[Bf-committers] Blender unified shading language/Renderman export: 2.38?

Jonathan Merritt j.merritt at pgrad.unimelb.edu.au
Fri May 13 09:00:09 CEST 2005


Chris Burt wrote:

> The answer is pretty obvious. Blender isn't so "cutting edge and
> experimental". I would say it probably runs better on slow hardware
> than most other applications with the same features. That is the
> reason why hardware based shaders aren't implemented.


Well, there's no reason that the shaders can't be enabled for the
hardware that supports them, and disabled otherwise.  You asked for an
example, so take a look at Celestia ( http://www.shatters.net/celestia/ ).

However, the bigger problem might be in providing shaders that can
produce useful scene previews.  As I've already indicated, RenderMan
shaders would have needed a *lot* of parameters passed to them if they
were to truly mimick every aspect of Blender's own shading system.

It is probably very straightforward to create a Cg or OpenGL shader that
implements the basic Blender shader types (Lambert, Oren-Nayar, Toon,
Minnaert, etc.), and I wouldn't be bowled over if I saw this come out
within a release or two.  It takes me 5 minutes to do this in the
RenderMan shading language, afterall!  This basic stuff may be the first
simple step that could be taken towards hardware shading.  If I had a
graphics card that supported hardware shading then I'd be looking at it
myself!

However, as I've said, when you throw in color ramps, textures and all
of their various mappings, and the rest, it's a *MUCH* bigger task!  It
may even be beyond the capabilities of the current hardware shading
systems (eg: the required shader may be too big to fit into the shader
VM).  Even if it is "physically possible", then the job of implementing
it still shouldn't be underestimated.

-- 
Jonathan Merritt BE(Mech)/BSc
PhD Student - Equine Biomechanics
The University of Melbourne
Veterinary Clinical Centre, Werribee



More information about the Bf-committers mailing list