[MPlayer-dev-eng] Colored subtitles

[Jindrich Makovicka]

Alban Bedel wrote:

>Imho color space conversion should be done only once. Converting all the time
>is a waste of cpu time (even if you have too much idle time there's some better way
>to use it ;)
>
It's done once per subtitle. I don't think that pre-rendering all 
characters in all thinkable colors is a good idea. It also means you 
need to pass the information about used video format to yet another 
layer of code and you need another bunch of functions for various 
colorspaces which will render into the osd buffer.

>>Stuff that (probably) works: xv (YV12) x11 (tested with 32bit), sdl (32bit)
>>Stuff that surely doesn't work: spudec, aalib, anything yuy2 & uyvy
>>
>>Interface changes: vo_draw_text and vo_remove_text now have one more 
>>argument for desired image format, vo_draw_alpha_yv12 has additional 
>>args for more layers.
>>    
>>
>I think that it must be done in some other way. First I still want the old osd that
>only use the Y plane as it's faster. And i don't think that it's a good idea to modify
>the existing vo_draw_alpha_* funcs as most of them (all ?) are optimized for x86.
>
My code is still not optimized and I don't see any special performance drop. There is more source data but the amount of written data is the same.

>Imho we should keep the existing stuff in the vo as is. Write some _new_ drawing
>funcs that support colors, etc. Use this in the expand filter as an alternative osd
>mode. Then it could be added to the vo wich really need a special osd support
>via control.
>
This means duplicating of osd.c and sub.c (and probably spudec). One 
version for color, one for mono.

--
Jindrich Makovicka