Yes, I was wondering about that. In this very simple network topology, the equations can be phrased tridiagonally, and so admit solution by the algorithm of Thomas, which requires linear time, but I wasn't sure how that generalized to the arbitrary topologies admitted by Gnucap, at least without internal renumbering as suggested by the other poster, and also how reasonable it was to expect a general code to recognize a relatively trivial special case. Presumably a completely connected network yields a full matrix, which can't be solved in less than cubic time, can it?
(I'm yet to convince myself that halving the number of nodes can change the index of complexity, but as you assure us, I'll ponder it further.) On Feb 7, 2012 6:11 PM, "al davis" <[email protected]> wrote: _______________________________________________ Help-gnucap mailing list [email protected] https://lists.gnu.org/mailman/listinfo/help-gnucap
