Variable Data for Particles about ls1-mardyn HOT 2 OPEN

As discussed with Fabio in our last meeting (10. Feb 2021), in some cases it is needed to store particle-specific information. E.g. I implemented the calculation of the heatflux which uses non-diagonal elements of the virial tensor. Therefore, I had to extend the existing virial vector in the FullMolecule class. Furthermore, I added the particle-individual potential energy to the FullMolecule class (which only works with the legacy cell processor to date).

In my opinion, all of the aforementioned quantities do not need to be transfered via MPI, since they are zeroed and calculated new in each timestep.

Nevertheless, in some cases, the possibility to transfer more/arbitrary particle-specific information via MPI would be handy. I think, Isabel would need such a feature for her implementation of the long range correction for droplets.

from ls1-mardyn.

My approach was very much the same as Simons. I need normal and tangential components of the virial per molecule, so I added VirN and VirT in the FullMolecule class.

They are zeroed in every time step.

I did not encounter any problems when I use mpi.

from ls1-mardyn.