[MPI3 Fortran] What to do with mpif.h in MPI-3?
nmm1 at cam.ac.uk
Tue Jan 26 07:53:34 CST 2010
On Jan 26 2010, Jeff Squyres wrote:
>On Jan 25, 2010, at 11:19 PM, Torsten Hoefler wrote:
>> That doesn't matter, if it's not in the F77 standard then we can't have
>> it in the F77 bindings.
>I remembered 2 other things that bothered me about this statement:
>1. In MPI-2.2:16.2.1, we say:
> "In the rest of this section, "Fortran" and "Fortran 90" shall refer
> to "Fortran 90" and its successors, unless qualified."
Given Fortran's attitude to compatibility, that's fine. There are more
official ways to phrase that, but it's clear.
> I haven't checked them all, but I would *hope* that all the places we
> explicitly mention "Fortran 77" in the document are within proper context
> and not globally scoped.
Yes. I haven't checked, either.
> 2. In the same section, we explicitly list what in MPI is incompatible
> with both Fortran 77 (including the 6 character thing that was already
> mentioned -- you couldn't even have MPI_SEND or MPI_INIT!) and Fortran
> 90. Both lists are already non-empty.
'_' wasn't allowed, either!
> I think this is a pretty strong statement that we do *not* need to be
> 100% compatible with the Fortran specification as it was 30+ years ago.
I want a Fortran II interface :-)
More seriously, Fortran 77 compilers are dead. Essentially all are options
to Fortran 90 compilers, and the meaning is usually to allow all of the
features of Fortran 90 that do not break the ABI, often combined with the
detection of all the major syntactic extensions introduced in Fortran 90.
I don't even know any significant, currently active codes that are still in
pure Fortran 77.
More information about the mpiwg-fortran