GL Redirection to offscreen
ufz6 at rz.uni-karlsruhe.de
Sat Oct 30 04:43:03 PDT 2004
On Sat, 2004-10-30 at 13:19, Keith Whitwell wrote:
> Amir Bukhari wrote:
> >>Have you made any progress with the redirect issues? It seems like in
> >>September you were coming to terms with some fairly deep questions about GL
> >>and composite - how is that panning out?
> > I have done some experiment with software Rendering in Xorg and I end up
> > that Composite work OK with, but for DRI it needs alot of change in DRI
> > structure.
> > what I am thinking about how GL redirection work is like following:
> > 1- Composite Extension should define interfaces for others X parts (GL,
> > Xv ...), which it call those interfaces if they exists.
> > 2- each Part of X need a redirection should implement those interface
> > and register them by Composite extension.
> What would this interface look like?
Composite define array of a structure, which define well known functions
(like redirectWindow, unredirectWindow). those functions tell the other
part that window is redirected to a pixmap. for example DRI should now
tell its subpart to put the end image to associated Pixamp. I don't know
if the last work (put the end image to that pixmap is possible or not),
but I think if not work we should find how Xserver can access the
pbuffer of video card.
> It seems like it would almost have to replace the current GLX/DRI interfaces,
> but maybe I'm reading too much into it. All the open drivers currently use
> shared Z and depth buffers, for instance - I don't see how that could continue
> under composite.
> > 3- the same should be done with Damage Extension, which is related also
> > with Composite.
> > this appraoch give more flexibilty for others GL Vender, to support GL
> > redirection and could be also get others part of X which can not be
> > handled directly with Composite extension like Xv.
> xorg mailing list
> xorg at freedesktop.org
More information about the xorg