Eric Salo replied:
>Well, except for this annoying little problem called backward compatibility. On
>our machines, REAL means 32 bits and I don't expect that to change for a *very*
>long time. Yes, 64-bit values are the future (and even the present) but I claim
>that they will rarely be called REAL; ...
The issue here is not whether there will be a backward compatiblity problem.
There *will* be such a problem, no matter which way we go.
The real issue is who will have the problem. Most of our users are
accustomed to using Cray Research systems. These users have all known
for years that REAL means 64 bits. Moreover, they actually need 64
bits.
Eric continued:
>...they will be REAL*8, or REAL(kind=8), or
>DOUBLE PRECISION. But not REAL.
The first, while prevalent in some circles, is not standard. The
third does not address the needs of those who do 64 bit arithmetic
with complex numbers (DOUBLE COMPLEX is also not standard).
Building the solution to this with KIND seems, at first blush, a
sensible approach.
-Ian
-- Ian E. Stockdale MRJ Technology Solutions, NASA Ames Research Center ies@nas.nasa.gov (415) 604-4628 (use ies@acm.org for non-work email)