Florian Weik
Florian Weik
Currently thee parallel core (mpi synchronous) and frontend interface (callbacks) are fused together. This should be separated for clarity and for making sure only the latter is called from the...
Currently the whole particle fetch cache is dropped when any particle is touched. This is not needed, because in most of the cases it is know which particle is modified,...
If the reference particle of a virtual site is removed, and virtual sites relative is enabled, this leaves the system in an invalid state, and it will crash on next...
The front end particle access is sped up by a cache, whose size is currently hard-coded to 100MB, which corresponds to ~100k-1M particles, depending on what features are activated. As...
There is a mechanism for tracking the number of particles per type in the head node particle data interface. This does so by tracking type changes thru the front end,...
In and around `on_observable_calc` there are potentially some unneeded calls to `cells_update_ghosts`. It should be checked if all calls are needed.
The cell systems are actually sorted by `p_old`, which should also be used to find the current cell for the particle. This always works, and makes the cell system independent...
The Thole dampening should be part of the coulomb short range calculation, and its parameter stored per particle (looking at [LAMMPS](https://lammps.sandia.gov/doc/pair_thole.html), this should be possible to store the a and...
The inertialess traces consist of two parts, an Euler integrator for the position and the LB coupling. Neither one is a virtual site mechanism. The two parts should be separated,...
Depends on #3036 for non-bonded, #3019 for bonded. Currently the functions setting the interaction parameters also select on which interaction struct they operate, e.g. they take the types and the...