Multiple FMUs and events handling

[nl78]

Hello,

Event and clocked variables management may become tricky when considering multiple FMUs. I tried to draw a diagram describing the sequence of API calls :

Notes:

• fmi3Get / fmi3Set use common buffers to share values between FMUs.
• Importer Clocks refer to the clocks "managed" by FMI-importer, typically the TX_Clock of bus.fmu (in LS-BUS example).
• Initialization of FMUs is not detailed here.
• next ticks(s) is (are) the closest one(s) returned by fmi3GetInterval. The communicationStepSize is set accordingly to this scheduled tick.

I would be glad to have your feedback to improve, fix or clarify this diagram.

Kind regards

[nl78]

It seems that FMI importer should tick its clocks in step mode. Here is an updated version the diagram.

I would be glad to have your feedback to improve, fix or clarify this diagram.

[pmai]

I don't know where this idea comes from that countdown clocks would be different from other clocks, but that is not true. For all clocks in CS (and ME) they can only be active in event mode (a clock being active implies an event, and can only occur in event mode). For scheduled execution this is slightly different, as there is no event mode and clock activation is handled very differently, but for CS/ME there is no distinction between the clocks in terms of their relationship to event mode (note that not being in event mode implies that you are in a continuous mode like step mode).

Also the flow charts seem to miss the fact that clocked variables (i.e. variables associated with a clock) can only be accessed when that clock is active, hence all of the Set/Get access to clocked variables needs to happen in event mode at the appropriate times.

Also the sequence of clock activation/variable accesses/event iteration seems unclear in the flow chart (different clocks can be active for different event iterations/super-dense time instants during one event mode phase), and the influence that scheduled events have on the selection of time-steps for the normal stepping seems not reflected.

Just as a quick drive-by the more obvious points to me...

[nl78]

Thank you for this feedback Pierre. (I focus only on CS)

Regarding clocks, my diagram is confusing: I made a difference between

• clocks managed by the fmi-importer (at the moment only "countdown", but you're right, it should be generalized. I will update my diagram)
• and clocks which are outputs of FMU (and connected to other ones).

For clocks managed by the fmi-importer, should their activation happen only in Event Mode ? I faced a FMU which raised a error when fmi-importer uses fmi3SetClock() during Event Mode.