> > > Read Mark Kilgards paper on it aat SGI
> >
> > I'm not affiliated with the GGI project. The only reason I suggest KGI drivers
> > is: they exist.
>
> Read the direct render paper by Mark Kilgard, then you'll understand why the
> KGI model isnt relevant.
I searched SGI's site and only found Kilgard's paper on "multi rendering" (multi
threaded rendering within the X server for interactivity). He does mention direct
rendering in passing, which seems to be simply SGI's term for an X client itself
(not the server) writing to the hardware. Certainly this needs to be possible (it's
what I want to do), but:
a) I don't think the X clients should be restricted to X APIs for direct rendering.
b) This has nothing to do with what graphics API the server itself is using to do
the rendering.
I simply think that *all* graphics within Linux should go through a device driver
to avoid hardware contention. If that is done, then there is no reason why an X
client can't "direct render" using whatever API it chooses - if the X server and
all graphics APIs go through the device driver, then it'll be fine.
>
> > The sooner we put at firm foundation under the Linux graphics the better. It
>
> Never pour concrete until you have planned what goes on top in detail. Its
> very embarrasing trying to put a square house and a round foundation, however
> firm.
True, but I don't think the goal is to just support one "house" /API such as Xlib,
but rather to provide an extendable infrastructure that will support many APIs and
take full advantage of graphics cards with many differing capabilities. This is
what my proposed architecture is intended to do.
I really think that the details of the initial driver interface/functionality, as
long as extendible, are somewhat unimportant! The goal is simply to provide
portable, safe direct access to graphics hardware capabilities, not to predict what
API edifices people wish to construct on top of that foundation.
Ben Bridgwater
Please CC: bennyb@ntplx.net
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/