[MPlayer-dev-eng] [PATCH] A new ASS renderer

wm4 nfxjfg at googlemail.com
Sun Sep 23 18:48:44 CEST 2012


On Sun, 23 Sep 2012 17:44:45 +0200
Reimar Döffinger <Reimar.Doeffinger at gmx.de> wrote:

> So I am also kind of curious if there was a specific reason why you
> implemented your variant for -vo gl? (you mentioned OS X I think)
> I just don't particularly like that it has some known pathological
> cases where it will behave vastly worse than it could (and also
> vastly worse than software rendering).

Reasons why I did that included:
1. It was using display lists, a deprecated OpenGL feature, which
triggered my OCD.
2. It potentially wasted video memory.
3. It didn't work well on OSX.

I also tested on Linux + nvidia binary blob. Loading a very complex ASS
script (~70 MB) and playing 5 seconds takes 46 seconds with my hacked
-vo gl, and 70 seconds with vanilla vo_gl. That test was very flawed
though; these two builds weren't even from the same git revision /
compiler / ffmpeg, and I don't remember the difference being that high.

(Btw. MPlayer 1.1-4.7 died trying by allocating > 1 GB of memory when
loading the script. Not sure what's happening here.)


More information about the MPlayer-dev-eng mailing list