[Mesa-dev] Geometry shader update, and a course correction

[Kenneth Graunke]

On 07/22/2013 10:42 AM, Paul Berry wrote:
[snip]
> I don't mind if someone wants to do some research in the background on
> questions 1 and 2 above, but IMHO we can safely proceed with this plan
> even without that information.  In all likelihood, the only thing we're
> going to learn by answering questions 1 and 2 is whether we need to
> implement both GLSL 1.50 and ARB_geometry_shader4, or just GLSL 1.50.
> So in either case it makes sense to work on GLSL 1.50 first.  In the
> unlikely event that we run into another snag down the road and need to
> switch back to ARB_geometry_shader4, we can always pick up where we left
> off--I've archived the state of my "gs" branch in "gs-arb".

I agree wholeheartedly - implement 1.50-style GS first, since it's 
mandatory for OpenGL 3.2+ support, which everybody clearly wants.  We 
can later implement ARB_gs4 if we find applications that want it but 
don't have a 3.2-Core path.

Those array hacks sound pretty awful.  Not insurmountable, but in light 
of that, I'd much rather see the 1.50-style GS support be fundamental, 
with ARB_gs4 built atop that infrastructure, rather than the other way 
around.

--Ken