Main situations and principles

Basically, two situations can be distinguished:

1. Your system can only be decomposed in several subsystems having similar dynamics. In this case, the use of multiple cores on the real-time target simply better distributes the CPU load.

2. Your system can be decomposed in subsystems having very different dynamics (ideally weakly coupled and such that the low frequency parts are the most CPU consuming ones). This situation is the one where the most significant speedups can be obtained.

In both situations, you need to split your Simcenter Amesim system into several parts. In the first case you can use the same fixed-step solver settings for all your subsystems whereas in the second case, it is advised to setup a co-simulation to fully exploit the multicore capability of your target. This basically consists in selecting relevant fixed-step solver settings for each of your slave subsystems, as well as optimized communication intervals with the master. Please refer to the Generic Co-Simulation manual for more information on co-simulation.

Even without co-simulation, testing your coupled model "offline" (i.e. prior to the export phase) is a safe practice that maximizes the chances to get it to work online. In addition, this allows you to already get a rough idea of the speedups you will obtain by applying similar solver and communication settings on the target.

Once these operations are completed, you can export your model on your multicore real-time target. We will describe this process more in details in the next sections.

Note

In the next sections, it is assumed that the Simcenter Amesim model export workflow of the considered multicore real-time target is based on the Simulink Coder™. This is the most common situation (see Appendix A: Supported Real-Time target workflows).