> 1. If two implementations produce exact results for > a computation that did not involve exact intermediate > results, the two ultimate results will be mathematically > equivalent. I asusme you mean "did not involve inexact intermediate results". Kent