SC²S Colloquium - August 03, 2016

From Sccswiki
Jump to navigation Jump to search
Date: August 03, 2016
Room: 02.07.023
Time: 3:00 pm, s.t.

Michael Obersteiner: Parallel Implementation of the Fast Multipole Method

Over the past decades the sizes of Molecular Dynamics simulations have grown rapidly. Hence, fast summation techniques such as the Fast Multipole Method are required to reduce the computational complexity of the electrostatic force calculation from O(n²) to O(n). Highly scalable implementations of the Fast Multipole Method are therefore needed to enable efficient simulations in parallel computing. In this work MPI is used to optimize the overlap of communication to computation during the Fast Multipole algorithm as well as optimizing import loads and waiting times through zonal methods.