[CIG-SEISMO] On "USE"ing MPI

[Elliott Sales de Andrade]

Hi all,

With the somewhat recent updates to default compiler options to use
Fortran 2003 standards, I have noticed that most MPI implementation
headers are pretty non-compliant.

With GNU Fortran, this tends to abort the compilation entirely. On IBM's
xlf, it prints nearly 100 warnings for each file using mpif.h. I'm not
sure that the other compilers have the 2003 standard enabled.

I believe this stems from the use of the Fortran MPI header file instead
of the module. That is, the code does "include 'mpif.h'" instead of "USE
mpi". The former seems to invoke Fortran77 style code, while the latter
is Fortran90 style.

Anyway, the question is, whether we use a Fortran90 compliant compiler +
MPI library everywhere? Can I just do a mass replace on everything?
(Actually, it's a bit more complicated than a simple replace, since the
order is different, but you get the idea.)

-- 
Elliott Sales de Andrade <esalesde at physics.utoronto.ca>
University of Toronto (Physics)