Xinerama infrastructure

[Mike A. Harris]

On Mon, 16 Feb 2004, Steve Salazar wrote:

>Hi.  I am not an X developer so forgive me if I get some of my facts/ideas 
>wrong.

You're forgiven for being wrong.

>Xinerama is currently not accelerated under XFree86.

Actually, that makes no sense.  Xinerama combines multiple
screens into a single display, thus giving a root window that
spans 2 or more monitors.  Each display can be on a single video
card, or can be multiple heads on a dualhead card such as a
Radeon 7500 for example.  "Accelerated xinerama" does not make 
any technical sense really.

XAA is the acceleration architecture and unless you purposefully 
disable it intentionally, it works just fine with Xinerama.

>My understanding is that xinerama and acceleration were simply
>not built into the XFree 4.x architecture in a way that would
>make implementing accelerated xinerama in XFree86 a reasonable
>task for any but the most highly motivated individuals (none yet
>have emerged apparently)

Again, "accelerated xinerama" is a nonsensical term.


>So, my question is this:  is accelerated xinerama something that
>is even being considered at this early architecting stage of
>xserver development so that when the time comes to implement it,
>it will be reasonably easy and will actually happen?  Is the
>fd.o xserver architected in a way that xinerama will fall
>conveniently within some abstraction that will make it so that
>we essentially get the multi-head support for free or with
>little overhead in terms of performance/coding
>effort/maintenance?

It's hard to 'consider' something that doesn't make any technical 
sense.  Either you don't understand what acceleration is, or you 
don't understand what Xinerama is, or perhaps both.


>Right now I am actually using Xig's x-server because I use dual
>head at work and the unaccellerated performance on dual head was
>totally unacceptable with XFree.  Please tell me that I will not
>still be in this same situation one year from now.

I don't know of any XFree86 DDX driver that disables acceleration 
when Xinerama is used, so the performance should be identical 
both with and without Xinerama.

Perhaps you are just not explaining yourself clearly.



-- 
Mike A. Harris