[Mpi3-tools] MPI-3 Tools TelCon 10/3 - PMPI Wrappers for F2008
schulz6 at llnl.gov
Tue Sep 27 19:33:55 CDT 2011
With the MPI_T ticket progressing, it has been a while since we had an
MPI-3 tools TelCon. However, during the MPI forum last week a topic came
up that may be an issue for tools and Marc-Andre, who was at the forum as
well, and I thought that this would be worth discussing in this group. I
also cc-ed a few tool developers that we know work on MPI wrappers - if I
missed anyone, please feel free to forward this.
The MPI forum is currently discussing the inclusion of new bindings for
Fortran 2008, as well as an update of the existing F90 bindings. This has
an impact on the linker symbols that need to be intercepted by PMPI tools.
In particular, tools will have to intercept up to 6 symbols for each MPI
routine in order to get full coverage. Further, it will not be guaranteed
that all "P" routines are always present in each installation, since this
depends on a combination of compiler and linker options.
The complete ticket is available at:
The most relevant parts are in Chapter 14 and 16.
Marc-Andre and I believe that we have reached a workable compromise (it
won't make it easier for tools, but it will be at least feasible), but it
would be good to have further reviews of this issue as well as a
discussion on this topic to make sure we haven't missed anything. Change
suggestions for this ticket need to be made two weeks in advance of the
next forum, which is already end of next week. Hence, unfortunately there
isn't much time.
We will cover this topic at the next MPI-3 tools TelCon, which is
scheduled for 10/3 at 8am PDT, 11am EDT, 5pm MSZ. Anybody is welcome to
join. Here is the link (it is also on the MPI-3 tools Wiki):
Monday, October 3, 2011, 11am US Eastern time
* Webex meeting:
* Webex password: tools
* Agenda: PMPI Wrappers for Fortran 2008
I apologize to the German participants since I know this is a holiday, but
we need to get the feedback in time for the next forum meeting. If you
can't make it, but have comments, please send email to this list.
Martin Schulz, schulzm at llnl.gov, http://people.llnl.gov/schulzm
CASC @ Lawrence Livermore National Laboratory, Livermore, USA
More information about the mpiwg-tools