Efficient Domain Decomposition of Dissipative Particle Dynamics via Choice of Pseudorandom Number Generator

Domain decomposition of dissipative particle dynamics is complicated by the use of random pairwise forces as a component of a momentum-conserving thermostat. The conventional use of a pseudorandom number generator for each processor core leads to the need for an additional communication step to correctly assign random forces to particles in boundary halos. To circumvent this communication, the use of a three-seed pseudorandom number generator is proposed to allow multiple processor cores to evaluate the same forces. This kind of pseudorandom number generator will be applied to the general-purpose mesoscale modelling package DL_MESO to improve its parallel scalability for large processor core counts.