mpi + shmem issues

[Tim Robbins]

On Sun, May 25, 2003 at 05:17:30PM -0400, Anthony Schneider wrote:

> so, does this mean that mpich somehow exhausted all shmem locks?
> after running the program only 10 times, i see this as infeasible,
> considering
> 	a) mpich (presumably in MPI_Init()) would only want 1 or
> 	   2 locks on init
> and
> 	b) any shared memory locks mpich grabs should be freed
> 	   upon process completion (whether clean or not) by the
> 	   operating system, no?

No, semaphores stay around until they are removed. Perhaps there's an MPI_Xyz
function you should be calling before your program exits. In any case, you can
use "ipcs -s" to list semaphores, and "ipcrm -s semid" to remove semaphores.

It looks like the program is leaking semaphores -- notice how your program
fails when you run it for the 11th time, and the default maximum number of
semaphores in the GENERIC kernel is 10 (sysctl kern.ipc.semmni).


Tim