I recently wrote an application that uses the NURBS package from GLU. In my model code, I represent NURBS with double precision andI ensure that knot multiplicity does not exceed the order of the Spline to within double precision…
What I have found is that sometimes a Spline that is valid with double precision is not valid with float precision because the knot multiplicity becomes larger than the order due to the loss of accuracy.
Are there any simple techniques/heuristics for repairing this problem?
Thanks for any suggestions,