[MPlayer-users] "shit" S3 drivers? (fwd)

[Arpi]

Hi,

> >And what about YV12 support? The driver has very slow conversion code, it
> >should be optimized. And what about using DMA for vidoe transfer? It
> >eats too many CPU time now.
> 
> Optimizing YV12 transfers would come in about #68 on my to-do list.  I cannot
> imagine a multimedia application that produces ONLY YV12.
Ok, but, then, why do you support YV12 at all? If a program sees the video
driver supports YV12 too, it uses YV12 because it needs less bandwith and no
conversion at the program's side. If a program sees YUY2 only, it will do
the conversion (with MMX).

> The streams engines for all the early chips are fairly similar.  Maybe you
> would want to take a shot at porting the ViRGE xvideo driver to the Trio64?
no, thanks :) i have no such card. i've matrox and ati and developing
drivers for those.

> That's the way most of these drivers get developed: someone who cares digs in
> and makes it happen.
I know. We've just talked about quality of S3 drivers.

> >And what about the very poor performance of AGP Trio3D/2x drivers? I
> >required 5 times more CPU time to transfer a 320x200x32bpp image as my
> >oldy PCI ET6000. (it was with 4.0.2, maybe already fixed, but I already
> >replaced that card by ATI mach64)
> 
> No clue.  The Trio3D/2X, as you probably know, is really a ViRGE that was
> cursed by marketing with a bad name.  I thought the ViRGE driver was in
> pretty good shape.
AFAIK 'real' Virge cards works much better.


A'rpi / Astral & ESP-team

--
mailto:arpi at thot.banki.hu
http://esp-team.scene.hu