[Mpi-forum] MPI supported languages

Skjellum, Anthony Tony-Skjellum at utc.edu
Sun Apr 25 14:47:16 CDT 2021


Benson and Dan, thanks.

Martin Ruefenacht, Derek Schafer, and I plan to submit a proposal for a new language interface (not language binding) for C++17/20 when MPI-5 convenes.  The aim is to generate more performance and productivity.

My group is also working on two separate themes with colleagues at PNNL and LLNL on exploratory C++ interfaces for MPI for specific learning/prototyping.  I am sure we will publish some joint papers before MPI-5 starts with these colleagues.

We are likely to extend that thinking to Python also.

Martin and I are hoping to form a WG around these OO languages.  We also, want to do a type theory for the entire standard; there are many explicit and implicit types in MPI (not just containment by integers or such).  This has become clear as Martin and Jeff Squyres developed the Python technology for the MPI-4 document itself.

Note that (much) more can be done with Fortran-2018 too that goes beyond the imperative language binding that is pretty much identical to the LIS.

One hope is that we can separate out the language interfaces from the main standard, and let them have space to develop properly and be extended to more languages than C and Fortran.

Among other learning, some whole classes of helper objects may fall by the wayside in a modern C++ language interface, such as MPI_Info.

Tony



Anthony Skjellum, PhD

Professor of Computer Science and Chair of Excellence

Director, SimCenter

University of Tennessee at Chattanooga (UTC)

tony-skjellum at utc.edu  [or skjellum at gmail.com]

cell: 205-807-4968


________________________________
From: mpi-forum <mpi-forum-bounces at lists.mpi-forum.org> on behalf of Dan Holmes via mpi-forum <mpi-forum at lists.mpi-forum.org>
Sent: Sunday, April 25, 2021 10:10 AM
To: Benson Muite <benson_muite at emailplus.org>
Cc: Dan Holmes <danholmes at chi.scot>; MPI Forum <mpi-forum at lists.mpi-forum.org>
Subject: Re: [Mpi-forum] MPI supported languages

Hi Benson Muite,

Thanks for this good list of unofficial bindings and projects using MPI or MPI-like interfaces.

The short answer to your question is: yes, there is an effort underway to provide better support for other programming languages.

Some work was done for the MPI-4.0 release of MPI (due this year) to firm up some of the key semantic terms, which will aid us when attempting to express those semantics in new languages. That work is continuing and more results from that are targeted at the upcoming MPI-4.1 and future versions of MPI. Figuring out the dividing line between what is currently being expressed (semantics) and what is dictated by the current mechanism of expression (C or Fortran) is a first step towards new forms of expression.

Beyond that work, additional language expertise will be needed from some folks who are regular users of these other programming languages so that appropriate expressions of the essential MPI semantics are chosen — language mechanisms that “fit” with the normal/natural ways that a typical programmer would recognise. One possible strategy that has been discussed is to split the MPI Standard into a single MPI Core Standard and multiple MPI Bindings for <language> Standards. This might happen as early as MPI-5.0 (perhaps 5-10 years away) depending on how controversial it turns out to be and the rate of progress, which will depend heavily on the amount of interest and sustained contribution from a community of users.

Cheers,
Dan.
—
Dr Daniel Holmes PhD
Executive Director
Chief Technology Officer
CHI Ltd
danholmes at chi.scot<mailto:danholmes at chi.scot>



On 25 Apr 2021, at 09:36, Benson Muite via mpi-forum <mpi-forum at lists.mpi-forum.org<mailto:mpi-forum at lists.mpi-forum.org>> wrote:

MPI has official bindings for C and Fortran. There are unoffical bindings in

C# - https://github.com/rogancarr/MpiDotNet
D - https://github.com/DlangScience/OpenMPI)
Go - https://pkg.go.dev/github.com/cpmech/gosl/mpi and https://github.com/marcusthierfelder/mpi
Java - https://www.open-mpi.org/
Julia - https://github.com/JuliaParallel/MPI.jl
Lua - https://github.com/jzrake/lua-mpi and https://peter.colberg.org/lua-mpi
Ocaml - https://github.com/xavierleroy/ocamlmpi
Perl - https://metacpan.org/pod/Parallel::MPI
Python - https://bitbucket.org/mpi4py/mpi4py
R - https://github.com/RBigData/pbdMPI and https://cran.r-project.org/package=Rmpi
Ruby - https://github.com/gfd-dennou-club/ruby-mpi
Rust - https://lib.rs/crates/mpi
Scala - https://github.com/mpokorny/scampi


It is also used as a possible backend for among other PGAS languages XcalableMP - https://xcalablemp.org/

Some people did not expect MPI to be used by many developers directly. Are there any thoughts on how better support for other programming languages and MPI as a backend might be improved in future?
_______________________________________________
mpi-forum mailing list
mpi-forum at lists.mpi-forum.org<mailto:mpi-forum at lists.mpi-forum.org>
https://lists.mpi-forum.org/mailman/listinfo/mpi-forum

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpi-forum.org/pipermail/mpi-forum/attachments/20210425/9e0f8e0b/attachment-0001.html>


More information about the mpi-forum mailing list