Fast Parallel Algorithms for Short-Range Molecular Dynamics

We have found that it, is only the current generation of massively parallel MIMD machines with hundreds to thousands of processors that have the computational power to be competitive with the fastest, vector machines for Ml) calculations

Steve Plimpton


Scholarcy highlights

  • Classical rnolecular dynamics is a commonly used computational tool for simulating the properties of liquids, solids, and molecules
  • The algorithms are suitable for molecular dynamics models which can be difficult to parallelize efficiently -- those with short-range forces where the neighbors of each atom change rapidly
  • They can be implemented on any distributed-memory parallel machine which allows for message-passing of data between independently executing processors
  • We have recently developed a new parallel algorithm which we present here in the context of Ml) simulations for the first time. lt offers the advantages of both simplicity and speed for small to medium-sized problems
  • In this paper we present the culmination of our efforts: several algorithms we have found, through implementing and testing a variety of ideas on different parallel machines, to be the fastest methods for short-range molecular dynamics across a wide range of problem sizes
  • We have developed methods for doing this in organic MI) simulations where the connectivity of the bond groups is static . tlowever, we know of no simple way to use the force--decomposition idea for the more general case of dynamically changing connectivities, such as for silicon three--body potentials
  • It suffers more from load-imbalance and is more difficult to implement efficiently

Need more features? Save interactive summary cards to your Scholarcy Library.