Abstract

One of the most significant drawbacks of the PCISPH algorithm (predictive corrective SPH), which is often used for the simulation of incompressible viscous liquid dynamics, is the complexity of the organization of distributed calculations on a computational cluster. At the same time, the algorithm supports the data parallelism relative to a single device. In this paper, new methods and algorithms for overcoming this drawback are suggested, based on an effective model of distributing the data between independent computing units. The resulting algorithms have been realized and tested on Novosibirsk State University’s cluster.

DOI
10.31144/bncc.cs.2542-1972.2021.n45.p59-68
File
Issue
Pages
59-68