I have a serial code that repeatedly calls OpenMPI mpirun on a parallel
code. Each run takes either 10 or 100 seconds, and the whole process
repeated
thousands of times. Each invocation of mpirun is gradually slower
(adding maybe
15-20 seconds per run after about 1000 runs). This is with OpenMPI
1.3.2, under
SGE (using the built-in gridengine integration). Is there any way I
can get mpirun
to give me timing information for various things it's doing, so I can
try to figure out
what's slowing it down?
Noam
