Differentiation of recursive function fails

[Rüdiger Franke]

This is the smallest example I'm aware of that causes most trouble in OpenModelica.

model SortTest2
  input Real u; // prevent evaluation by frontend
  Real[:, :] M = [3, 4; 1 + u, 2];
  output Real[size(M, 1), size(M, 2)] N;
equation
  N = Modelica.Math.Matrices.sort(M);
  annotation(uses(Modelica(version="3.2.1")));
end SortTest2;

It not only knocks out the Cpp code generation (#3263), while it's reduced version crashes OMEdit (#3262). Also FMI export fails to generate the model structure -- translateModelFMU(SortTest2, version="2.0"):

Error: Internal error BackendDAEOptimize.generateSymbolicJacobian failed
Error: Internal error BackendDAEOptimize.createJacobian failed

[Rüdiger Franke]

It's not related to Cpp. We had similar issues with failed Jacobians in January.

Willi, could you have a look?

[Adeel Asghar]

Well the C runtime works fine with the attached example. I checked it before changing it to FMI-CPP. So i don't understand why you set the component to FMI.

[Rüdiger Franke]

Also when using the C runtime, translateModelFMU(SortTest2, version="2.0") gives the same error message:

Error: Internal error BackendDAEOptimize.generateSymbolicJacobian failed
Error: Internal error BackendDAEOptimize.createJacobian failed

The modelDescription.xml contains an empty ModelStructure section. Can it be that you checked using OMEdit -- it ignores this error message recently (did not ignore it back in December/January)?

[Rüdiger Franke]

I might be wrong with my files from January/February -- I was mostly generating FMUs from the command line.

Can you please also check #3262 -- there are still issues with error handling OMEdit.

[Adeel Asghar]

Which revision you have?
Post the output of getVersion().

[Rüdiger Franke]

My current OMEdit has version r25313 -- the one crashing in #3262. Moreover I tried a local build of r25402. The Jacobian failure is the same in both versions.

[Adeel Asghar]

See my comment in #3262 for OMEdit crash.
About Jacobian error, you are right. I wasn't checking for the errors when translateModelFMU returns true. I fixed it now in r25442.

[Rüdiger Franke]

Btw. I just did

$ svn update
$ make omc

The update says "At revision 25447". getVersion() still reports "1.9.3+dev (r25402)". I like the treatment of dependencies a lot when building omc, saving a lot of time when only recompiling things that have changed.

But where does the result from getVersion() come from?

[Adrian Pop]

It comes from Compiler/runtime/omc_config.h and is updated only if you runt configure I think.

[Rüdiger Franke]

Right, configure updates the version -- and the compilation still treats dependencies. Thanks!

[Rüdiger Franke]

One reason could be that there are no states in the model. ModelVariables start with <!-- Index of variable = "3" -->.

[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]

The milestone of this ticket has been reassigned to "Future".

If you think the issue is still valid and relevant for you, please select milestone 1.13.0 for back-end, code generation and run-time issues, or 2.0.0 for front-end issues.

If you are aware that the problem is no longer present, please select the milestone corresponding to the version of OMC you used to check that, and set the status to "worksforme".

In both cases, a short informative comment would be welcome.

[Rüdiger Franke]

The situation has improved meanwhile. The example simulates in OMEdit now.

Remaining issues are:

1. FMI export fails: [C:/dev/OpenModelica32bit/OMCompiler/Compiler/Util/Graph.mo: 990:5-990:101]: Interner Fehler Graph.addForbiddenColors failed. -- FMI export works with the debug flag -disableColoring
2. OMEdit crashes during FMI export with the flag -disableDirectionalDerivatives

These remaining FMI issues can hopefully be solved as well.

[Willi Braun]

Fixed the differentiation of recursive function in c15e7bd/OMCompiler.

[Rüdiger Franke]

The crash is gone. Instead the following error message is raised:

[6] 18:37:22 Symbolic Error
[OpenModelica/OMCompiler/Compiler/BackEnd/Differentiate.mo: 395:3-395:155]: Derivative of expression "(N, _) = Modelica.Math.Matrices.sort({{3.0, 4.0}, {M[2,1], 2.0}}, true, true)" w.r.t. "dummyVarFMIDER" is non-existent.

[Willi Braun]

Yes, I try to fix it with ​PR2203.

[Rüdiger Franke]

Since ​PR2206 the FMU can be generated when using default OMC flags.

Do you think the ticket can be closed, i.e. the error message in comment:23 with the flag -disableDirectionalDerivatives is valid?

[Willi Braun]

Actually, the differentiation of the function sort still fails and I'm try to fix this in ​PR2203. I think I'm almost done, currently only one issue with tuples return functions in algorithm section is failing in code generation phase.

[Willi Braun]

It seems the differentiation fails for FMU's silently.

[Willi Braun]

​PR2203 is pushed, so I close it now.