HIGH PERFORMANCE COMPUTING (HPC)
Particle simulations make it possible to observe dynamic processes in great detail. In order to do that it is often necessary to divide the simulated medium very finely, into very many particles. There are several reasons for that which also depend on the type of simulation method used. For simulating granular media using the DEM method the process to be simulated often already contains a large number of particles.
For particle-based computational fluid dynamics, however, the quality of the results depends, as for most other simulation methods, on the degree of discretization. Thus, achieving a high-quality result also necessitates taking a large number of particles into account.
When large simulations are unavoidable for solving a specific simulation task even modern workstations reach their limits, especially if the softwares used don’t utilize the computer’s processing power optimally. To make sure that your resources are always optimally utilized Pasimodo® offers special features for increasing performance. Your simulation problems are automatically structured (parallelized) in a way that makes it possible for several processing units to work on them simultaneously. Modern workstations from generation to generation increasingly offer more processing cores available for parallel computing. Pasimodo® always ensures that, if necessary, all cores of your workstation are used for the simulation in order to fully utilize the available processing power.
If the processing power of a single workstation is not enough to cope with very large simulations another way for increasing performance is using Pasimodo® on HPC clusters for simulations. A special adaptive MPI-based load balancing process helps to ensure that even for highly dynamic simulations the available processing power is always optimally utilized.
Thanks to its sophisticated program architecture Pasimodo® Unleash makes it possible for users to do parallel simulations with program add-ons developed of their own, without them needing to concern themselves with the complex topic of parallelization. Program add-ons developed by a user will be available both on workstations and on HPC clusters in the same way. So, if there are several simulation platforms available, the most suitable one for a given simulation task can easily be chosen without making it necessary to adapt or port the self-developed program parts.