Re: R: R: [Plib-users] Transparent texture

[Steve B. <sjb...@ai...>]

Paolo Leoncini wrote:
> 
> Thanks, Steve, for the useful suggestions - I finally saw those
> semi-translucent textures in my database. What I did? these simple three
> things:
> 
> 1. glBlendFunc( GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA )
>    so to not actually need alpha planes still getting roughly good blending;

Yep - only a few rather esoteric algorithms actually need destination
alpha - and quite a few graphics cards don't support them.

> 2. mat->enable( GL_BLEND ) and mat->setAplhaClamp(0.01)
>    the latter to get the full [0:255] alpha range;

Yep.
 
> 3. switch for tests to a AccellGraphics GMX2000-equipped NT box (!!!!)
> instead of my G200 on NT, which soon switch to software with as few as
> 2.5Mtexels loaded and doesn't want to know about alpha-ed textures (latest
> driver, of course).

I'm not familiar with Windoze drivers - but a board shouldn't switch
to software rendering just because you are using too much texture memory.
It might take a while to swap maps in and out of memory - but it's probably
still hardware accellerated.

Dunno why G200 would have trouble with alpha - it works OK under Linux.

> Since I would like to rely on (3DS) modeller's information as much as
> possible, a further little work will focus on automatically set those
> properties for transparent textures (only the image loader knows about the
> 2/4-components) - the AppStateCallback seems once again the final resource.

Well, I guess whatever the problem is should be fixed - but I still believe
that most applications will want to use AppStateCallback in the end.

> Furthermore one has also to "isolate" translucent objects from the rest of
> the database for rendering with z-buffer disabled after all other objects,
> for best blending results.

SSG does that automatically for you.

-- 
Steve Baker                  http://web2.airmail.net/sjbaker1
sjb...@ai... (home)  http://www.woodsoup.org/~sbaker
sj...@ht...      (work)