Hello LAMMPS,
There occurs a Segmentation Fault on the latest version when running a simulation containing a ‘compute msd,’ and then using ‘change_box all triclinic’ on the resulting restart file. I’m guessing that the per-atom info stored in the restart file is causing a problem, because when going old school and writing out the data file and manually converting to triclinic, there is no issue.
Thanks,
Jake
Please post a simple script that triggers the segfault?
Thanks,
Steve
Hey Steve,
I’m sorry, was unable to reproduce error with a smaller system. However, here is my restart file in case you want to take a look and simple input script which runs fine without ‘change_box all triclinic’, but throws a seg fault with it. As stated before, previously input was just as simple but included:
group polymer type 1:6
compute polymer_msd polymer msd
Looks like seg fault occurs after attempting PPPM initialization.
restart_after_5_ns
Thanks,
Jake
input2.txt (195 Bytes)
jake, your restart was written with a rather old version of LAMMPS (1 Feb 2014). restarts in general are only guaranteed to work for the same version of LAMMPS (unless there is a bug). otherwise there is no guarantee that restarts can be read correctly. they often can, because changes are usually made in a backward compatible way, but occasionally there are changes after which data would be initialized incorrectly. the way to approach this is: go back to the older version of LAMMPS, convert your restart to a data file and the start from that data file instead. if you want portable restarts, you must use data files. since you are doing significant changes to the system, you don’t lose anything this way anyway.
axel.
Better yet, see if you can reproduce your problem in the current version
of LAMMPS. It may be that some bug you are seeing has already been fixed.
Steve