M1 issue with steadystate

[luciansmith]

In addition to some of the tests failing on the M1, Paul Jost also reports different results on the M1 vs. other platforms. Here's his script:

roadrunner_instance.conservedMoietyAnalysis = True
# here a call is missing where we set parameter values
roadrunner_instance.getSteadyStateSolver().setValue(
                "allow_presimulation", True
            )
# steady state output = observables + state variables
steady_state_selections = observables_ids + state_variables
roadrunner_instance.steadyStateSelections = steady_state_selections
steady_state = roadrunner_instance.getSteadyStateValuesNamedArray()

And here's the model in question:
environment.xml.txt

[adelhpour]

I can confirm that we are getting different results on macOSs with apple silicon chips for this example.

[luciansmith]

@PaulJonasJost Working on this...

[PaulJonasJost]

@luciansmith thanks, what is the current status of this? :)

[luciansmith]

@adelhpour is still tracking this down, I believe. We're closer, but don't have a final solution. One issue seems to be that due to numerical differences, the calculations are literally different on MacOS M1 vs. other platforms.