Simulation speed under Linux / Docker

Hi all,

I set up a Docker file for running Simulations on a more powerful machine than my laptop. The model itself uses Simscape Electrical, several Stateflow charts (triggering function call subsystems), many MATLAB function blocks and a good amount of custom C-code. I was able to get everything running in the Docker container on my workstation using SSH / X11 (with MATLAB GUI) and did a benchmark run of my model on the two machines.

• Laptop: DELL XPS13 9310, i7-1165G7 @ 2.8 GHz, 16 GB RAM, 500 GB SSD, Windows 11
• Workstation: Custom built, Ryzen 5900X @ 3.7 GHz, 64 GB RAM, 1 TB SSD, Ubuntu (Docker)

However, the simulation turns out to be three times faster on my laptop (24 minutes vs. 7 minutes) for whatever reason. I also tried to run it on the native Ubuntu system (without Docker) and it is even slower than in the Docker container.

In another test I tried one of the Mathworks Example Models ('ac6_IPMSM') and this one was a bit faster on the Linux workstation compared to my laptop, not significantly, though.

What could the reason be? A bad hardware configuration? Memory issues? Or something related to the model, e.g. custom C-code?

I am thankful for any hint that will speed up the simulation.

Best regards

Julian

[JulianSTABL]

I checked with the Simulink Profiler: The main cause are indeed the custom C functions. When I uncheck the option "Simulate custom code in a separate process", the simulation is faster on the workstation compared to the laptop. Still no idea why this happens, though.