After all, i would be glad if someone could produce some benchmarks on real data in real Java3D vs. GL4java projects. I doubt that it would be a big surprise... (but who knows? Personally never made one...)
Problem is that by reducing point of interest to 'real projects' you open a way to ignore any benchmark/test which can be given...
I'll say it this way. For functionality that java3d exposes, I do not see a point in proving which API is faster. Java3d can be a bit slower versus heavily optimized opengl implementation, quite faster than naive opengl version. But we are talking about same order of maginitude here.
For me, main problem is with all advanced functionality. Try to implement shadow volumes in java3d. Any kind of funny effects using pixel shaders. Even not so complicated tricks like bumpmapping with ambient+diffuse+specular. Only way to do it, is to do everything in software - probably 100 or 1000 times slower.
I'm not sure how fast render to texture is with current by-ref textures, but I have a bad feeling that it has to be copies between GPU and main memory there and back (please correct me if I'm wrong).
You can tell that no 'real application' needs shadow volumes, pixel shaders, true bumpmaps, etc. This is not valid argument for me - same as saying than nobody will ever need more than 640kb of memory :) There is also a chance that most of this functionality will be exposed in java3d 1.4 - but before we will see at least API project (or even better beta reference implementation) it is non-existent.
I use java3d for all of my for-fun visualisation projects and I'm 100% happy with it's performance and capabilities. But sometimes I run NV Effect Browser and a bit of jealousy appears somewhere inside...
Artur
=========================================================================== To unsubscribe, send email to [EMAIL PROTECTED] and include in the body of the message "signoff JAVA3D-INTEREST". For general help, send email to [EMAIL PROTECTED] and include in the body of the message "help".
