Hi all, Sorry for the late notice - our next MPI-3 Tools TelCon is scheduled for next Monday 2/21 at the usual time 8am PST, 11m EST, 5pm MEZ. I know that this is a holiday for some of you (including us at LLNL), but moving it would probably cause more problems than it would help and I didn't want to cancel since we only have a few weeks before we have to have the MPIT document ready for a first reading. If you can't make it, please feel free to send comments to the email list, for the rest Jeff has setup a Webex as usual: https://cisco.webex.com/ciscosales/j.php?ED=160088522&UID=0&PW=NMzcyMGEyNjNi... Password: tools The information is also on the Wiki. The main point on the agenda is to go over the feedback on the MPIT draft from the last forum meeting and to discuss how to proceed. The open issues are (as far as I can see): - Naming: people wanted us to move into the MPI namespace - Which prefix should we use? - How do we distinguish the semantics? - Types: people didn't like a separate type system - Can we use the MPI type systems? - Can we address this otherwise? - Threading: what are the implications on threading? - Do we need to specify thread support? - What is the impact on sessions? - What are the semantics of MPIT_Init if called before MPI_Init? - Do all processes have to call it and what does this mean? - How do we handle the case of an implementation that uses fork in MPI_Init to spawn rank processes? - Smaller issues - Inclusion of BOOLEAN in the type system and implications for C - How to handle variables that are strings? - Connection between timer variables and values from MPI_Wtime - Overlap/Distinction of MPI and MPIT return codes - How do we deal with 30 character limit on names? - Clarification on NULL buffer passed to routines returning strings If I missed anything, please add it to the list - also, any further feedback on the text would be great! The draft is on the tools Wiki and on ticket #266. Thanks! Martin ________________________________________________________________________ Martin Schulz, [email protected], http://people.llnl.gov/schulzm CASC @ Lawrence Livermore National Laboratory, Livermore, USA