Frontend replaces constant array with parameter subscripts

[Jens Frenkel]

In the model Modelica.Electrical.Digital.Examples.DFFREG the frontend replace the line of an algorithm

nextstate[i] := T.StrengthMap[L.'0', strength];

to

dFFREG.dFFR.nextstate[i] := .Modelica.Electrical.Digital.Interfaces.Logic.'0';

but strength is a parameter, hence the replacement is illegal.

With the debugflag "setDebugFlags("scodeInstShortcut");" it is the same.

[Adrian Pop]

Hi,

I guess T.StrengthMap is a package constant. We replace package constants where they appear but it seems we don't calculate the seconds index (strenght) constant-ness correctly in this case. I'll look what we do wrong there.

Cheers,
Adrian Pop/

[Adrian Pop]

Hi again,

It seems we do constant evaluation for all subscripts that are constant or parameter.
So I guess we shouldn't do it for parameters. I'll see what tests I break when I only constant-evaluate constants.

Cheers,
Adrian Pop/

[Adrian Pop]

Unfortunately the fix breaks too many models and we cannot fix this bug right now.
The problem is that we don't have a way yet to detect which parameters are final and we should constant evaluate and which are not and we shouldn't constant evaluate.
The fix is in the svn commented out.
Note for me: maybe add final prefix to DAE.ATTR so we can access it in Static.mo.

As far as I see the model works fine, but needs to be re-translated if strength changes.

Cheers,
Adrian Pop/

[Martin Sjölund]

It is allowed to evaluate any parameter you like in a Modelica front-end. It is only an error if strength was not marked as a structural parameter. And else it's a minor loss of functionality (user cannot modify the parameter between the build and simulate steps).

[Jens Frenkel]

So if the frontend evaluate a parameter it would be good to mark it as evalutate=True. Then it can be also replaced by the backend and the user can get an information about that.
I know that this model works fine with the replacement but work also models fine with parameter(fixed=false)? (see also Ticket #1964)
Why do we not inserd the row with the parameter as index:

  = {L.'0', L.'0', L.'L', L.'0', L.'Z', L.'L', L.'L', L.'Z', L.'0', L.'L'}[strength]; 

I know minor less effizient, but at least correct.

[Martin Sjölund]

I do believe it is because (previously?) the backend would be unable to handle code like:

arr[x] = 1.5;

Unless it knew the index x.

As for the evaluated parameters, you have them in: Env.CACHE(evaluatedParams=(_,lst))

And they should be marked final through Compiler/FrontEnd/DAEUtil.mo: elts := List.map1(elts,makeEvaluatedParamFinal,Env.getEvaluatedParams(cache));

Maybe it is possible to modify that code to also mark them Evaluate=true, and force a ceval of the binding?

[Adrian Pop]

These are not the problems we face.
My fix does this, basically as Jens said:

      elseif dFFREG.dFFR.reset_flag == 2 then
        dFFREG.dFFR.nextstate[i] := {{Modelica.Electrical.Digital.Interfaces.Logic.'U', .Modelica.Electrical.Digital.Interfaces.Logic.'U', .Modelica.Electrical.Digital.Interfaces.Logic.'U', .Modelica.Electrical.Digital.Interfaces.Logic.'U', .Modelica.Electrical.Digital.Interfaces.Logic.'U', .Modelica.Electrical.Digital.Interfaces.Logic.'U', .Modelica.Electrical.Digital.Interfaces.Logic.'U', .Modelica.Electrical.Digital.Interfaces.Logic.'U', .Modelica.Electrical.Digital.Interfaces.Logic.'U', .Modelica.Electrical.Digital.Interfaces.Logic.'U'}, {.Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W'}, {.Modelica.Electrical.Digital.Interfaces.Logic.'0', .Modelica.Electrical.Digital.Interfaces.Logic.'0', .Modelica.Electrical.Digital.Interfaces.Logic.'L', .Modelica.Electrical.Digital.Interfaces.Logic.'0', .Modelica.Electrical.Digital.Interfaces.Logic.'Z', .Modelica.Electrical.Digital.Interfaces.Logic.'L', .Modelica.Electrical.Digital.Interfaces.Logic.'L', .Modelica.Electrical.Digital.Interfaces.Logic.'Z', .Modelica.Electrical.Digital.Interfaces.Logic.'0', .Modelica.Electrical.Digital.Interfaces.Logic.'L'}, {.Modelica.Electrical.Digital.Interfaces.Logic.'1', .Modelica.Electrical.Digital.Interfaces.Logic.'H', .Modelica.Electrical.Digital.Interfaces.Logic.'1', .Modelica.Electrical.Digital.Interfaces.Logic.'Z', .Modelica.Electrical.Digital.Interfaces.Logic.'1', .Modelica.Electrical.Digital.Interfaces.Logic.'H', .Modelica.Electrical.Digital.Interfaces.Logic.'Z', .Modelica.Electrical.Digital.Interfaces.Logic.'H', .Modelica.Electrical.Digital.Interfaces.Logic.'H', .Modelica.Electrical.Digital.Interfaces.Logic.'1'}, {.Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W'}, {.Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W'}, {.Modelica.Electrical.Digital.Interfaces.Logic.'0', .Modelica.Electrical.Digital.Interfaces.Logic.'0', .Modelica.Electrical.Digital.Interfaces.Logic.'L', .Modelica.Electrical.Digital.Interfaces.Logic.'0', .Modelica.Electrical.Digital.Interfaces.Logic.'Z', .Modelica.Electrical.Digital.Interfaces.Logic.'L', .Modelica.Electrical.Digital.Interfaces.Logic.'L', .Modelica.Electrical.Digital.Interfaces.Logic.'Z', .Modelica.Electrical.Digital.Interfaces.Logic.'0', .Modelica.Electrical.Digital.Interfaces.Logic.'L'}, {.Modelica.Electrical.Digital.Interfaces.Logic.'1', .Modelica.Electrical.Digital.Interfaces.Logic.'H', .Modelica.Electrical.Digital.Interfaces.Logic.'1', .Modelica.Electrical.Digital.Interfaces.Logic.'Z', .Modelica.Electrical.Digital.Interfaces.Logic.'1', .Modelica.Electrical.Digital.Interfaces.Logic.'H', .Modelica.Electrical.Digital.Interfaces.Logic.'Z', .Modelica.Electrical.Digital.Interfaces.Logic.'H', .Modelica.Electrical.Digital.Interfaces.Logic.'H', .Modelica.Electrical.Digital.Interfaces.Logic.'1'}, {.Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'X', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W', .Modelica.Electrical.Digital.Interfaces.Logic.'W'}}[Modelica.Electrical.Digital.Interfaces.Logic.'0', dFFREG.dFFR.strength];

The problem is that for other models (mainly Fluid) we get a lot of flowModel.states[1] not found in scope and I have no way of fixing that right now.

I could probably add Evaluate=true, but where is that stored? In the parameter comment/annotation?

[Jens Frenkel]

Yes in the comment of the Variable, see also BackendDAEOptimize.hasEvaluateAnnotation

[Adrian Pop]

Ok. I'll see if I can add the Evaluate=true for these.

[Martin Sjölund]

Postponed until 1.9.1

[Martin Sjölund]

This ticket was not closed for 1.9.1, which has now been released. It was batch modified for milestone 1.9.2 (but maybe an empty milestone was more appropriate; feel free to change it).

[Martin Sjölund]

Milestone changed to 1.9.3 since 1.9.2 was released.

[Martin Sjölund]

Moved to new milestone 1.9.4

[Martin Sjölund]

Milestone pushed to 1.9.5

[Martin Sjölund]

Milestone renamed

[Martin Sjölund]

Ticket retargeted after milestone closed

[Martin Sjölund]

Milestone moved to 1.12.0 due to 1.11.0 already being released.

[Francesco Casella]

This issue is still valid with the NF

[Per Östlund]

Replying to casella:

This issue is still valid with the NF

For the mentioned equation the NF right now generates:

dFFREG.dFFR.nextstate[i] := {
  Modelica.Electrical.Digital.Interfaces.Logic.'0',
  Modelica.Electrical.Digital.Interfaces.Logic.'0', 
  Modelica.Electrical.Digital.Interfaces.Logic.'L',
  Modelica.Electrical.Digital.Interfaces.Logic.'0', 
  Modelica.Electrical.Digital.Interfaces.Logic.'Z', 
  Modelica.Electrical.Digital.Interfaces.Logic.'L', 
  Modelica.Electrical.Digital.Interfaces.Logic.'L', 
  Modelica.Electrical.Digital.Interfaces.Logic.'Z', 
  Modelica.Electrical.Digital.Interfaces.Logic.'0', 
  Modelica.Electrical.Digital.Interfaces.Logic.'L'}[dFFREG.dFFR.strength];

The way constants are handled in the NF was recently changed, but even before that the strength parameter wasn't evaluated since it's not structural. So I don't think the NF has ever had the issue this ticket is about (it had other unrelated issues with Digital models though).

[Francesco Casella]

OK.

@jfrenkel, feel free to reopen this if needed