Re: [AMBER] segmentation faults and randomly stopping simulations

From: Daniel Roe via AMBER <amber.ambermd.org>
Date: Thu, 24 Oct 2024 09:21:01 -0400

Hi,

On Wed, Oct 23, 2024 at 11:46 AM Ryan Woltz <rlwoltz.ucdavis.edu> wrote:
> However, the reason for the instability was I had 4 overlapping atoms that the parm7 validate missed.

Note that the ParmEd 'checkValidity' command only checks that your
parameters/topology look OK - it does *not* check structure. To do
that you should use the 'checkstructure' command of cpptraj. That will
tell you about bad overlaps, distorted bonds, etc.

> I did also test out the double precision and I didn't see any errors for the few steps I made but the speed drops from 90ns/day to 3.5ns/day on a local machine. Exported to a slower HPC with the max speed at 35ns/day I think it would take 100 days to get my 100ns of equilibration needed so this may not be the best use of these resources?

Note you would only have to use the double precision code as long as
it takes to resolve the really bad clashes - in my experience 1 ns is
usually more than enough for this (and indeed most really bad clashes
can be resolved on the order of 10s of ps, as long as you don't have
something really bad like a ring intersection or something).

-Dan

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Oct 24 2024 - 06:30:02 PDT
Custom Search