Re: Similar project for Windows. How about dual-head

[Fredrik T. <fr...@do...>]

On Thu, 2005-04-28 at 10:58 -0700, yitzhak bar geva wrote:
> Here's a link to an article about a similar,
> proprietary project done here in Israel for MS
> Windows:
> 
> http://www.haaretz.com/hasen/spages/569358.html
> 
> I spoke to the developer today. What is interesting in
> his work is that he uses Nvidia dual port boards and
> gets a separate session on each one.
> The Nvidia drive documentation has an XConfig layout
> (near the end) which describes such a scenario. I had
> asked Aivils about this and he said it won't work with
> Ribui (Ruby). I never tried it.
> What are the chances that we can get separate sessions
> on each port of an Nvidia dual-port board?
> Does it require work on the part of the Xorg
> developers?
> What are the technicalities involved?

The problem is that the X server handles one (or more) *card* at a time.
Two X servers cannot share a single physical card.

The way to solve this would be to write a kernel framebuffer driver that
presents one framebuffer device to userspace for each head, and then run
two X servers that use these framebuffer devices instead of controlling
hardware directly (as they should be doing either way). The problem with
this, as far as I know, is that either the framebuffer device interface
isn't accelerated, or that there is no X server capable of using the
acceleration. I also don't think that there is a framebuffer driver that
does this with nVidia cards, but writing such a driver would probably be
the smallest subproblem involved.

At least this is my take on the situation.

Fredrik Tolf