What makes Rapid Accelerator regenerate code?

I have a very simple model containing a top model and two Model blocks that reference the same model. When I run with Rapid Accelerator, the Model reference SIM target for the referenced model is always out of date when I hit the play button and always has to be rebuilt, even if I don't do anything between one run and the next. I don't get this behavior when running with Accelerator. What's the difference between Accelerator and Rapid Accelerator when it comes to needing a rebuild?

Also, I tried to use the getChecksum command to sort it out, but that resulted in an error, which is kind of weird.

Answer 1

[MarkB]

You can control how "sensitive" Simulink is about rebuilding model references by going to the "Model Reference" pane of the Configuration Parameters. The "Rebuild" section of the dialog lets you set how often these rebuilds occur.

The behavior that you described (rebuilding even if nothing changed), doesn't sound like expected behavior. Could you please confirm that it is actually rebuilding, and not just taking a long time for the other things that go on during update? If this is the case, please send it in to us as a bug. The way to confirm this would be to look at the "created on" dates on the files generated for the model references. Depending on the models, sometimes it takes some time even to confirm that a rebuild is unnecessary.

As far as the criteria that should trigger re-builds, you can find more information here:

MathWorks Help

The key difference is that Rapid Accelerator has the solver "baked into" its code, whereas Accelerator just uses Simulink's solver. This makes Rapid Accelerator rebuilds more sensitive to changes in solver settings.