Aaron, Depending on what you mean by "large decimal numbers." If you require "arbitrary-precision", then the BigDecimal class is the way to go. There are other nice things about the BD class that you don't get with the primitive float/double type, but nothing is ever free. Using a class is going to cost you in performance. Unless you need the overhead, I recommend sticking with a primitive double which has The java.lang.Double.MAX_VALUE = 1.79769313486231570e+308!!! -Peter
Joined: Aug 16, 2001
Maybe I should add that I would be willing to sacrifice some performance for accuracy. The more accurate the better, but how much better is the accuracy between the two implementations.