[mpich-discuss] Is it okay to remove F77 support?

Zhou, Hui zhouh at anl.gov
Thu Jan 28 08:37:13 CST 2021


No worries. For the record, your first letter of panic is not minded and welcome. It is important for us to care about what user worries.

Just to explain, MPICH, as an implementation of MPI specification, will never propose dropping part of standard such as mpif.h. In fact, we even keep deprecated features as long as we can unless it conflicts with new feature. Even then, we’ll consider configure options if the change will have significant impact. Rest assured, mpif.h is very much safe.

Just to explain my original question. There is always a risk when MPICH is compiled by certain Fortran compiler, and the user code is by another Fortran compiler. Different compilers are not necessarily binary compatible and may not link against each other. So if MPICH requires newer compiler, and your code requires some other compiler, then that is a potential concern that needs study. My question was to see if such concern exist.

--
Hui Zhou


From: Gus Correa <gus at ldeo.columbia.edu>
Date: Thursday, January 28, 2021 at 1:01 AM
To: Zhou, Hui <zhouh at anl.gov>
Cc: discuss at mpich.org <discuss at mpich.org>
Subject: Re: [mpich-discuss] Is it okay to remove F77 support?
Dear Hui Zhou.

Thank you.
On your second posting on this thread I understood
the problem boiled down to a misunderstanding caused by the
subject line, and that the support for the MPICH Fortran-77 API
was not at risk of being dropped.
That took away the worries, and I recognized this in my subsequent posting.
After that, other than Jeff Squyres clarifying posting,
everything else didn't really matter, was just noise,
as the issue had been already clarified,
the problem was solved.

I appreciate your efforts and those of the MPICH team,
to keep the project user oriented.
In the past I have posted questions on this mailing list,
and I received valuable (and valued, rest assured) help
from Pavan Balaji and other list subscribers.

Although my initial concern was genuine
- there is a lot of Fortran-77 code here and out there that uses and depends on MPI and MPICH
and that would be orphaned if the Fortran-77 MPI API was to be dropped -
I should have toned down my first posting,
and I apologize to you if it sounded rough.

Thank you,
Gus Correa


On Wed, Jan 27, 2021 at 11:23 PM Zhou, Hui <zhouh at anl.gov<mailto:zhouh at anl.gov>> wrote:
Dear Gus Correa,

There are some misunderstanding, not all posts are coming from or representing MPICH team. The original question, by me, is from MPICH team, and I personally, not the team nor Argonne, take the full responsibility of “badly written subject line”. Most other posts are from different entities and most of them only reflect their personal views anyway. As a discussion mailing list, we should assume all posts represent personal view by default. That said, all inputs are valued here. Frank opinions and voices are what makes better software and better user experiences. Since we have spent several rounds of email clarifying the question, and since you understood the initial false alarm and misunderstanding, I am hoping we can put this behind us and not to hold grudges or at least not to hold it long.

If you believe, after clearing the initial misunderstanding, that the topic is not of interest to you, please ignore rest of the thread. Meanwhile, I do believe the question is meaningful to ask. So please don’t dismiss the discussion just out of your perspective.

--
Hui Zhou


From: Gus Correa via discuss <discuss at mpich.org<mailto:discuss at mpich.org>>
Date: Wednesday, January 27, 2021 at 9:27 PM
To: MPICH-Discuss <discuss at mpich.org<mailto:discuss at mpich.org>>
Cc: Gus Correa <gus at ldeo.columbia.edu<mailto:gus at ldeo.columbia.edu>>
Subject: Re: [mpich-discuss] Is it okay to remove F77 support?
Oh Mr. Hammond, you guys are just die-hards, insisting to defend what you just did,
namely alarm your community of users with your terribly phrased question.

The original question, at the very subject line of the email,
was very poorly formulated, and suggested the end of mpi.f and such,
which are essential in code written in all sorts of
Fortran-77 code (and various F77 extensions) that abound
among your users.
Don't you care about these users (us)?

Being pedantic, belittling your users alarm after your horrible question,
or berating your users knowledge of what is and what is not strict F77 ISO standard
doesn't help at all.
After all, we, your users, in one way or another,
as taxpayers and in other ways,
support the development of publicly available high quality
software in national labs, such as mpich, expecting that it will
be useful and usable, and work together (not against) the codes we have.
So presumably we  deserve a little better.

The bad question triggered a whole discussion,
which subsequent answers showed to be a discussion about nothing.
Period.

Jeff Squyres pointed out that it was never possible to build
MPI with a compiler strictly adherent to Fortran-77 ISO standard.

So, why, as the initial question asked, do you need to remove
support for something that never existed to begin with?
Why do you need to ask this non-question to the community,
and raise the false alarm?
Huh.

Indeed, I never paid much attention to that,
but admittedly I have been compiling OpenMPI, MPICH and MVAPICH
all with F90 compilers (gfortran, pgfortran, ifort), although often
times using their Fortran-77 versions (with extensions to the ISO standard I suppose),
to compile code that depends on
Many of those codes do depend on mpi.f, won't work with F90 modules (or 'use mpi').
I guess this is not unique at all, and a lot of the code base out there,
in various areas of knowledge, works this way.
The original question clearly suggested suppressing this possibility.


Gus Correa

On Wed, Jan 27, 2021 at 12:16 PM Jeff Hammond <jeff.science at gmail.com<mailto:jeff.science at gmail.com>> wrote:
Squyres' point is that there is not and has never been an MPI Fortran 77 API and every time someone implies there is, it confuses people.  mpif.h requires ISO Fortran 90 - it just happens to support fixed-source form and does not require any of the Fortran 90 features that were not ubiquitous in compilers ~15 years ago.

The original question was correct, but could not be interpreted correctly by most people.  It asked whether MPICH could drop support for compilers that support no Fortran dialect beyond Fortran 77.  The only compiler I have ever used that would be impacted by this is g77, which I used in the same era that I used Itanium.

Jeff


Well, then the question was very poorly formulated.
It suggests suppressing the Mpich Fortran-77 API.

Hard to gauge the impact of removing support of F77 compiler (ONLY),
but it may cause trouble when building complex code,
and linking it to MPI (mpich),
as some may actually require a compiler that is specific for Fortran-77,
or have Makefiles and other utilities that rely on such a compiler.

Gus Correa
_______________________________________________
discuss mailing list     discuss at mpich.org<mailto:discuss at mpich.org>
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss


--
Jeff Hammond
jeff.science at gmail.com<mailto:jeff.science at gmail.com>
http://jeffhammond.github.io/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpich.org/pipermail/discuss/attachments/20210128/ca4f6db7/attachment-0001.html>


More information about the discuss mailing list