apillekeit / fmi-standard Goto Github PK

Please excuse if this may have been discussed before or already clarifed, but in the current version of the state machine diagram for cs fmi3: /docs/images/calling-sequence-co-simulation.svg
shows the "event mode" as the entry point.
However if the canHandleEarlyReturn (maybe other new features as well) feature is set to false, how should an FMU transition into StepMode, which is the only mode for 2.0 Co-Simulation?
I see the following options:

• Option one: Change the default entry point to someone that is common to all co-simulation fmu types i.e.: default entry point for fmi3 cs would be step mode.
• Option two: add more than one state machines / behiaviours to the spec document
• Option three: add an additional mode common to all cs fmu types?

There may be someone who is in favor of having a cs state machine similar to model exchange with event mode as entry point. but in this case what to do with fmi2 cs fmus?

The working group during the meeting in Regensburg suggests to merge fmi3EarlyReturn and fmi3ClockedCoSimulation into one fmuType fmi3HybridCosimulation.

Reason:
Reduce the number of fmuTypes in order to reduce complexity in the standard document and cross checks and real world interoperability issues. Exporting tools which support early-return only would simply not define any clock in the modelDescription.XML.

Contra:
There are importing tools that do not support clocks but want to support a Hybrid-Co-Simulation without clocks. The issue would be that such tools can not describe their support based on fmuType support.

For completness: the build proces described in CONTRIBUTING.adoc using Ruby and asciidoctor gems won't produce an output with appropriate translated equations (currently using latexmath). Should we extend the description on how to do this? (Im not sure if this is possiblie in html)
Or should we translate the equations to texmath or similar? @t-sommer what do you think?

Option 1: No – FMU would have to stop computation close to CP
Option 2: Yes – FMU and Master could exchange information from the „simulation future“

Needs more discussion
This is related to the new state machine: modelica#535 (comment)