[gmx-users] Re: Diatomic in MeCN NPT (NH and PR) simulation segfaults after 1 us
jalemkul at vt.edu
Tue Nov 6 15:25:29 CET 2012
On 11/6/12 9:01 AM, benjfitz wrote:
> Thanks for the suggestion, I should have mentioned Tcoupl and Pcoupl in my
> previous post. I changed them from Nose-Hoover/Parrinello-Rahman to
> V-rescale/Berendsen and one permutation mixing the two, but it crashed all
> the same.
> The nodes have at least 8 GB of ram and free -m shows that I'm not running
> out, even when it gets close to the magical 1.07 us. I'm currently running a
> 2 us NVT simulation with CS2 and PME. Do you have any other thoughts as to
> options I can change to narrow this down? Perhaps a third coulombtype?
> Thanks in advance.
Rather than run the entire job in one sitting, I would suggest running in
smaller chunks of time (e.g. 500 ns or so) and restarting from checkpoints as
necessary to reach the desired 2 us. It would be interesting to see if there is
some accumulated error that arises from a very long simulation that is mitigated
or fixed by using smaller increments of time.
Justin A. Lemkul, Ph.D.
Department of Biochemistry
jalemkul[at]vt.edu | (540) 231-9080
More information about the gromacs.org_gmx-users