[Mpi-forum] Draft standard
rabenseifner at hlrs.de
Mon Nov 8 14:47:58 CST 2010
I'm clearly against producing a draft that is already known to
be incorrect, because new text (here a sentence about status.MPI_ERROR)
in the new nonblocking collectives chapter is inconsistent with the
existing MPI-2.2 standard.
We discussed this problem on the ticket, and Bill and Bronis could
agree that we can correct the text.
It is a major job of the chapter's committee to make such corrections.
The sense of such a MPI-3.0 draft is not to show up at SC10,
the sense is to have a clear basis for all MPI library developers
to implement nonblocking collectives in their MPI libraries.
This basis should be as correct as possible.
Please take the time to review the technical problem instead of
being only formal.
If you want to be formal, then please feel free to make a
formal on-line vote on this small change.
I believe that such a vote is needed only for the final MPI-3.0,
not for a corrected draft-3.0.
----- Original Message -----
> From: "Richard L. Graham" <rlgraham at ornl.gov>
> To: "Main MPI Forum mailing list" <mpi-forum at lists.mpi-forum.org>
> Sent: Monday, November 8, 2010 9:22:24 PM
> Subject: [Mpi-forum] Draft standard
> A lot of people have put in quite a bit of time over the last several
> weeks to get the first draft standard ready for release, so thanks all
> - this is VERY much appreciated. The document is not perfect, but that
> will always be the case. An issue with the lack of definition of the
> return status was discovered in the nonblocking collectives section.
> Since this is a matter of specification, and not a simple wording
> this, a change for this will need to come through the full forum.
> Trunk r689 is what I would like to put out as the draft standard.
> Please, by Wed., 11/10/2010, get back to list with any show stoppers
> you may have spotted. I am not posting the standard to the list to
> minimize possible confusion, as the document should not go out for
> public distribution until it is actually ready for public consumption.
> Thanks again to all,
> mpi-forum mailing list
> mpi-forum at lists.mpi-forum.org
Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner at hlrs.de
High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530
University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832
Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner
Nobelstr. 19, D-70550 Stuttgart, Germany . (Office: Allmandring 30)
More information about the mpi-forum