dly wrote: > I understand that when I design algorithms I must be wary of all > these things that you mention > > But if I use a J primitive or a library function that I am not able > to examine, how am I to determine the limits of its capability?
Donna: You have to keep in mind that floating-point operations are always approximate. It is reasonable to assume that J primitives and library functions work correctly, but within this limitation: even addition is not associative. Your algorithms must take this into account. Many problems are numerically stable, and the errors will not matter, but unstable problems are not rare. The computer will do what you tell it, but you have to interpret the result, and there are no easy answers. Best wishes, John ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
