MPI 2.1:
********
As we agreed to in Reno, I am soliciting for people who are willing
to be
responsible for a careful reading of a chapter the MPI 2.0
standard. The
intent is to have a deliberate effort to fix obvious errors, and
find items
that need clarification so that we can finish the 2.1 process by
mid year
2008. What we need to have as an outcome is a list of proposed
changes to
the standard that can for the basis for an e-mail discussion, and then
eventually a vote. All are encouraged to review the standard, and
participate in this process. It at all possible, it would be nice
to have a
first draft for review by the Jan 14-16 meeting. Note that there
is already
a list of changes, as well as proposed changes on the MPI 2.1 web
site.
Please send the list a note, if you are willing to be responsible for
reading a chapter.
Chapter 2: MPI-2 Terms and Conventions
Chapter 3: Version 1.2 of MPI
Chapter 4: Miscellany
Chapter 5: Process Creation and Management
Chapter 6: One-sided Communications
Chapter 7: Extended Collective Operations
Chapter 8: External Interfaces
Chapter 9: I/O
Chapter 10: Language Bindings - Jeff Squyres
MPI 2.2
*******
I would like to publish an agenda in the next couple of weeks for
the Jan
14-16 meeting. Please send in items that you would like to
discuss. Please
remember that 2.2 is intended to be "small" fixes to the API that
do not
change the nature of the standard itself. Sending in an item also
implies
that you will come prepared for a detailed discussion of the
proposal - why
this change is needed, pros and cons, impact on users, impact on
implementers, impact on other stake-holders. I expect to spend
about one
day on MPI 2.2 items, so we can most likely handle up to four such
items.
It would also be good to discuss these items via e-mail, just like we
discussed the "const" issue that was raised, so all of us can come
prepared
to discuss these items.
MPI 3.0 (?)
***********
Again, I would like to have items for the agenda in the next could
of weeks.
The first item will be a discussion of what the MPI 3.0 (if this is
what it
is called) be, and I will set aside a half day for this
discussion. If you
have opinions on what MPI 3.0 should include, please send these to
the list
- I will start the discussion with a summary of the issues raised,
to give
us a concrete starting point. Please send in items that we can
consider,
based on the out come of the initial discussions. These should
include only
items that people have already some experience with implementation and
usage, or will be working on this in the process of this effort
(speculative
items should not go into the standard).
Administrative items:
*********************
As of now, we don't have sufficient commitment to attend the
meetings to
reserve a block of rooms for the meeting (we will make meeting room
arrangements). I expect this is just because people have not had
time to do
this. If we can get on the order of 30-35 commitments, ORNL will
go ahead
and write a contract with the hotel, so please let me know in the near
future if you plan to attend, and would like to reserve a hotel
room for
this. This is the current list of people who have indicated they will
attend the first two (some only one) meetings (does not include
people local
to Chicago, who will not need rooms):
Will Attend August meeting in Europe Org
*********** ************************
********
Purushotham Bangalore UAB
Gil Bloch/others yes - no weekend
Mellanox
George Bosilca (1st meeting) UT
Ron Brightwell yes SNL
Edgar Gabriel yes UH
Patrick Geoffray
Myricom
David Gingold yes
SiCortex
Rich Graham yes ORNL
Bill Gropp
Illinois
Erez Haba yes
Microsoft
Torsten Hoefler (or others)
Indiana U
Matt Koop
Ohio State
Quincey Koziol
HDF/Illinois
Kannan Narasimhan/ HP
David Solt
Yarek Nieplocha yes PNNL
Jeff Squyres (1st, ? on rest) yes
Cicso
Alexander Supalov yes
Intel
Thanks,
Rich