Arthur, I was planning on doing the coarse-grained backface culling once, before the graphics pipeline is even set up. Basically, I would start with an indexed face geometry. And then store several sets of indices. Each set of indices would represent the back face culled version of the original geometry, but from some different point in space. Perhaps I would choose 6 or 8 reasonabbly distributed points in space.
The idea is not fully cooked of course.. thuus making it half baked! But I wanted to toss it around to see what people thought. I am starting to feel that although this could improve performance in some ways, it will come at the cost of lots and lots of memory. Thoughts? Ian On Tue, 8 Jul 2003, Artur Biesiadowski wrote: > Ian M Nieves wrote: > > > My thought is that the graphics card still has to transform ALL faces that > > are sent to it, before it can determine which are back facing. And then > > perhaps, by eliminating back facing faces, some processing is saved in > > texturing, lighting, etc. > > And to determine it yourself, you need to transform vertices on CPU - > which is a LOT more costly that throwing it at GPU (for modern cards at > least). > > Generally, think about object-level optimalizations, not-triangle based > ones. For triangles, just find a fastest way to bulk transfer it to GPU > and let it handle the rest. > > 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". > > =========================================================================== 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".
