A simple Lambertian reflection model is used.

That is and analogously for .

Memory unscalabilities can be removed very cheaply; see [Skjellum:90c].

And more at trial timepoints which are discarded by the integration algorithm.

Optimality per se hinges on what our objective is. If, for instance, we want minimum time for LU factorization, the objective of minimum fill-in still does not guarantee minimum time in a concurrent setting.

Guy Robinson
Wed Mar 1 10:19:35 EST 1995