Missing support for operator interval

[Lennart Ochel]

This can be used as test case: Modelica_Synchronous.Examples.CascadeControlledDrive.SubClocked.

[Lennart Ochel]

It's the same issue for Modelica_Synchronous.Examples.SimpleControlledDrive.ExactlyClockedWithDiscreteController.

[Lennart Ochel]

I think the generated code for interval operator should be something like this:

/*
 equation index: 129
 type: SIMPLE_ASSIGN
 PI._Ts = interval(PI.u)
 */
void Modelica_Synchronous_Examples_CascadeControlledDrive_SubClocked_eqFunction_129(DATA *data, threadData_t *threadData)
{
  TRACE_PUSH
  const int equationIndexes[2] = {1,129};
  $PPI$PTs = data->simulationInfo.clocksData[clock_index].interval;
  TRACE_POP
}

The issue is that clock_index need to be resolved. I think it should be idx within the function handleTimers in synchronous.c, right? Should we just pass it through the call hierarchy?

[Rüdiger Franke]

The clock index is known for the system for which the equation is generated. In the C++ runtime we generate a constant clockIndex for each equation. This is passed through one level (from the system to the equation).

void SubClocked::evaluate_129()
{
  const int clockIndex = 2;
  _PI_P_Ts = _clockInterval[clockIndex - 1];
}

[Maksimov Doe]

According to Modelica specification partitioning algorithm, variables on left and right sides of interval operator belong to the same subpartition. So in fact this operator will always return the interval of current subpartition.

[Lennart Ochel]

Thanks for the input. I adjusted the code generation similar to the cpp runtime.

[Rüdiger Franke]

Do you also consider models with inferred sample time?

This is important for FMU export. The interval is not known in this case and must be obtained at simulation time.

[Lennart Ochel]

Compilation is fixed with 3a9e48.
All models mentioned in this ticket do simulate now. However I didn't compare the results yet.

Replying to rfranke:

Do you also consider models with inferred sample time?

I think that was already implemented, but I am not sure right now. Once we have verified simulations available it will be much easier to identify what is missing.

[Lennart Ochel]

Fixed in 3a9e48.

[Martin Sjölund]

Milestone renamed

[Martin Sjölund]

Milestone renamed