On 6/8/06, Mikko J Rauhala <mjrauhal@...> wrote:
> On to, 2006-06-08 at 16:31 +0200, Roland Scheidegger wrote:
> > Mikko J Rauhala wrote:
> > > In particular, were I to buy a couple
> > > of 1920x1200 screens, would I still be able to get 3D rendering for
> > > the whole 3840x1200 desktop area? Given the development of XGL and
> > > AIGLX, I'd like not to paint myself into a corner where I could not
> > > use a fully GL-rendered desktop.
> > I thought the limit was 2560 too, but obviously not. There's a limit in
> > the driver for the surface regs, which is 3968. But these are only used
> > for tiling, so it _might_ work up to 4096. No guarantees...
> So one could probably cautiously suppose that it would work for 2x1920,
> then. Actually, come to think of it, my whole question was kind of silly
> since I should be able to test this myself by defining a virtual
> resolution 3840 pixels wide, correct? Or does the driver do some tricks
> like defining the rendering area to start at the origin of the viewport,
> and not the framebuffer? (In that case I _might_ still be able to coax
> some crappy 1920 pixels wide mode from my current CRT displays for
> testing purposes...)
A big virtual desktop would be a fine way to test.
> I'll probably do some testing later when I'm back home. Verification on
> the above is welcome.
> > This is hard to believe as the 0x554d id is in cvs since about half a
> > year before xorg 6.9/7.0 appeared so is definitely in there (secondary
> > isn't needed).
> Ah. Well.
> > It may not work however if you use the ati driver wrapper (it should be
> > fixed in cvs, but there are some reports it still doesn't work), you may
> > need to specify radeon as driver.
> I did specify radeon as the driver, yes. Reminded of this, I do not
> recall exactly in which order I did the xorg.conf tuneups that were
> necessary, so it might be that it would've sufficed to do this. However,
> also all DRI clients complain about an unknown ChipID, but then decide
> to just assume a plain r300, which seems to mostly work. (Is this
> assumption problematic in some cases, by the way, or is the warning
> something I should expect?)
> Anyway, if the ChipID is recognized in your code as you said, probably
> not your problem :]
> Mikko J Rauhala <mjrauhal@...>
> University of Helsinki