espressomd
WIP Collision detection rework
will fix #3296
Description of changes:
- reworking collision detection to not use different modes anymore, which will make it more flexible to use.
I think the whole construction would be much easier if there where a set of parameter struct for every type pair, much like for the non-bonded interactions. Also the cutoff needs to be properly calculated, and respected in the global cutoff calculation instead of throwing an error if the cutoff requirement is not fulfilled. The current approach is needlessly complicated, and the implementation with the vectors has considerable performance bugs.
In my opinion this functionality can not added in the core at the moment without increasing the technical debt. This is what triggered all the core refactoring that I eventually failed to deliver in time. I still think that the (MD) core code needs to be brought into better shape before this can be implemented, and I think this requires some strategic planning on what actually the requirements are and how methods can be hosted by the core without introducing massive interdependencies. It is just one of the things where change can not be introduced without making things worse because of the bad state the code is in.
Planning for collision detection and bond breakage and the core refactoring which is a pre-requisite should be discussed at an Espresso meeting.
Planning for collision detection and bond breakage and the core refactoring which is a pre-requisite should be discussed at an Espresso meeting.
that might be true... I just wanted to have an update because we have a few PRs that seem to be stuck or not being worked on anymore
