Input Devices

[Michel Dänzer]

On Thu, 2005-08-04 at 18:48 -0700, Eric Anholt wrote:
> On Thu, 2005-08-04 at 15:48 -0400, Michel Dänzer wrote:
> > On Thu, 2005-08-04 at 15:08 -0400, Jim Gettys wrote:
> > > Hopefully we'll be using GL or exa drivers soon enough that this will be
> > > moot.
> > 
> > I don't see how EXA would fix this though as it stands; it still seems
> > to have a static partition of the framebuffer into screen and offscreen.
> > To be able to grow beyond the original size, the screen would have to be
> > allocated from the EXA memory manager, wouldn't it? Also, can all parts
> > of the X server deal with the screen moving around in the framebuffer?
> 
> All the previous EXA (KAA) drivers allocated the real screen statically,
> but that's not necessary.  You could just allocate it out of the normal
> allocator locked, and free it and try to allocate the new size when you
> need to resize.  

Sounds intriguing.

> I can't really imagine any problems with the screen moving, except for 
> the DRI needing to notify of front/back/depth moving (anyone know better 
> how much of a problem this would be?  particularly client driver 
> notification).  

AFAIK it would require extending either the DRI extension protocol or
the driver specific DDX<->3D interface. The tricky part would be how to
deal with old clients that assume the buffer offsets to be static.

Incidentally, this is probably also one of the major obstacles for
proper integration of the DRI and Composite.


-- 
Earthling Michel Dänzer      |     Debian (powerpc), X and DRI developer
Libre software enthusiast    |   http://svcs.affero.net/rm.php?r=daenzer