Further Boost.MPI development

Dear Boost.MPI users, With the MPI-3 standard having been finalized we want to intensify Boost.MPI development as soon as my crazy teaching schedule this fall gives me time from the middle of next month. With the deprecation of the - little used - official MPI C++ API in MPI-3 Boost.MPI has a chance of becoming a de facto standard C++ API for MPI. We first want to focus on features from MPI-2 and MPI-3 that are widely used or useful. For this purpose I would like to collect suggestions. Can you please reply and mention the functions of MPI that you would most like to see supported by Boost.MPI. Best regards Matthias

Nonblocking collectives are at the top of my list.
It would also be nice if the environment class could initialize thread support.
On Oct 28, 2012, at 5:33 AM, Matthias Troyer
Dear Boost.MPI users,
With the MPI-3 standard having been finalized we want to intensify Boost.MPI development as soon as my crazy teaching schedule this fall gives me time from the middle of next month. With the deprecation of the - little used - official MPI C++ API in MPI-3 Boost.MPI has a chance of becoming a de facto standard C++ API for MPI.
We first want to focus on features from MPI-2 and MPI-3 that are widely used or useful. For this purpose I would like to collect suggestions. Can you please reply and mention the functions of MPI that you would most like to see supported by Boost.MPI.
Best regards
Matthias _______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

Hi,
My priority would be the vector types with striding... The current way of
doing this is complicated in my opinion. I would like an API close to C,
but with all the C++ RAII stuff. For efficiency/HPC code, it is mandatory.
Cheers and thanks for this (I would like to see Boost.MPI being the only
C++ API :) ).
Matthieu
2012/10/31 Daniel Mitchell
Nonblocking collectives are at the top of my list.
It would also be nice if the environment class could initialize thread support.
On Oct 28, 2012, at 5:33 AM, Matthias Troyer
wrote: Dear Boost.MPI users,
With the MPI-3 standard having been finalized we want to intensify Boost.MPI development as soon as my crazy teaching schedule this fall gives me time from the middle of next month. With the deprecation of the - little used - official MPI C++ API in MPI-3 Boost.MPI has a chance of becoming a de facto standard C++ API for MPI.
We first want to focus on features from MPI-2 and MPI-3 that are widely used or useful. For this purpose I would like to collect suggestions. Can you please reply and mention the functions of MPI that you would most like to see supported by Boost.MPI.
Best regards
Matthias _______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users
_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users
-- Information System Engineer, Ph.D. Blog: http://matt.eifelle.com LinkedIn: http://www.linkedin.com/in/matthieubrucher Music band: http://liliejay.com/

Combine Boost.MPI, your array, multi_array, vector w/STAPL efforts. I'm just a minor user who knows nothing about the internals of any of these & who does not want to know but who wants more of boost making some obvious [to me] things easier. E.g., we already know arrays & decomposition so why not have processor tracking of who has data instead of user writing code to get processors to do it? I'm thinking better than just a wrapper for tools like PETSc, FFTW (MIMAS does fine but none allow basis functions other than Fourier like Hermite... note how I'm more interested in user items over API:). Imagine: a boost::vector where I don't have to specify extent, stride, serialization, env, comm, .... I'll shut up now since I'm obviously not in your field but want too much from your field. I do hope something happens that will help someone even if it occurs too late to help me. Thanks anyway! On Oct 31, 2012, at 2:54 PM, Matthieu Brucher wrote:
Hi,
My priority would be the vector types with striding... The current way of doing this is complicated in my opinion. I would like an API close to C, but with all the C++ RAII stuff. For efficiency/HPC code, it is mandatory.
Cheers and thanks for this (I would like to see Boost.MPI being the only C++ API :) ).
Matthieu
2012/10/31 Daniel Mitchell
Nonblocking collectives are at the top of my list. It would also be nice if the environment class could initialize thread support.
On Oct 28, 2012, at 5:33 AM, Matthias Troyer
wrote: Dear Boost.MPI users,
With the MPI-3 standard having been finalized we want to intensify Boost.MPI development as soon as my crazy teaching schedule this fall gives me time from the middle of next month. With the deprecation of the - little used - official MPI C++ API in MPI-3 Boost.MPI has a chance of becoming a de facto standard C++ API for MPI.
We first want to focus on features from MPI-2 and MPI-3 that are widely used or useful. For this purpose I would like to collect suggestions. Can you please reply and mention the functions of MPI that you would most like to see supported by Boost.MPI.
Best regards
Matthias _______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users
_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users
-- Information System Engineer, Ph.D. Blog: http://matt.eifelle.com LinkedIn: http://www.linkedin.com/in/matthieubrucher Music band: http://liliejay.com/
_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

On Sun, Oct 28, 2012 at 11:33 AM, Matthias Troyer
We first want to focus on features from MPI-2 and MPI-3 that are widely used or useful. For this purpose I would like to collect suggestions. Can you please reply and mention the functions of MPI that you would most like to see supported by Boost.MPI.
I'd like to see "ssend" and friends implemented. I actually have a patch ready for this, but it hasn't received much attention: https://svn.boost.org/trac/boost/ticket/5245 Thanks, Riccardo -- Riccardo Murri Grid Computing Competence Centre, http://www.gc3.uzh.ch/ Winterthurerstrasse 190, CH-8057 Zürich (Switzerland) Tel: +41 44 635 4222 Fax: +41 44 635 6888

As far as I am concerned: Collectives on intercommunicator, multi-threading, cart, IO (this one is tricky, not sure what kind of interfaces...). Also, regarding arrays, I'd like to see some support for the situations when the types (dimensions) are different on both sides of the communications (in the case of a transposition for example). I guess a solution involving boost's multi-array could work ? I am willing to help provided someone take care of the review and commit process. Best regards. On Sun, 2012-10-28 at 11:33 +0100, Matthias Troyer wrote:
Dear Boost.MPI users,
With the MPI-3 standard having been finalized we want to intensify Boost.MPI development as soon as my crazy teaching schedule this fall gives me time from the middle of next month. With the deprecation of the - little used - official MPI C++ API in MPI-3 Boost.MPI has a chance of becoming a de facto standard C++ API for MPI.
We first want to focus on features from MPI-2 and MPI-3 that are widely used or useful. For this purpose I would like to collect suggestions. Can you please reply and mention the functions of MPI that you would most like to see supported by Boost.MPI.
Best regards
Matthias _______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users
-- Alain Miniussi Responsable Tech. Centre de Calcul Haute Performance Obs. de la Côte d'Azur |Mont Gros: +33 4 92 00 30 09 BP 4229 |Sophia : +33 4 83 61 85 44 06304 Nice Cedex 4 |https://crimson.oca.eu
participants (6)
-
Alain O Miniussi
-
Daniel Mitchell
-
Jacques C. Richard
-
Matthias Troyer
-
Matthieu Brucher
-
Riccardo Murri