[mythtv] Proof of concept: Chromakey OSD

Blammo blammo.doh at gmail.com
Thu Mar 24 22:38:15 UTC 2005


> It's possible that you're encountering the fact that while this code is safer
> (I believe), it is necessarily a bit slower than the code in CVS... primarily
> because it holds the mutex longer and more often.
> 
> You mentioned "freezing" issues.  Were these when using XvMC?  If so, has the
> patch improved that situation any?

Not sure.. the resultant video would have dramaticly lowered my WaF
(wife acceptance factor) so I didn't leave it in place. It really was
unwatchable through the whole program.

I'll try to describe better what it looked like.

If you split the screen up into two screens, each one taking every
other line. (a trick I know some 3d glasses use) Like 1,3,5,7,etc was
screen1, and 2,4,6,8,etc was screen2. Then you started the same
program on each one, only you started the 2nd one about .5 - 1 second
behind the first one.

As obtuse as that probably sounds, that's what it looked like.
tearing, stuttering, but sometimes right one if there was little to no
motion for 2-3 seconds.

Yes, I'm using XvMC. Have to for HD content, even with a beefy CPU, if
I want any deinterlacing.

The "freeze" issues, which happen several times a week, seem to be
related to errors in the bitstream, IE I see DVB errors on the backend
log for the same time, and most of the time, the audio also has
dropped out. I'm aware these are known issues and under repair by
Taylor/others.

The OSD / Flicker issues irratating, but not stability affecting,
unlike early on in 0.17 where they would crash almost 100% of the
time. Hats off to the devs for progress.


More information about the mythtv-dev mailing list