Automatic parallelization of multi-rate fmi-based co-simulation on multi-core

Salah Eddine Saidi (1) , Nicolas Pernet (1) , Yves Sorel (2)

1 IFPEN - IFP Energies nouvelles
2 AOSTE2 - Models and methods of analysis and optimization for systems with real-time and embedded contraints

Co-simulation refers to simulating a complex system using several coupled numerical models. Engineers define the rate of data exchange between the models by setting communication steps. FMI is a standardized interface which easily allows coupling and co-simulation of numerical models. The RCOSIM approach allows the parallelization on multi-core processors of co-simulations using the FMI standard. In this paper, we tackle the limitations of this approach. First, we extend the co-simulation to multi-rate, i.e. with different communication steps. We present graph transformation rules and an algorithm that allow the execution of each model at its respective rate while ensuring correct data exchange between models. Second, we present an acyclic orientation heuristic for handling mutual exclusion constraints between operations that belong to the same model due to the non-thread-safe implementation of FMI. We evaluate the obtained speedup on a multi-core processor and the effect on the accuracy of the numerical results.

Mots clés: parallel; scheduling; acyclic orientation; multi-core; co-simulation

Domaines: Calcul parallèle, distribué et partagé [cs.DC]; Systèmes embarqués; Modélisation et simulation; Système d'exploitation [cs.OS]; Recherche opérationnelle [math.OC]; Systèmes et contrôle [cs.SY]; Génie logiciel [cs.SE]

Salah Eddine Saidi, Nicolas Pernet, Yves Sorel. Automatic parallelization of multi-rate fmi-based co-simulation on multi-core. TMS/DEVS 2017 - Symposium on Theory of Modeling and Simulation , Society for Computer Simulation International San Diego, CA, USA, Apr 2017, Virginia Beach, United States. pp.Article No. 5. ⟨hal-01610268⟩