Hi Bronis, all, Thanks for the feedback - I have to admit, though, I have no idea how to represent this graphically in a useful way. If anyone has suggestions, please feel free to draft something. Also, Rich limited each WG to 2-3 slides, so adding two more full slides wouldn't work. It would have to be sketches on the current slides. The other points, that this presentation needs to be more targeted towards the SC audience and be more technical is certainly right. I have attached a modified version - I will ask Rich for a few more days, so that we can discuss this a bit more (incl. on Monday's TelCon). Thanks, Martin On Oct 27, 2010, at 4:45 AM, Bronis R. de Supinski wrote:
Martin:
Re:
As some/most of you have probably seen, Rich asked for 2-3 slide overview/status update from each WG for his presentation on the MPI-3 efforts during the SC MPI BoF. Attached is a first draft for our activities in the tools WG. Please let me know if have any comments or suggestions. Rich needs this by Friday, so please answer as soon as possible.
At a high level, these slides are rather dry. I have found the BoF presentations to be somewhat boring in the past and these slides won't change that. I think it would be good to have two more slides that graphically illustrate the main thrusts. The first could show how MPIT will work (and, at least to some extent, what it is good for). The second could illustrate the MPIR acquisition interface and why we need to extend it before we incorporate it into the standard.
At a slightly lower level, what these slides are missing is related to what you need to show graphically. You say MPIR is not ideal but you provide nothing that backs up that statement other than "has scalability problems". I think you need to show it has those problems and you need to state that it cannot support dynamic processes (and that one can expect dynamic processes will become important to fault tolerance so they are likely to be used in the future even if they are not today). Similarly, you don't give any real feel for what MPIT is or how one would use it and why. Covering those issues are of much more interest to a general audience than a status report on something they do not know enough about to evaluate.
Bronis
_______________________________________________ Mpi3-tools mailing list [email protected] http://BLOCKEDlists.mpi-forum.org/mailman/listinfo.cgi/mpi3-tools
________________________________________________________________________ Martin Schulz, [email protected], http://people.llnl.gov/schulzm CASC @ Lawrence Livermore National Laboratory, Livermore, USA