[gmx-users] restarting the crashed run

Justin Lemkul jalemkul at vt.edu
Sat Sep 21 22:13:43 CEST 2013

On 9/21/13 3:46 PM, Nidhi Katyal wrote:
> Dear all
> I would like to know the difference between restarting our crashed runs by
> 1) first generating next.tpr using tpbconv -extend option
>      then running grompp with this *.tpr file

Why would you run grompp?  If you're using it as a source of coordinates, you're 
going to be dealing with the initial state, not the last state of the previous 
simulation, so that's garbage.  If you're restarting a crash, then presumably 
there is no need at all to invoke tpbconv or grompp.

>      and finally running mdrun but with no cpi option

Makes no sense.  You're basically obliterating the previous simulation.

> 2) same as 1 but with -cpi option

Still no need for grompp, but if providing -cpi to mdrun, you're resuming from 
the correct state.

> 3) using only mdrun command with cpi option and with previous *.tpr
>      (ie not creating new tpr by tpbconv option)

This is the correct way to proceed.  The run will pick up from the state stored 
in the .cpt file and proceed with the number of steps originally specified in 
the .tpr file.

> 4) using procedure 3 but with no state.cpt file

The run should start over.

> Secondly, if state.cpt contains all the information to continue the
> simulation then why the simulation should continue at all without providing
> these files as in procedure 1 and 4

Without a .cpt file, the run starts over from the beginning.



Justin A. Lemkul, Ph.D.
Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 601
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441


More information about the gromacs.org_gmx-users mailing list