I do not quite understand how the “Model Settings” block works… I get the same result
The Electrical Model Settings block has two key roles.
First, on the RT Box 2/3 it can assign the electrical circuit to be simulated on a CPU Core (default) or on the FPGA-based FlexArray solver. Note you can only have 1x FlexArray solver instance per RT Box.
For the CPU, you can also configure different matrix coding style. Tweaking the matrix coding style can impact the execution time on the CPU, as different approaches may be more suitable for different types of electrical networks (depending on sparsity, etc.). Sometimes the impact is negligible.
Additionally, I have not seen the task transition block in any demo files; how should I use it?
The short answer - you don’t need to use it with the RT Box.
The task transition block can be used to modify the default behavior and decrease the latency when transitioning from a slower task to a faster task, as described in the documentation. Rather than having a deterministic latency based on the slow task execution rate, the latency can be shortened so as the fast task receives the data as soon as available.
However, this is specific for task-to-task transitions. For CPU to CPU transitions this behavior is defined by the TSP and for the RT Box, the task transition has no effect. See this example: CpuLatency.plecs (18.1 KB).
I’d recommend reading the “Task Transitions in Multi-Tasking Mode” to understand the various task transition behaviors.
And so, in general, when I want to divide a model into different tasks, what is the best method and which blocks should be used?
The mechanics of this, as shown in the model, are quite straightforward. You must use the task frame to assign the different tasks to a CPU core. Only signal data can be exchanged between the cores.
The challenge is often finding an appropriate place to split your model, balancing load across the cores, determining the optimal step size for each of your cores, and ensuring the latency introduced does not significantly degrade the model fidelity.
Also, fantastic profile picture!
