DRI2 status

[Kristian Høgsberg]

On Thu, Feb 21, 2008 at 12:33 PM, Tiago Vignatti <vignatti at c3sl.ufpr.br> wrote:
> Jerome Glisse escreveu:
...
>  > Yes please put DRI2 in separate files so i can import them in
>  > gallium without having to mess with older dri interface sitting
>  > in gallium right now
>
>  Also, with DRI2 we have to explicit pass --disable-dri2 flag to autoconf
>  because Xorg uses dri_sarea.h which is external to X server (from mesa).
>  Is this good enough?

Nope, these are definitely problems that need to be fixed.  Please
file bugs, and assign to me, otherwise I'll just forget.  I got sucked
into a Red Hat black hole shortly after committing the first DRI2
patches, and had to leave them upstream in a rather rough state.  Bad
timing, but I'm working fixing this stuff now.

I recently merged the intel_context.c files for i915 and i965 in mesa,
which should add DRI2 for i965, but I'm still testing that.  The
roadmap to getting DRI2 into a state where we can enable it by default
includes:

 - Implement DRI2 protocol so direct rendering can work.  Includes a
DRI2 protocol module, protocol code in the DRI2 module in the X server
and protocol code in libGL.

 - Implement DRI2 support in libGL.  I want to do this more like how
AIGLX works, with a vtable of GLX functions that the glX* functions
can call into.  There will then be a vtable for indirect, XF86DRI and
DRI2.

 - There's problems with resizing redirected direct rendering windows.

 - Get the xf86-video-intel intel-batchbuffer branch up to decent
performance and merge to master.  Not sure what state this is in at
this point, but I merged Eric's intel-batchbuffer branch from his repo
to the main repo, so all this work is now at least in the same branch.

Kristian