Towards Efficient Distributed Simulation in Modelica using Transmission Line Modeling
Martin Sjölund, Robert Braun, Peter Fritzson, Petter Krus · Equation-Based Object-Oriented Modeling Languages and Tools · 2010
The current development towards multiple processor cores in personal computers is making distribution and parallelization of simulation software increasingly important. The possible speedups from parallelism are however often limited with the current centralized solver algorithms, which are commonly used in today’s simulation environments. An alternative method investigated in this work utilizes distributed solver algorithms using the transmission line modeling (TLM) method. Creation of models using TLM elements to separate model components makes them very suitable for computation in parallel because larger models can be partitioned into smaller independent submodels. The computation time can also be decreased by using small numerical solver step sizes only on those few submodels that need this for numerical stability. This is especially relevant for large and demanding models. In this paper we present work in how to combine TLM and solver inlining techniques in the Modelica equation-based language, giving the potential for efficient distributed simulation of model components over several processors.