Re: TiBook Dual Display
On Thu, 2001-10-18 at 19:37, Michael Schmitz wrote:
> > > Sure, as a first approximation I'd perhaps recommend to go without
> > > acceleration anyway. I'm not sure how much more synchronization than plain
> > > 'wait until accel idle' would be required (which happens anyway IIRC).
> >
> > You don't necessarily have to wait for the engine to be idle. You do have to
> > realize the current context isn't the one of the head you want to do an
> > operation on. If you want to do that transparently inside the framebuffer
> > device, you have to map the MMIO area twice and fault on access to the
> > inactive one, saving and restoring registers which could make a difference,
> > potentially all of them. I doubt that would be very efficient.
>
> Sounds like an interesting concept though (I forgot about mapping MMIO
> twice but you're right on that; that's the cleaner way).
> Does Linus allow hacks like having special page fault handling code for
> that purpose? Or would you rather have a signal handler in user space
> handle that?
Probably the former, anyhow I'd rather avoid having to deal with this. ;)
> What are the relevant registers for the accel BTW? Just a few might not be
> too much overhead.
If you want to go down this road, you have to be prepared for anything, not
just the typical 2D acceleration as done by the X server.
> > > > > X gurus please correct me if the current X release already supports
> > > > > Xinereama on rage128 out of the box
> >
> > Actually, that's the wrong question BTW. Xinerama is a driver independent
> > extension. My answer is to a question along the lines of 'does the r128
> > driver support single-chip dualhead', which I'm sure you really meant to ask.
> > :)
>
> Right - having Xinerama on two different cards should be OK already.
> I meant to say Xinerama on two framebuffers that map to the same card
Again, this isn't about Xinerama. Xinerama automatically works with any
screens, however they may be implemented.
This is about implementing two screens with a single chip.
(I suspect you say 'Xinerama' when you mean 'multihead', which isn't
equivalent)
> (this would imply UseFBDev probably?).
Depends if the implementation of the two heads is done in the
framebuffer device or in the X server driver.
--
Earthling Michel Dänzer (MrCooper)/ Debian GNU/Linux (powerpc) developer
XFree86 and DRI project member / CS student, Free Software enthusiast
Reply to: