[gmx-users] atom moved too far

[Justin Lemkul]

On 1/6/20 2:59 PM, Christos Deligkaris wrote:
> Justin, thank you.
>
> I have implemented the pull code but that also exhibits the same error when
> I use 12 cores (failed at about 2ns) and the simulation goes on fine when I
> use 6 cores (now at about 32 ns).
>
> I tried using the v-rescale thermostat (instead of Nose-Hoover) and
> Parinello-Rahman barostat, which failed. I also tried the v-rescale
> thermostat and the Berendsen barostat but that also failed. It seems to me
> that this is not an equilibration issue.
>
> So, to summarize, only if I decrease the time step to 0.001 ps or decrease
> the number of cores seem to allow the calculation to proceed.
>
> In this email list, I read that someone else was trying to use different
> arguments supplied to mdrun (-nt, -ntomp etc) to solve the same problem. Is
> it possible that the problem arises due to my running gmx_mpi on a single
> node? This is the command I use in my submission script:
>
> mpirun --mca btl tcp,sm,self /opt/gromacs-2018.1/bin/gmx_mpi mdrun -ntomp
> \$ntomp -v -deffnm "${inputfile%.tpr}"
>
> If you think that this is not due to a physics issue I can continue doing
> calculations with 6 cores and try to install gromacs 2020 (both gmx and
> gmx_mpi) to see if my problem persists there or not...

If you're running on a single node, there's no need for an external MPI 
library. Perhaps you've got a buggy implementation? Have you tried using 
12 cores via the built-in thread MPI library?

-Justin

-- 
==================================================

Justin A. Lemkul, Ph.D.
Assistant Professor
Office: 301 Fralin Hall
Lab: 303 Engel Hall

Virginia Tech Department of Biochemistry
340 West Campus Dr.
Blacksburg, VA 24061

jalemkul at vt.edu | (540) 231-3129
http://www.thelemkullab.com

==================================================