2018/02/08 by Sebastian Eibl, Eibl, Sebastian, Ulrich Rüde +1 · 1 citation
Computer Science · Engineering · #Cellular Automata and Applications #Computational Physics (physics.comp-ph) #Distributed #FOS: Computer and information sciences #FOS: Physical sciences #Modular Robots and Swarm Intelligence #Opportunistic and Delay-Tolerant Networks #Parallel #Performance (cs.PF) #and Cluster Computing (cs.DC)
paper · pdf · doi:10.48550/arxiv.1802.02765
openalex publication_date 2018/02/08 · openalex created_date 2022/10/03 · openalex updated_date 2026/07/28
The simulation of large ensembles of particles is usually parallelized by\npartitioning the domain spatially and using message passing to communicate\nbetween the processes handling neighboring subdomains. The particles are\nrepresented as individual geometric objects and are associated to the\nsubdomains. Handling collisions and migrating particles between subdomains, as\nrequired for proper parallel execution, requires a complex communication\nprotocol. Typically, the parallelization is restricted to handling only\nparticles that are smaller than a subdomain. In many applications, however,\nparticle sizes may vary drastically with some of them being larger than a\nsubdomain. In this article we propose a new communication and synchronization\nalgorithm that can handle the parallelization without size restrictions on the\nparticles. Despite the additional complexity and extended functionality, the\nnew algorithm introduces only minimal overhead. We demonstrate the scalability\nof the previous and the new communication algorithms up to almost two million\nparallel processes and for handling ten billion (1e10) geometrically resolved\nparticles on a state-of-the-art petascale supercomputer. Different scenarios\nare presented to analyze the performance of the new algorithm and to\ndemonstrate its capability to simulate polydisperse scenarios, where large\nindividual particles can extend across several subdomains.\n