Hi everybody, I added my comments to the proposal for lack of feedback per email. Please consider at F2F. Slight correction to item 5: Reads "Likewise, the MPI-stall can probably be fairly well statistically approximated by mapping PC to the progress engine routines and its dependants." Should read "Likewise, the MPI-stall can probably be fairly well statistically approximated by mapping PC to the progress engine routines and its dependants, except memcpy and equivalents, and wire access routines". Best regards. Alexander -----Original Message----- From: Supalov, Alexander Sent: Tuesday, May 12, 2009 3:42 PM To: '[email protected]' Subject: RE: [Mpi3-tools] Meeting Notes from 5/11, next meetings Hi, I missed this meeting due to a conflict with another one, sorry. How can one comment on the proposals offline? Briefly on the MPI state proposal: 1. The implementation impact and the associated runtime overheads seem to be rather high. 2. The syntax of the proposed interface is OpenMPI specific. 3. The proposed interface discovery method may not work on all platforms/linkage models. 4. The comparison to the API tracing is skewed in favor of sampling. If one can sample one representative process, one can just as well trace one representative process at comparable or lower overheads. Message matching is not required in this case anyway. 5. Finally, and most importantly, one may be able to get comparable information thru sampling without changing the MPI implementation. For example, the inside-MPI/outside-MPI data can be deduced from the program counter mapping to the program address space occupied by the MPI library and its dependants. Likewise, the MPI-stall can probably be fairly well statistically approximated by mapping PC to the progress engine routines and its dependants. They will be implementation specific, but the set of pertinent address ranges should be possible to specify in any given case. Any existing sampler should help here. Best regards. Alexander -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Martin Schulz Sent: Monday, May 11, 2009 7:00 PM To: [email protected] Subject: [Mpi3-tools] Meeting Notes from 5/11, next meetings Hi all, The meetings notes from today are posted on the Wiki - please let me know if I missed anything, in particular from the first 15min. Here are the dates for the next TelCons: - TelCon: Monday, May 18th, 2009 - Canceled - TelCon: Monday, June 1st, 2009 - Pavan is leading a discussion on PMI, notes will be posted on Wiki - Face to Face: Tuesday, June 9, 2009 (as part of the MPI Forum Meeting in Menlo Park, CA) - Continue discussion on MPI State Interface, focus on implementation impact - Continue discussion on inclusion of PMI - First proposal on Handle Information Interface - TelCon: Monday, June 15th, 2009 - Agenda: tbd (we will decide on this on 6/9) Also, I saw that Rich already put up an agenda, which lists the tools WG for 6/9 from 1-3pm PT. I am not sure if we can change this at this point, but I will try. Martin _______________________________________________________________________ Martin Schulz, [email protected], http://people.llnl.gov/schulz6 CASC @ Lawrence Livermore National Laboratory, Livermore, USA _______________________________________________ Mpi3-tools mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-tools --------------------------------------------------------------------- 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.