[Mpi-forum] Reserved MPI_ prefix & namespace in C and Fortran

William Gropp wgropp at illinois.edu
Sun Aug 26 10:49:45 CDT 2012


On 2, I'm ok but I do want to emphasize that the original *intent* of the MPI Forum was to reserve the MPI and PMPI prefix for future versions of the standard - thus *any* limitation on what sort of entity that applies to violates the original *intent*.  Yes, the text failed to make that clear, but there is no change or extension intended here.

This should also be a cautionary tale about attempting to enumerate everything, particularly in a way that will apply in the future.  

Bill

William Gropp
Director, Parallel Computing Institute
Deputy Director for Research
Institute for Advanced Computing Applications and Technologies
Paul and Cynthia Saylor Professor of Computer Science
University of Illinois Urbana-Champaign



On Aug 26, 2012, at 10:19 AM, Rolf Rabenseifner wrote:

> Bill, you ask for two things (here the Fortran text, but C is the same):
> 
>> Programs must not declare 
>> variables, parameters, or functions with names 
>> beginning with the prefix MPI_.
>> To avoid conflicting with the profiling interface, programs 
>> should also avoid functions with the prefix PMPI_. 
> 
> The first sentence, we substituted already by
>> Programs must not declare
>  names, e.g., for variables, subroutines, functions,
>  parameters, derived types, abstract interfaces, or modules,
>> beginning with the prefix MPI_.
> 
> You ask in the second sentence for 
> 1. Substitute "should" by "must" in the last line
> 2. Substitute "functions" by "name".
> 
> About 1.: Yes, agreed. Always must and should are used as
> synonyms in the scope of standards. To use the same
> words in both sentences is clarifying that there is
> no relevant difference in the meaning.
> 
> About 2.: I do not see that it is needed,
> i.e., that there is an inconsistency,
> and the MPI-1 Forum decided to make a clear difference
> between MPI 
> - where all names must be restricted,
> - as we now do,
> and PMPI
> - where conflicts are only for routines because we
>   do not have any PMPI name for other areas.
> 
> Therefore for me, the 2. topic would be out of the scope 
> of this erratum.
> And I want to keep it as an erratum, because it should
> be done before we vote on MPI-3.0 and therefore automatically
> included in MPI-3.0.
> 
> Okay?
> 
> Rolf
> 
> 
> ----- Original Message -----
>> From: "William Gropp" <wgropp at illinois.edu>
>> To: "Main MPI Forum mailing list" <mpi-forum at lists.mpi-forum.org>
>> Sent: Sunday, August 26, 2012 4:28:22 PM
>> Subject: Re: [Mpi-forum] Reserved MPI_ prefix & namespace in C and Fortran
>> I would use "must" instead of "should", esp in reference to the PMPI
>> interface. I'd also reserve all PMPI and MPI names, not just the
>> function/subroutines for PMPI. The intent was always that users avoid
>> these two prefixes for any use; anyone who used them was courting
>> trouble.
>> 
>> Bill
>> 
>> William Gropp
>> Director, Parallel Computing Institute
>> Deputy Director for Research
>> Institute for Advanced Computing Applications and Technologies
>> Paul and Cynthia Saylor Professor of Computer Science
>> University of Illinois Urbana-Champaign
>> 
>> 
>> 
>> On Aug 26, 2012, at 9:03 AM, Rolf Rabenseifner wrote:
>> 
>>> I updated the ticket 343 with the now hopefully final text
>>> for the terms chapter
>>> (after Bronis has moved the tools aspect to MPI-next),
>>> see
>>> https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/343
>>> and the attached files.
>>> 
>>> I hope that now all can agree with this correction of the
>>> existing inconsisteny as detected in the public review
>>> and discussed in the previous emails.
>>> 
>>> --------------------------------
>>> Latest summary of the necessary MPI-2.2 erratum
>>> in Sections 2.6.2 - 2.6.4:
>>> 
>>> C:
>>>> "Programs must not declare
>>>     names (identifiers), e.g., for variables, functions,
>>>     constants, types, or macros,
>>>> beginning with the prefix MPI_."
>>> 
>>> Fortran:
>>>> "Programs must not declare
>>>     names, e.g., for variables, subroutines, functions,
>>>     parameters, derived types, abstract interfaces, or modules,
>>>> beginning with the prefix MPI_.
>>> 
>>> and for Fortran also "subroutines and functions" instead of
>>> only "functions" in the sentence about PMPI_.
>>> 
>>> C++ (not visible in MPI-3.0 due to removal of C++):
>>>> "Programs must not declare
>>>     names (identifiers), e.g., for variables, functions,
>>>     constants, types, or macros,
>>>> in the namespace MPI."
>>> 
>>> And change-log entry
>>> 
>>>  Sections 2.6.2 and 2.6.3 on pages 18 and 19, and
>>>  MPI-2.2, Section 2.6.2 on page 17, lines 41-42,
>>>  Section 2.6.3 on page 18, lines 15-16, and
>>>  Section 2.6.4 on page 18, lines 40-41.
>>>  This is an MPI-2 erratum: The scope for the reserved prefix MPI_
>>>  and the C++ namespace MPI was extended to any name.
>>> -------------------------------
>>> 
>>> Best regards
>>> Rolf
>>> 
>>> 
>>> ----- Original Message -----
>>>> On Sun, 26 Aug 2012, Bronis R. de Supinski wrote:
>>>> 
>>>> I think we should defer the issue to MPI Next.
>>>> 
>>>> On Sun, 26 Aug 2012, Rolf Rabenseifner wrote:
>>>> 
>>>>> Hello Bronis, Kathryn, Dave,
>>>>> 
>>>>> Do you agree with Martin's opinion and his suggested addition
>>>>> for the tools chapter?
>>> ...
>>> 
>>> --
>>> 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)
>>> _______________________________________________
>>> mpi-forum mailing list
>>> mpi-forum at lists.mpi-forum.org
>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
>> 
>> 
>> _______________________________________________
>> mpi-forum mailing list
>> mpi-forum at lists.mpi-forum.org
>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
> 
> -- 
> 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)
> _______________________________________________
> mpi-forum mailing list
> mpi-forum at lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum





More information about the mpi-forum mailing list