Evaluation of Clock parameters

[Rüdiger Franke]

The following model has the evaluated parameter dt to parameterize its clock (see also Modelica_Synchronous):

model SID
  parameter Real dt = 0.1 annotation(Evaluate=true);
  Real x(start = 0);
  input Real u(start = 1);
  output Real y;
equation
  when Clock(dt) then
    x = previous(x) + u * interval(u);
    y = previous(x);
  end when;
end SID;

Exporting it as FMU with

translateModelFMU(SID, version="2.0");

gives the Clocks section:

    <Clocks>
      <Clock><Triggered/></Clock>
    </Clocks>

Triggered is used because the value of dt is not known during code generation. It should be known due to annotation(Evaluate=true). The correct Clocks section reads (e.g. replace parameter Real dt with constant):

    <Clocks>
      <Clock><Periodic
              baseInterval="0.1"
              /></Clock>
    </Clocks>

How can the evaluated value of dt be inquired during code generation?

[Lennart Ochel]

I think this should get already replaced in the back end.

[Lennart Ochel]

It seems that parameters get not replaced within the clocks, since they were introduces recently and the eval-module does not consider clocks yet.

[Lennart Ochel]

​PR279 adds missing replacement rules for clock expressions.

[Lennart Ochel]

Replying to rfranke:

gives the Clocks section:

    <Clocks>
      <Clock><Triggered/></Clock>
    </Clocks>

Triggered is used because the value of dt is not known during code generation. It should be known due to annotation(Evaluate=true). The correct Clocks section reads (e.g. replace parameter Real dt with constant):

    <Clocks>
      <Clock><Periodic
              baseInterval="0.1"
              /></Clock>
    </Clocks>

Where do I find this information?

[Bernhard Thiele]

Replying to rfranke:
I guess you are experimenting with adding clock features to FMI 2.0 since Clocks and DiscreteStates are not elements of FMI 2.0?

[Rüdiger Franke]

Super. The change works. 8847a737b601ca36fe2bbe4ff070b8bb8d83e37a/OpenModelica-testsuite updates the respective test.

Find the proposed spec in FCP-001_Clocks:

​https://trac.modelica.org/fmi/ticket/353

The originating time events proposal had been under discussion since 2011. FCP-001 simplifies the original proposal (minimal extension to C API, no sub-clocks in the XML description). Moreover it attempts to make the extension forward compatible with FMI 2.0 and it combines time events with discrete states that had been removed shortly before the release of FMI 2.0, lacking test implementations.

By now we have test implementations in the exporting tool OpenModelica (+simCodeTarget=Cpp) and the importing tool HQP, besides the original Dymola test implementation. Hopefully this will bring us forward in the FMI standardization, because we really need such models in our applications.

Modelica was lacking behind other simulation technologies prior to Clocks -- FMI is lacking behind until today.

[Martin Sjölund]

Milestone renamed

[Martin Sjölund]

Milestone renamed