Parallelization in Modelica
Kaj Nyström, Peter Aronsson, Peter Fritzson · International Modelica Conference · 2005
The better the computer, the larger and more precise simulations can be carried out, and the more beneficent modeling can be. It is well known that faster computers enable more precise and computationally expensive simulations to be carried out, which allow more pre-cise mathematical models. This paper gives an overview of certain methods for expanding the limits of what can be done in the area of simulation by parallelizing simulations based on Modelica [18, 16] models. This is an efficient and less expensive way of achieving better simula-tion performance. In the following, we will restrict ourselves to describing various ways of parallelizing a simulation in Modelica, ranging from coarse grained high level parallelization to fine grained task merging at a very low level. It is very difficult to say which approach is the most successful or promising since little research has been done in most of the subareas of parallelizing Mode-lica models. Task merging seems to be the most developed approach and does give significant performance increases [1] but the other areas are largely unexplored. We can therefore only guess that based on parallelization research in other areas, there is little to gain for a normal user in parallelizing a small simulation. Larger, more complex simulations on the other hand can benefit greatly from parallelization, especially if it can be done automatically.