Dear all,
I have been working on the nudged elastic band method and I have implemented several new features (a different parallel spring force which is similar to the first string method and which is more efficient at keeping the replicas equispaced, free end images, a perpendicular contribution of the spring force which in some cases stabilises the method...).
I plan to clean up my code and to submit it for making it available to everybody. However I heard somewhere (but I am not sure that it's true) that somebody is currently working on improving the parallelization of the NEB (in order to allow the use of multiple processors per replica).
If this is true, I guess that it would be best that I coordinate with this person.
Thank you to let me know,
Emile Maras
Dear all,
I have been working on the nudged elastic band method and I have implemented several new features (a different parallel spring force which is similar to the first string method and which is more efficient at keeping the replicas equispaced, free end images, a perpendicular contribution of the spring force which in some cases stabilises the method...).
I plan to clean up my code and to submit it for making it available to everybody. However I heard somewhere (but I am not sure that it's true) that somebody is currently working on improving the parallelization of the NEB (in order to allow the use of multiple processors per replica).
If this is true, I guess that it would be best that I coordinate with this person.
i cannot help you much with this specific question, but it may be a
good idea to "advertise" your plans and some specifics on the LAMMPS
github project as an "issue". https://github.com/lammps/lammps/issues
there used to be a page on the LAMMPS homepage about features that
people have asked for and features that were in progress, but that
wasn't kept up-to-date, since it required manual and explicit action.
we hope that the github issue tracker will make this a bit more
dynamic, since feature suggestions and plans for features can be added
by anybody.
axel.
There hasn’t been any recent work on allowing mutliple MPI ranks per replica.
That would be orthogonal to everything you mention I think. So please
submit your NEB enhancements - they sound useful.
Steve