Re: amber-developers: fortran95 help requested

From: Michael Crowley <>
Date: Fri, 2 Dec 2005 16:37:54 -0700

Dear all,
Upon reading carefully the f95 spec as it is in the ibm manual,
it seems like we have little choice but to only use variables and arrays
exactly as declared in both calling routine and called routine.
We probably only violate this in the real(8)/complex declarations in the
fft work but watch out for it. Basically, I think we, or at least I, have
to stop thinking of fortran passing pointers that we can do with what we
please as long as it is correct.

Now the compiler forces you to do
something much more restrictive when it can enforce it, and, more
dangerous to people like me, assumes that you have followed the rules and
may do something I was not expecting based on that assumption.

Ignore my call for help, I will recode according to what fortran wants me
to do even though it makes the code somewhat non-intuitive in the case of
the real-to-complex fft routines.

As long as you do not use pointers of targets, you can still change
dimensions and ranks of arrays between subroutines and calling routines.

Physical mail: Dr. Michael F. Crowley
                 Department of Molecular Biology, TPC6
                 The Scripps Research Institute
                 10550 North Torrey Pines Road
                 La Jolla, California 92037

Electronic mail:
Telephone: 858/784-9290
Fax: 858/784-8688
Received on Wed Apr 05 2006 - 23:49:46 PDT
Custom Search