[MPlayer-dev-eng] Patch: ao_jack for smoother playback

Ed W lists at wildgooses.com
Sat Jul 9 21:04:54 CEST 2005


Hi Reimar,

I did send this message to you a couple of times offlist, but either you 
didn't get it, or didn't think it was even worth commenting on...  Just 
in case it was the former I am resending it to the list in the hope that 
it might catch your eye.

Basically if the Jack buffer size is very high then playback in mplayer 
is extremely jerky.  I have tweaked the sync code so that it's in a 
"pretty accurate" state by default, and there is a define to put it into 
super accurate mode as well.  Personally I would prefer to go with the 
super accurate mode by default, but both look quite reasonable on my system

Thanks for listening

Ed Wildgoose

-------------------------------

Hi Reimar,

Just been re-testing the CVS Jack stuff and I have another sync patch 
for you.

Justification is as follows.  I startup Jack with param "-p 4096" to 
give a sensible load to the system, ie period size of 4096 samples

What happens is that mplayer then only calculates video delays in around 
100ms increments and consequently the video is extremely jerky

So I enable the "ESTIMATE_DELAY" option in the code and now it's all 
extremely smooth.  So I think that we really need this option enabled by 
default.

I submit the following patch which a) enables the option by default, and 
b) then includes another option which tries to get things even smoother 
by simply incrementing the predicted pointer unless we deviate too far 
from reality. 

I think we need the normal delay estimation, but I'm hoping that since 
it's a #DEFINE you will also commit the rest of the patch which gives 
this high quality sync option...

Thanks for writing this code - it seems to be working pretty well (OK, 
apart from the channels reversing when you fast forward...)

Thanks

Ed W




-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: ao_jack.c.2.diff
URL: <http://lists.mplayerhq.hu/pipermail/mplayer-dev-eng/attachments/20050709/de34beb8/attachment.asc>


More information about the MPlayer-dev-eng mailing list