Re: OpenGL-based framebuffer concepts

[D. Hazelton]

On Friday 02 June 2006 04:28, Jon Smirl wrote:
> On 6/1/06, Jon Smirl <jonsmirl@xxxxxxxxx> wrote:
> > On 6/1/06, Dave Airlie <airlied@xxxxxxxxx> wrote:
> > > > > 15) re-use as much of the X drivers as possible, otherwise it will
> > > > > KGI.
> > > >
> > > > I would broaden this to use the best code where ever it is found. Of
> > > > course X is a major source.
> > >
> > > I'm not considering using knowledge from X drivers, I'm considering
> > > using the X drivers, I don't personally care about things like X's
> > > over use of typedefs and that sort of stuff, that is what I term
> > > semantic, people who work on X drivers know X drivers, and writing the
> > > drivers is the biggest part of any graphic systems.
> >
> > I have considered that option too. It is a good place for a quick
> > start but it is not maintainable in the long run. The driver code has
> > to be divorced from X and not require having the entire X system
> > around to build a new driver.
> >
> > Have you checked the dependencies needed for loading X drivers?
> > Modularization may have helped but loading an X driver used to
> > effectively suck in the entire X server due to dependencies. Sucking
> > in all of X is not fair to alternative windowing systems.
> >
> > I do agree that this is a workable starting point but it can't be the
> > long term solution.
>
> I just checked the Xorg R7 drivers. The ones I checked are statically
> linked to their X components so there are no big X dependencies. That
> makes them usable as standalone drivers.
>
> What do you think about wrapping them with EGL instead of using their
> entry points directly? That would remove the temptation to use
> acceleration code in the X drivers and encourage use of DRM instead.
>
> Wrapping them with EGL was my plan for getting Xegl up last summer
> when Nvidia wouldn't implement the API. Using the X driver was the
> only solution available for Nvidia/ATI hardware.
>
> This still needs to be classified as a temporary solution. Long term
> the code needs to be extracted from X and converted to a standalone
> build system. They could be turned in to real EGL drivers at that
> point.

Exactly. Using the X7 drivers would be a good starting point for the userspace 
side of things. I've always planned on this. Moving away from using the X7 
drivers towards ones built specific for the purpose is also in the plans I 
have.

DRH
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/