[Mplayer-cvslog] CVS: main/libmpcodecs vd_ffmpeg.c,1.26,1.27

[Arpi]

Hi,

> > Ah, ok. Btw, whiel you're there :)
> > I noticed that lavc now expect persistent I/P buffers (i think your old
> > skip-mb optimization). I mean, every time lavc request a buffer, it gets
> > one, different from previous, but maybe not the same, as before previous.
> > To reproduce, increment number of buffers 2->3 ni vo_xv.c, and change type
> > _IPB to _IP in vf_ffmpeg.c (i'm working on IPB right now).
> hmm, i allready expected sth like that ... perhaps the mb_skip table in lavc 
> should be increased for each skiped MB and if >= buffer num than skip ...
hmm, good idea. i thought just disabling that for -dr :)

> > anyway with mga_vid or vidix (dr buffers in vram), and IP only (no B
> > frames) video this -dr has no effect.
> i know
i know you know :)
it was there for atmos, nexus and others testing -dr and asking why is it slow

> > ah, one more question for today :)
> > what's that CODEC_FLAG_EMU_EDGE ? i think it's a hack for UMC, but not sure
> > :)
> yes, u can turn it on even without dr1, it was even slightly faster then 
> normal UMC when i tried it

faster? how? let me guess: it doesn't store umc at all, but the MC code
calculates edges on the fly. btw i should RTFS instead of guessing :)


A'rpi / Astral & ESP-team

--
Developer of MPlayer, the Movie Player for Linux - http://www.MPlayerHQ.hu