[gmx-users] MPI termination

[Erik Lindahl]

David L. Bostick wrote:
> Hello,
> 
> I send a paralell job with gmx3.1 to 8 processors on a SGI Origin machine
> using mpi.  No matter how many timesteps I give for the simulation it gives
> the error
> 
> MPI: MPI_COMM_WORLD rank 0 has terminated without calling MPI_Finalize()
> MPI: aborting job
> 
> anywhere from 1/2 to 3/4 of the way through the calculation.  What might
> cause this?
> 

Looks like a crash/error where the first node terminated the run by 
calling exit() directly. Check the logfiles for information, or (in 
worst case) try to run it through the parallel debugger. I'm not sure if 
that's trivial, since I've never used parallel SGI's myself!

Cheers,

Erik