g95 errors

radsci2008 9srh8rw... at jetable.org
Fri Mar 21 09:08:26 UTC 2008

Hi All-

On Mar 20, 6:41 am, Teodoro Laino <teodor... at gmail.com> wrote:
> Let me comment on that.
> Few of us had a discussion yesterday about this point and we all
> realized that the g95 version you downloaded is non standard.
> In fact quoting these guys:
> http://www.mcs.anl.gov/research/projects/mpich2/support/index.php?
> s=faqs#g95
> "the Fortran standard requires that INTEGER and REAL be the same size"

Point taken, thanks.  In the event, -i4 caused many more problems.

> So.. if you download the other g95 (the standard one with 32-bit
> default integer) you won't need any modification to the arch file.
> And most important.. why should be worry for something that is not
> even standard?
> Already with standard stuff there are so many troubles around..

Actually, 64-bit default integer g95 DID compile cp2k sources from
20070928 or thereabouts, at a time when gfortran 4.3 had
difficulties.  Note that even the same arch file or flags therefrom
fail with 2008 sources.

> And back to my previous question:
> What is the reason why you downloaded the g95 default 64-bit integer
> version (if there's any..)?

It produced very good results with other codes (even before gfortran
was operational,) though this is not always the case.

Since at least some of these use 64-bit reals, it seems reasonable to
use compilers that give that on 64-bit machines.  If cp2k is really
intended for only the 32-bit version, a comment to that effect might
be added to the respective arch files to indicate that.


More information about the CP2K-user mailing list