Re: canonical external representation of Fortran 90 datatypes

W. Saphir (wcs@nersc.gov)
Wed, 19 Mar 1997 10:15:09 -0800 (PST)

> Yes, 64-bit values are the future (and even the present) but I claim
> that they will rarely be called REAL; they will be REAL*8, or REAL(kind=8), or
> DOUBLE PRECISION. But not REAL.

Anyone trying to write a code that is portable between an SGI
and a Cray^H^H^H^H SGI machine will probably use REAL(kind=something).
Note that REAL(kind=8) falls into the category I'm trying to address
(base class of REAL with modification through KIND).

An amusing declaration I've seen a few times is:
real(kind(0.0d0)) x

Bill