I am working on a viewshed* algorithm that does some floating point arithmetic. The algorithm sacrifices accuracy for speed and so only builds an approximate viewshed ...