Shay wrote:
>...
> You said in a reply in your previous mesh post that the parsing time is over
> an hour. How reasonable do you think the parsing time might be in C?
>...

During the parsing of my "A Trivariate NURBS function" image
POV-Ray had to evaluate 3 NURBS functions for each sphere (one
function for each x, y and z coordinate, because POV-Ray does
not allow for arbitrary vector functions yet).

And when the sphere grid is 80*80*80 = 512 000 spheres, then
POV-Ray has to do 1 536 000 evaluations of NURBS functions in
order to place these spheres.

The way I have chosen to build the blending functions in the
macros I posted for the NURBS functions, are quite "quick and
dirty".

The resulting blending functions are quite non optimal,
because the recursive algorithm chosen causes some of the
calculations to be repeated several times.

At the moment I'm looking into other ways of building these
functions.

If one writes a C-program and uses more optimal blending
functions, then I think (or hope ?) that its performance
will be several hundred times better.


Tor Olav

Post a reply to this message