Re: [Celestia-developers] GLSL Betterness!

[Chris L. <cl...@ww...>]

Pat,

This is great news! I was beginning to think that I'd never get this GLSL
stuff right on ATI cards.

> So, I finally had a chance to test the changes made by Chris 5/6 weeks
> ago to the GLSL programs, and there is definitely progress! Only two
> minor rendering errors exist, from what I can see...
>
> I have before and after shots that should explain everything nicely.
>
> Before:
>   http://pat.suwalski.net/celestia/misc/ogl2-clouds.jpg
>
> After:
>   http://pat.suwalski.net/celestia/misc/ogl2-earth-clouds.jpg
>
> As you can see, the clouds are fine now, the specularity isn't. But it's
> probably something really simple.

Interesting . . .  Are you using the standard Earth texture here? I take
it that the oceans are not supposed to look sky blue. It seems like the
specular highlight is getting smeared over the entire globe. It's not
clear to me what's happening here, but it sort of looks like the effect
you get with a very low specular exponent.

> Before:
>   http://pat.suwalski.net/celestia/misc/windows-ati-saturn-ogl2.jpg
>
> After:
>   http://pat.suwalski.net/celestia/misc/ogl2-saturn.jpg
>
> Much better! Only one shadow. The shadow from the rings on the planet is
> a little strong, but certainly much better than the one from the vertex
> path:
>
>   http://pat.suwalski.net/celestia/misc/windows-ati-saturn-vertex.jpg
>
> I assume that shadow looks better on the nVidia cards with the vertex
> path? Perhaps more like with the ogl2 path on the ati's?

With the vertex path, nVidia cards render ring shadows exactly as shown in
your image. Only the ogl2 path (and the soon-to-be-deprecated NV30 path)
render ring shadows correctly with ambient light. The darkness of the
rings in the ogl2 path is correct.  There was a lot of discussion a while
ago with Grant and Fridger about how dark the ring shadows should be.  The
dark shadows looked strange at first, but after seeing at all the images
from Cassini, the shadows seem quite natural to me.
>
> All in all, we're pretty close now!
>
> What else needs lots of work before the next release? I still have plans
> to split up gtkmain.cpp into several files to make it manageable. Also,
> there needs to be some work done on the autotools stuff, as usual.

I've been really slow with Celestia development, and for that I apologize.
We need to get 1.4.0 released. I'm encouraged by your positive reports on
GLSL on ATI hardware. If I can fix the specular highlight problem, then
I'd feel OK making the ogl2 path the default on systems that can support
it. Does anyone feel that there are outstanding bugs serious enough to
block a 1.4.0 release? I'd like to aim for an early May release. An
official release with OpenGL 2.0 and multiple star system support is long
overdue; we can worry about implementing further new features after 1.4.0.

--Chris