Triggering Clock Update Mode from a preemptable partition results in unpredictable activation time?

[yoerg]

The non-normative text in 5.2.2. State: Clock Update Mode explicitly states that the triggering clock for callbackClockUpdate neither can be nor needs to be identified. Countdown clock intervals (and, I assume, output clock activation times as well) are instead relative to the "time of the current [...] Clock Update Mode", which I read as the last preceding call to activateModelPartition?

Yet that would lead me to conclude that for an SE-FMU, triggering Clock Update Mode from any preemptable partition (i.e. usually every partition except the highest priority partition) would effectively result in an unpredictable (though still well defined) activation time for the triggered output and countdown clocks.

What is the intention of this design? It seems to make Clock Update Mode needlessly complicated for both FMU and importer implementers. Or did I read that wrong somewhere?

[MBlesken]

Hello @yoerg by

that the triggering clock for callbackClockUpdate neither can be nor needs to be identified

you propably refer to "In Scheduled Execution it cannot be determined which model partition has called fmi3ClockUpdateCallback... ", right? That doesn't mean that the scheduler couldn't activate the next task with an appropriate activationTime. Maybe you can draw some conclusions from Figure 53. "Scheduled Execution Example ThreeInputClocks" where the aperiodic task is triggered for time t_{a0}.

It would be nice if you could leave your contact information in https://github.com/modelica/fmi-design/blob/master/Members/Members.md then maybe we could organize a meeting and look at it together.