Charles R Harris wrote: > > I don't think it's that bad. Leaving out the ppc and sticking to ieee, > there is only double precision, extended precision and quad precision > versions of long double and they are easily determined at run time. How would you determine this at runtime ? cheers, David