[gmx-users] segmentation fault, very high KE but no high velocity

[Berk Hess]

>From: Dongsheng Zhang <dong at pampas.chem.purdue.edu>
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>Subject: RE: [gmx-users] segmentation fault, very high KE but no high 
>velocity
>Date: Thu, 20 Jul 2006 08:51:43 -0400
>
>Dear Berk,
>
>Thank you for your reply. The velocities are from trr file. Could you
>please tell me what those values are (for t-dt/2 or t+dt/2 or t)? If
>they are not for time step t, could you change the source code to output
>the value for t to keep consistency with KE? By the way, I had set
>nstfout  = 1, but I can't find the information for force in trr file.
>

The velocities in the trr are at step t-dt/2.
With the leap-frog integrator the velocities are never
determined at time t (read the manual).

>
>segmentation fault usually means that one array is out of bound. I set
>cutoff as 1.4, the table extension as 2.0. My table covers the
>interaction arrange of 5.0. I thought it should large enough. Just now I
>extend my table to 10, I still get segmentation fault at the same time
>step. Therefore I guess there is another reason to cause segmentation
>fault. Could you please give me some suggestions? With your experiences,
>could you please give me some general ideas what are possible causes for
>segmentation fault?

The most probable cause in an instability in your system, which
causes very high forces. If you are lucky you can see it in
the configuration or the forces one integration step before the crash.

Berk.