I agree with Dick's inital skepticism about the usefulness in general of such user survey's (that always more or less generate the feedback the authors want in the first palce...), and the below is in my opinion definitely not going to generate any useful input (it's too vague, and even more dangerously, speaking about things where the ramifications and pro's and cons are not even clear to the "experts" - what guidance do the Forum possibly expect from the causual filler-outer here???) I think the initial, more modest try by Jeff could, perhaps, (with emphasis on "could" and "perhaps") generate a useful feedback. Technical discussions as below will in my opinion definitely not. And: it might make the Forum look foolish, and/or raise false expectations about what's going to be in MPI-3 on behalf of some users. We should avoid that So I'd thread very carefully and very modestly (and very clearly) here... Jesper On Tue, Nov 17, 2009 at 10:57:57PM +0000, Supalov, Alexander wrote:
Thanks. Good idea. A first take on the subsetting follows:
The MPI-3 standard is going to provide several substantial extensions to the existing MPI standard. Some of them may or may not be needed for all MPI-3 applications. At the same time, runtime support for the complete set of the MPI-3 features may incur additional implementation complexity and hence probably performance and/or memory space penalty on the rest of the MPI implementation (think dynamic processes in the case of MPI-2, for one). Subsetting, i.e., a facility intended to determine a working subset of the MPI features needed for a particular application, was proposed as one way of managing the increasing complexity of the MPI standard on a per application basis. If such a facility were included into the MPI-3 standard, how useful might that be to you?
Very valuable Probably valuable Probably useless Useless Harmful
________________________________ From: [email protected] [mailto:[email protected]] On Behalf Of Richard Treumann Sent: Tuesday, November 17, 2009 6:48 PM To: Main MPI Forum mailing list Subject: Re: [Mpi-forum] MPI user survey
I will leave it to Alexander to for a question about subsetting if he wishes to. Here is one about assertions.
The MPI standard provides certain semantic guarantees that may not be required by a particular application. It also provides functions that many applications never use. If MPI-3 provided an "assertions" interface that would let an application identify specific functionality it does not depend on and an MPI library could improve performance or reduce memory usage by disabling that specific functionality, how valuable might that interface be?
Very valuable Probably valuable Probably useless useless harmful
Dick Treumann - MPI Team IBM Systems & Technology Group Dept X2ZA / MS P963 -- 2455 South Road -- Poughkeepsie, NY 12601 Tele (845) 433-7846 Fax (845) 433-8363
[email protected] wrote on 11/17/2009 11:40:03 AM:
[image removed]
Re: [Mpi-forum] MPI user survey
Jeff Squyres
to:
Main MPI Forum mailing list
11/17/2009 11:41 AM
Sent by:
Please respond to Main MPI Forum mailing list
On Nov 17, 2009, at 8:16 AM, Supalov, Alexander wrote:
No objections - let's go for 2 separate questions, e.g.:
"Do you want to achieve higher performance by disabling certain MPI-3 features in your program?
I do not think that this is a well-formed question.
It directly states something that has not been well defined (or even accepted) by the Forum. Specifically: it is *by no means guaranteed* that you will get higher performance by disabling certain MPI-3 features. Heck, we don't even know precisely what MPI-3 features will exist!
"If so, do you prefer subsetting or assertions?"
The others questions contain at least a hint of context so that respondents have a clue as to how to answer. This one explains neither "subsetting" nor "assertions". I suspect that 95% of respondents will not know how to answer.
-- Jeff Squyres [email protected]
_______________________________________________ mpi-forum mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
Intel GmbH Dornacher Strasse 1 85622 Feldkirchen/Muenchen Germany Sitz der Gesellschaft: Feldkirchen bei Muenchen Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer Registergericht: Muenchen HRB 47456 Ust.-IdNr. VAT Registration No.: DE129385895 Citibank Frankfurt (BLZ 502 109 00) 600119052
This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
_______________________________________________ mpi-forum mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum