search

OpenModelica / OpenModelicaLibraryTesting

Test script for OMCompiler+OpenModelicaLibraries
Other
9 stars 13 forks source link

#79 dramatically worsened simulation times and causes large simulation errors #80

Open casella opened 7 months ago

casella commented 7 months ago

@adrpo, please check https://libraries.openmodelica.org/branches/history/master-fmi-fmpy/2024-03-18%2022:22:01..2024-03-19%2020:08:50.html. Apparently, many simulations went from a few seconds to fifty-something.

I'm not sure if this is just a glitch in reporting, these timings are a bit suspicious. How can we check?

casella commented 7 months ago

The 50 s timeout likely plays some role, but we still get successful simulation and verification, so it's not failing. Not sure what is going on.

adrpo commented 7 months ago

I remember you said that if you have and output-interval as non integer that it would get stuck, maybe this is what it happens and the timeout just forces to finish (correctly).

casella commented 7 months ago

Yes, that could be, see CATIA-Systems/FMPy#655.

But then how comes we still get the verification right? Could it be that it is still using the previously computed result file? If so, then you should probably update the test script to delete all such file before starting the new simulation.

adrpo commented 7 months ago

Yes, that could be, see CATIA-Systems/FMPy#655.

But then how comes we still get the verification right? Could it be that it is still using the previously computed result file? If so, then you should probably update the test script to delete all such file before starting the new simulation.

We always delete the previous results so that's not the case. It is possible that the simulation is finished in way less time that 50 but is stuck, so the result files are correct when is killed.

casella commented 7 months ago

OK. Any clue what is actually going on there?

casella commented 7 months ago

One more piece of info: many of those models timing out fail on verification with really weird results, e.g.:

immagine

Something's really weird going on there. @adrpo, how can we debug this? Can you try to run the same fmpy command line on your computer and see what the result is?