> It is really a problem. > It does disqualify `float' and `double' > for a lot of applications on Intel. It rather disqualifies the those applications, which are not following basic principles of mathematical numerics. Please show me a standard which says that your program is guaranteed to output 1000.0. Regards, Martin