legacy code with real*16

[Pierre-Matthieu anglade]

If you keep using x86 hardware, you might not find some hardware
quadruple precision numbers. Still you may  give a try to libs that
implement it in software.
I don't know if this solution might be suitable for your case but if
you translate to C (you can use f2c to do that)  then use cephes
(http://www.moshier.net/) to implement extended precision you might
then get (slowly) the precision you want.

--
Pierre-Matthieu Anglade