I've been running several simulations for different trajectories all of which require restarting after a 48hr job timeout on the HPC. I have 20 trajectories running all of which ran for the full time, but only one of these has a restart file with anything in it. All the trajectories use the same input file (attached). Is there any reason why only one trajectory should have a restart file with any content or any way to solve this?
Change NSTEP so that the job finishes within the allowed time. This is the best way. Changing DIMENS will not have a great effect on the performance (if any at all), and it is cumbersome to stitch together tracectory pieces.
Lennart Nilsson Karolinska Institutet Stockholm, Sweden