[Mpi-forum] Request introspection

[Jeff Squyres]

Good feedback -- thanks!

/me tosses this idea to /dev/null


On Aug 2, 2011, at 1:01 PM, Martin Schulz wrote:

> I tend to agree with Bill - sure, tools could use such information, but,
> as Bill said, in most cases it is not sufficient information for tools
> anyway and they need to track additional request information. Also,
> most tools already have some kind of request tracking (since this
> is needed for many use cases) and adding a field for the request
> type to such a tracking mechanism is easy.
> 
> Martin
> 
> 
> On Aug 2, 2011, at 8:33 AM, William Gropp wrote:
> 
>> There are already requests for I/O operations on files (just not  
>> collective nonblocking file I/O).
>> 
>> For a tool, this isn't detailed enough - given a request at random,  
>> even for point-to-point, whether it is a send or receive, and whether  
>> it is persistent or not, is important.
>> 
>> However, requests are also on the critical path for performance.  I'm  
>> opposed to anything that adds yet more overhead; there are also  
>> alternatives for tools to use (though they require much more work on  
>> the part of the tool developer).
>> 
>> Bill
>> 
>> On Aug 1, 2011, at 6:50 PM, Jeff Squyres wrote:
>> 
>>> MPI now has 3 different types of requests that I can think of, with  
>>> pending proposals for (at least) 2 more:
>>> 
>>> - point-to-point
>>> - collective
>>> - generalized
>>> - timers (proposed)
>>> - file (proposed)
>>> 
>>> Would it be worthwhile to have a simple introspection routine that  
>>> can return an enum telling you the type of a request handle?  E.g.,  
>>> MPI_REQUEST_GET_TYPE that has an OUT parameter that will be set to  
>>> MPI_REQUEST_PT2PT, MPI_REQUEST_COLLECTIVE, MPI_REQUEST_GENERALIZED,  
>>> etc.
>>> 
>>> I can generally see how this might be useful to applications, but I  
>>> can't think of a "must have" kind of use case.
>>> 
>>> Does anyone else think that this would be valuable / useful?
>>> 
>>> -- 
>>> Jeff Squyres
>>> jsquyres at cisco.com
>>> For corporate legal information go to:
>>> http://www.cisco.com/web/about/doing_business/legal/cri/
>>> 
>>> 
>>> _______________________________________________
>>> mpi-forum mailing list
>>> mpi-forum at lists.mpi-forum.org
>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
>> 
>> 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
>> 
>> 
>> 
>> _______________________________________________
>> mpi-forum mailing list
>> mpi-forum at lists.mpi-forum.org
>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
> 
> ________________________________________________________________________
> Martin Schulz, schulzm at llnl.gov, http://people.llnl.gov/schulzm
> CASC @ Lawrence Livermore National Laboratory, Livermore, USA
> 
> 
> 
> _______________________________________________
> mpi-forum mailing list
> mpi-forum at lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum


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