Abstract
We present a new algorithm for automatic parallel load balancing in classical molecular dynamics. It assumes a spatial domain decomposition of particles into Voronoi cells. It is a gradient method which attempts to minimize a cost function by displacing Voronoi sites associated with each processor/sub-domain along steepest descent directions. Excellent load balance has been obtained for quasi-2D and 3D practical applications, with up to 440·106 particles on 65,536 MPI tasks.
| Original language | English |
|---|---|
| Pages (from-to) | 2608-2615 |
| Number of pages | 8 |
| Journal | Computer Physics Communications |
| Volume | 183 |
| Issue number | 12 |
| DOIs | |
| State | Published - Dec 2012 |
| Externally published | Yes |
Funding
This work was performed under the auspices of the US Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. We also acknowledge the support of the Lawrence Livermore National Laboratory Directed Research and Development Program.
Keywords
- Computational geometry
- Molecular dynamics
- Parallel load balancing