Many Simulink processes run single-threaded. There are several instances where Simulink will run multithreaded implicitly, for example:
- Simulink uses BLAS and LAPACK libraries that can use multiple threads for linear algebra.
- By default, Simulink will attempt to multithread the execution of "For-Each" subsystems, S-functions, FMUs, and Accelerator Model Reference. - Run Co-Simulation Components on Multiple Cores - MATLAB & Simulink Multithreaded Simulation Using For Each Subsystem - MATLAB & Simulink
It is also possible to scale your simulations to be "multi-process" by using the Parallel Computing Toolbox to:
- Parallelize model reference compile and code generation - Reduce Update Time for Referenced Models by Using Parallel Builds - MATLAB & Simulink
- Parallelize multiple simulation workflows - Running Multiple Simulations - MATLAB & Simulink Comparison Between Multiple Simulation Workflows - MATLAB & Simulink
Keep in mind that not all tasks (from a computing standpoint) can be sped up with multithreading. Overheads, hardware bottlenecks, and data dependencies can actually make some software perform slower multithreaded vs single threaded. If performance is your concern, many Simulink performance problems will see much better performance improvements by profiling and identifying hot spots in the workflow and using various strategies to address them as mentioned in the link below.
