> > > Some embedded devices are so slow that --enable-small-dither-
> > >mask is not enough, we have to completely remove dithering to
> > > get some speedups. This is the case with Nokia 770 or N800,
> > > where we have a [EMAIL PROTECTED] but not much CPU to handle all
> > > these pixels :-)
> >
> > Some of that conversion and dithering related code could
> > be sped up somewhat.. though it'd be a fair amount of work (just
> > to go thru it all is quite a bit) and might not be worth it in
> > all cases.
>
> Yep. Actually, I'm talking to raster about we start a 16bpp-only
> software_x11 engine. Problem is: today we need to convert
> 24->16[+dither] + alpha for each blit, I've removed the [+dither]
> step, improving performance in about 20%... but from his tests,
> having 16bpp packed in a special format, with alpha in another
> plane and scaled down to 5 bits, we can multiply both r5g6b5 in
> 1 32bit instruction, giving 2x speedup... that's a huge thing!
That would be better yes. I don't think you need a separate
x11 engine for this (and indeed this would be useful for the other
software based engines), just add compositing functions to blend
a8r8g8b8 to r5g6b5 16bpp dst images would be a great improvement
there (and adding suitable changes to get the correct version).
Sounds like you and raster have something going already
here.. excellent!
An interesting side question though would be to wonder how
many display devices out there are planning to have 16bpp displays
for very long...?
-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
enlightenment-devel mailing list
enlightenment-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/enlightenment-devel
