Concatenating trajectories in AMDAT

[drayer587]

In my current computing environment, I have a ceiling of 96 hours of compute time on a single execution of LAMMPS. As a result, if I want to analyze a trajectory that is longer, I need to manually stitch the trajectories together (by running two identical runs, and eliminating the one repeated frame: last of the first run, the first of the second run). I suppose I could instead have LAMMPS not write the first frame of the second block, but also, best practice is not to have a single file become too large anyway, and it's good to have that frame if you want to analyze each half of the run independently anyway.

Is this something that is within scope for AMDAT? This doesn't seem like it would be too cumbersome to implement, but I don't trust my impression. I experienced first hand that this is not too bad to do by hand, but is somewhat painful to do in a script with good performance. I'm happy to share the bash script I use to do this now (it's not pretty but it worked for me).

[drayer587]

I suppose an alternative option is to make the time-step in AMDAT be some time-step increment in LAMMPS (i.e. a time-step of 2 in LAMMPS, but make a trajectory for AMDAT which has a time-step of, say, 4 or 10). This seems like it should work without a problem but it's not clear to me if this is worthwhile or good (better?) practice that this proposal here.

[pkawak]

To be clear, is the proposal that AMDAT read trajectory frames from two files for the same analysis? If so, this is a great idea and should be implemented. In general, we need to improve AMDAT's versatility when it comes to read trajectory frames. In addition to the above, we should also be able to read only parts of the trajectory file and automatically identify the time scheme as well.