlayer-dev-eng] [PATCH] reduce latency for ao_pulse

[Reimar Döffinger]

On Wed, Jul 14, 2010 at 08:06:45AM +0200, Dan Oscarsson wrote:
> tis 2010-07-13 klockan 19:30 +0200 skrev Reimar Döffinger:
> > On Tue, Jul 13, 2010 at 08:44:51AM +0200, Dan Oscarsson wrote:
> > > It cannot be fixed by autosync as the
> > > problem is that pulseaudio reports a "get_delay" that do not match what
> > > it should be so mplayer calculates wrong playing audio pts.
> > 
> > Uh, broken get_delay is exactly what autosync is there for.
> 
> If you enable autosync mplayer drops frames to keep up with the faulty
> ao driver.

No, -framedrop does that. If autosync without -framedrop does that it would
be a bug.

> It is the audio that is the problem, not the video.

I repeat: autosync was explicitly designed for broken audio drivers, which
is exactly this case. It may not actually implement to detect and work
for this case, and I am not sure it is possible to do this reasonably
successfully.

> My patch
> changes so pulseaudio works much better without needing autosync.

Well, it also does nothing to avoid similar issues that I suspect other
aos like SDL and possibly OSS have, seems to rely on pulseaudio internals
that might change with unknown effects on our code and will continue to
cause higher CPU load and higher power usage due to more frequent wakeups
even if the bug is eventually fixed.
Do you have a link to a pulseaudio bug report so we have at least a chance
to notice if it is ever fixed?