Issue with Rapid Accelerator Mode in parsim After Upgrading to R2024b

I am encountering issues with the workers in parsim when using the Rapid Accelerator mode.

After upgrading from R2024a to R2024b one week ago, I noticed that running parsim simulations no longer works as expected. I aim to run 6 simulations on a system with 6 physical (P) cores, allowing Matlab to compute all simulations simultaneously. However, only one worker completes the simulation, while the remaining five remain stuck in the "Running" state.

I tested this behavior with 6 identical models, but the issue persisted. Uninstalling and reinstalling Matlab did not do the trick.

Another observation is related to CPU behavior. When the parsim simulation starts, my laptop's fan runs intensively, and some CPU cores reach 100% utilization—this is expected and matches the behavior in R2024a. However, as soon as the first worker completes its simulation, the fan slows down, and CPU utilization becomes erratic. The CPU usage fluctuates significantly, jumping between 20% and 90%.

Additionally, if I stop the parsim simulation via the Simulation Manager, Matlab becomes stuck at the "Cleaning up parallel worker" stage for the workers that were stuck during the simulation.

Interestingly, when switching from Rapid Accelerator to Accelerator mode, the parsim simulations work correctly, and some CPU cores consistently run at 100% utilization.

Does anyone have a clue what the cause for this behavior?

[Oliver Jaehrig]

Is this happening for any model?

If yes, can you check if your anti-virus scanner maybe blocks the executables?

[Christian Hanzl]

Hi Oliver,

Thank you very much for your reply.

I’ve investigated further, and the issue occurs with every model run in Rapid Accelerator mode using parsim in MATLAB R2024b, even with minimal examples. However, when the same models are run in R2024a, they work as expected.

I also checked the Windows Security history for any blocked files or events. There are no entries indicating that any files or actions related to MATLAB have been blocked recently. Actually, there is no entry at all of anything being blocked recently.

Let me know if there’s anything else I should check or if additional details are needed.

Best regards,

Christian

Answer 1

[Divyam]

Hi @Christian Hanzl,

I have encountered this bug with MATLAB R2024b. The rapid accelerator mode breaks during parallel simulations if there are other versions of MATLAB installed on your machine besides R2024b.

The current workaround for this bug is to uninstall all other versions of MATLAB except R2024b and remove any path variables related to other MATLAB versions from the environment variables. You will need to reboot your system after making the necessary changes to the path variables.

To ensure this workaround is correctly implemented, run the following command in your terminal and check that no paths including earlier versions of MATLAB or MATLAB runtime are printed:

echo %PATH%

If this does not resolve your problem, you could refer to the other troubleshooting links to MATLAB answers here:

• https://www.mathworks.com/matlabcentral/answers/2026934-why-does-my-computer-experience-high-memory-usage-while-attempting-to-run-parsim-multiple-times
• https://www.mathworks.com/matlabcentral/answers/1650450-when-do-i-need-to-call-simulink-blockdiagram-buildrapidacceleratortarget-when-using-parsim-or-b

[Oliver Jaehrig]

@Christian Hanzl I can confirm that there is indeed such a bug existing in R2024b and the development team currently works on bug fixes for it.

Please check if the above workaround helps.

[Dan]

Hi Oliver, I have the same issue and the workaround doesn't fix it. Is there an ETA for fixing it? Are there other workarounds? It's quite a disruptive bug.

[Dan]

@Oliver Jaehrig We received an email informing us that this bug is fixed in Update 4, however our testing suggests that's not the case. The workarounds don't seem to work either.

[Christian Hanzl]

Hi @Dan,

The official website does not mention that this bug got fixed with Update 4, see https://de.mathworks.com/support/bugreports/?product%5B%5D=ML&release=R2024b&release_filter=Fixed+in&fir%5B%5D=R2024b-Update-4

My workaround is to work with R2024a. Nothing works for R2024b.

Best regards,

Chris

[Oliver Jaehrig]

The bug report is:

https://www.mathworks.com/support/bugreports/details/3479106

If the bugfix in R2024b Update 4 does not fix your issue, please create a Technical Support case so we can check if you are facing any other issue:

https://www.mathworks.com/support/contact_us.html