[Mplayer-cvslog] CVS: main/libmpcodecs vd_ffmpeg.c,1.32,1.33

[Arpi]

Hi,

> > anyway, using draw_slice() to blit the macroblocks is imho a big overhead,
> > and probably means more slowdown than speedup.
> hmm,i would guess that the skiped and non skiped MBs are grouped togther so we 
> dont have much more calls to draw_slice, as multiple non skiped mb are merged 

hmm. we should benchmark it

> > imho, you should add skip-table support to the csp converters, or at least
> sounds like alot of work, dont expect me to do that anytime soon

yes, it is if you do it for everything.
anyway it can be done optionally only for the most used converterts, for
example yv12->rgb24/32. we can pass this array to the converter, and it
either uses it or ignore it.

> > a special version of agpmemcpy, doing linear copy with skips.
> hmm, ill perhaps implement that ...
would be great, it could speed up the most paths when no conversion is
needed (mostly yv12->yv12).


A'rpi / Astral & ESP-team

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