[Mpi-forum] Let's stop using the term "MPI Next"

Jeff Squyres jsquyres at cisco.com
Wed Oct 3 10:06:29 CDT 2012

Just to clarify...

There was a discussion in Vienna of what the value of <next> should be in the term "MPI <next>", and what kind of content should go into that version of MPI.

After some discussion, the room pretty much agreed on general definitions of "MPI-3.1" and "MPI-4.0":

*** MPI 3.1: should be small/errata items, analogous to MPI 2.1 and 2.2.  

  ==> With that premise, we should just gather errata over time (e.g., as implementors start/continue implementing, and users start using).  Once we get enough errata, and/or enough time goes by (e.g., 1 year? 2 years?), we just incorporate those errata in the document and pass it as 3.1.

  ==> One non-errata topic that might be suitable for 3.1 is the scalable vector collectives, because they've actually been done for a while, are small / self-contained, and simply ran out of time to get into 3.0.

*** MPI 4.0: should be Big New Things (in the same spirit as MPI-3.0 is Big New Things).

  ==> A brainstorm list of possible MPI 4 topics can be found http://lists.mpi-forum.org/mpi-forum/2012/09/2208.php.

Jeff Squyres
jsquyres at cisco.com
For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

More information about the mpi-forum mailing list