[Developers] MPITimer problem

Steve White swhite at aei.mpg.de
Thu Aug 12 07:23:58 CDT 2004


I have placed the MPITimer thorn in AlphaThorns.  It works, but a bug at
Shutdown has surfaced.

Here is the problem:

Cactus first shuts down all the non-Cactus stuff (such as MPI), then calls
a routine that shuts down all the Cactus stuff, and prints the final
timing talleys.  In doing so it calls the timer Stop function, which is
meant to get a final time measurement.

In the MPITimer thorn, the Stop function calls the MPI timer...
but...MPI has been shut down already.  Could be a crash.

Now, this is only a problem in certain situations.  On my laptop running
MPICH, I see no problem.  On Peyote, there is a problem.

Any ideas what to do about this?  

It's a bit of a conflict between the Cactus shutdown procedure and 
running MPI code to the end...

If I could somehow trigger this Stop in the non-Cactus Shutdown...

------------------------------------------------------------------------
Steve White : Programmer
Max-Planck-Institut für Gravitationsphysik      Albert-Einstein-Institut
Am Mühlenberg 1, D-14476 Golm, Germany                  +49-331-567-7329





More information about the Developers mailing list