> Hello,
> there was some related discussion on (to?) this topic a while ago. As
> far as I understood it, one can (should?) use something along
> integer, parameter :: sp = kind(1.0)
> integer, parameter :: dp = selected_real_kind(precision(1.0_sp)+1)
> ....
> when building the libraries, to create generic interfaces for the
> available precisions. The user of the library can then specify the
> necessary precision for his problem.
>
> Good luck
> Alois
The issue of writing generic procedures is covered in
detail in "Writing Portable Fortran Procedures",
and appeared in the August (?) 1999 issue of Fortran Forum.
It covers declarations, including the selection of maximum
precision, in a portable way.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|