On Mon, 21 Mar 2005 10:44:33 -0500, Timothy Miller <[EMAIL PROTECTED]> wrote:
> On Mon, 21 Mar 2005 17:14:01 +0800, Joe Hsu <[EMAIL PROTECTED]> wrote:
> > If I understood correctly, to mmap a sub-region of ram with
> > being process-related(some kind of privillage control) can be
> > done within kernel(driver). It seems to need no hardware support
> > for this kind of mmap.
> 
> We cannot selectively map windows, but we could selectively map other
> parts of the graphics memory (in page-sized units), using the usual VM
> stuff in the kernel.

So if direct rendering clients could be made to render to a texture or
pbuffer instead of directly to the framebuffer, it would all work out
nicely.

The question is if that can work in a sane (and high performance)
manner within the existing  XAA/DRI/DRM framework. If not, it might
make sense not to worry about it and instead focus on supporting the
features a compositing, GL-based X server needs.

-- 
Will Dyson
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to