Re: AMBER: Probelm with running Amber 9 parallelly

From: David A. Case <case.scripps.edu>
Date: Fri, 20 Oct 2006 08:24:16 -0700

On Fri, Oct 20, 2006, Rafi Ahmad wrote:
>
> when using a parallel version of sander (sander.MPI ) on ia64 it crashes
> in the middle of the run.
>
> on i386 it runs to the end, but hangs for several minutes before it
> crashes with an mpi-write eror.

It's best if we can narrow down the possible sources of problems. To that
end, please be sure that your installation passes the test suite. If so, try
one of the benchmarks (e.g. jac). These run longer jobs, and may expose the
error. If these work, make the number of steps in the benchmark longer, and
see if that works as well.

The idea is to separate possible problems that are specific to your simulation
from those that are generic to the compiler/OS/cpu you are using.

[Also, see if you get errors at exactly the same place when re-running these
jobs. The description given above sounds like hardware problems to me.]

...good luck...dac

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Oct 22 2006 - 06:07:25 PDT
Custom Search