[mythtv] [PATCH] vsync reimplementation

Doug Larrick doug at ties.org
Wed Jul 14 10:01:49 EDT 2004


[Sent this to Kenneth only; let's try that again.  Kenneth, please don't
CC me directly.]

Kenneth Aafløy wrote:

> On Tuesday 13 July 2004 13:42, Doug Larrick wrote:
> 
>> Can you make available a minute or so of a clip that starts up 
>> poorly so I can test it?
> 
> It appears that I introduced a bug when I was working with the
> teletext patch, which generated invalid pts data (in comparison with
> the original ts), so it should not occur in normal operation. But if
> you still want a sample with very bad pts info, I could put in the
> bug again, and make a sample.
> 
> This new patch will aquire sync faster even on normal streams, won't 
> it? I'm sorry if you did unecessary work on behalf of my bugreport.

No apology needed; it was a good idea anyway... and actually makes the
code cleaner IMHO.

Doesn't sound like I need a sample.  Thanks.

The new patch will only acquire sync earlier if it's pretty far out to
begin with... 20 frames.  It also avoids a bug in synching with audio
timebase if the recording has 0 as the first audio timestamp (i.e. no
audio for first frame).  This bugfix was already in video timebase.

>> Which avsync method are you using (video or audio  timebase), and 
>> which timer method?
> 
> On my desktop (w/GeForce TI4200 & 6106) I tested with OpenGL, RTC, and
> the busy wait method, on my epia m10k I tested RTC and busy wait. Both
> is using audio timebase, and all methods performed more or less equally.


Good to know.  I think the more-advanced methods (than busy wait) will
perform better on more heavily-loaded systems, both because they consume
less CPU and because the wait is more precise.  And of course the
methods that sync to retrace will work better on videos whose framerate
doesn't match the display, especially with video timebase.  Of course, 
video timebase won't work with AC3 passthrough.

-Doug
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 256 bytes
Desc: OpenPGP digital signature
Url : http://mythtv.org/pipermail/mythtv-dev/attachments/20040714/c0e092fc/signature.pgp


More information about the mythtv-dev mailing list