Hi All,
I am curious if anyone on the list has tried to set up a parallel replica dynamics (accelerated MD method developed by Art Voter of LLNL and coworkers) run using LAMMPS as a base. It is essentially running a group of nearly identical simulations in parallel, with periodic communications between them.
One of the things that is needed for PRD to work is a sort of ‘wrapper’ which will allow for the creation of the replicas and starting the simulations, allow the replicas to communicate if a transition has been detected in one of the replicas, and if so, to stop all the others, then communicate the state of the replica that had the transition to the other replicas(perhaps create some output), and repeat the process.
Add added complexity will come in the fact that I will have to deal with a batch scheduler system too, because I am using an HPC center machine.
Anyone ever try anything like this?
Dave
David E. Farrell
Graduate Student
Mechanical Engineering
Northwestern University
email: d-farrell2@…435…