sdv-univ wrote:
> Allright, perhaps you are correct that VESA would be the simpliest to
> implement after VGA. However, higher resolutions, and especially large
> color depthes, couldn't work fast in GUI without the hardware-assisting 2D
> acceleration. Imagine software-only scrolling 1024x768x32bit across
> the PCI/AGP bus - even these probably don't have enough bandwitch, and
> even if they had why waste it on operation that should be performed by
> accelerator locally?
> Also, 2D acceleration implementation shouldn't be THAT complex (in
> contrary to 3D). I didn't propose windowed accelerated 3D, only a
> full-screen one. The later also wouldn't be very simple I'd imagine,
> because it's card-dependent, so the plex86 should be provided with the
> hardware resources this particular card uses, to allow direct access
> only to these resources, also to save/restore the state of hardware.
> Yet simplier than building a whole wrapper.
> Another thing that bothers, is that competition is very hard among
> card manufacturers, nobody would know if the particular architecture would be
> popular in few years. This favours to abstraction of acceleration,
> however the need to write custom drivers for guest OS'es makes this
> good idea look less attractive..
> The idea of multiple video plug-ins for plex86 is nice.
I have to agree. I'm thinking that the best option is to use
an abstraction layer to talk to SDL, and special guest OS drivers
to talk to that layer. We could pump 2D/3D primitives to the
real card via SDL.
Though, I'm not familiar with how SDL works yet. If someone wants
to write a summary, please do. Anyways, it would be nice if we
could potentially map the video framebuffer into the guest
address space, and let the special guest OS driver write direct
to the card. I would imagine the key is letting an abstraction
layer take care of video card configuration etc.
-Kevin
