Help Please !

Hi,

Your code is wrong both in the C file and in the .mo file. See the attached zip.
Model.zip

  function mafonction 

    input Rec u1; 

    output Rec u2; 

    // send BOTH u1 and u2 to the external function as that is the way is defined!

    external mafonction(u1, u2) annotation(Library = "Chirp", LibraryDirectory="modelica://ChirpSignal"); 

  end mafonction; 

rec* mafonction(rec test, rec* test1) 

{ 

  memcpy ( test1, &test, sizeof(rec) );  // no addres of test1, is already a pointer

  return test1; 

} 

Running model.mos gets you:

adrpo@ida-liu050 ~/dev/OMTesting/forum/1665

$ /f/OpenModelica1.9.3-v1.9.3-dev-523-g1d963ac/bin/omc +locale=C model.mos

true

""

true

0

""

record SimulationResult

    resultFile = "c:/bin/cygwin/home/adrpo/dev/OMTesting/forum/1665/ChirpSignal_res.mat",

    simulationOptions = "startTime = 0.0, stopTime = 1.0, numberOfIntervals = 500, tolerance = 1e-006, method = 'dassl', fileNamePrefix = 'ChirpSignal', options = '', outputFormat = 'mat', variableFilter = '.*', cflags = '', simflags = ''",

    messages = "",

    timeFrontend = 0.7343596380537807,

    timeBackend = 1.148104848302405,

    timeSimCode = 0.006371707033899545,

    timeTemplates = 0.03741838078879516,

    timeCompile = 3.829548752874621,

    timeSimulation = 0.1699746779308428,

    timeTotal = 5.926586016890105

end SimulationResult;

""

"u1"

7.0

""

6.0

""

5.0

""

4.0

""

15.0

""

"u2"

7.0

""

6.0

""

5.0

""

4.0

""

15.0

""

Cheers,
Adrian Pop/

Hi Adrian,

First , thank you for your answer.
What you proposed doesn't change anything.
for the error in C code, I totally agree with you. It was just an error while copying the code.
for the OM code : 
"external u1 = mafonction(u) annotation(Library = "Chirp");"  and  "external mafonction(u1, u2) annotation(Library = "Chirp", LibraryDirectory="modelica://ChirpSignal"); " are totally equivalent (at least in this exemple), I still get the same result.

Actually the solution is to declare the first function as impure and pass it as argument of the second one.  :

impure function Chirp
    output Rec u;
 
    external Chirp(u) annotation(Library = "Chirp1", LibraryDirectory = "modelica://ChirpSignal");
  end Chirp;

  function mafonction
    input Rec u1;
    output Rec u2;
 
    external u2 = mafonction(u1) annotation(Library = "Chirp1", LibraryDirectory = "modelica://ChirpSignal");
  end mafonction;

  Rec u2;
equation
  u2 = mafonction(Chirp());
end ChirpSignal;

Try this and tell me what u get (It works just fine).

Anyway, thanks for your Help.

Cheers,
Olaama

The C code was wrong. you took the address of the pointer when doing memcpy.
And:

external u1 = mafonction(u) annotation(Library = "Chirp");

is not equivalent to:

external mafonction(u1, u2) annotation(Library = "Chirp", LibraryDirectory="modelica://ChirpSignal");

are not equivalent!

in the first case we generate u1 = mafonction(u) in the C code and in the second mafonction(u, &u1).
So if your C function is given as having 2 parameters you need to have the second style!
Do a diff of your code and the code in the zip i attached before to see what i changed!

Cheers,
Adrian Pop/

Hi ,
for the C code I said that I totally agree with u (there was a mistake thank u for reminding me).
for the OM code : OK , I thought it was equivalent . but still, even with what u gave me, the result is the same, the values are in the reverse order.
Have u tried what I gave u ? It works !!

Cheers,
Olaama

It seems I really don't get what you're saying
So, is the problem solved now?

The values are reversed when in the .mo file you have out = f(in) and in the C code you have f(in, out)
I guess because the stack is somehow corrupted when the function expects 2 parameters and gets only one.

Yes, the problem Is solved, the solution is to define the first function in OM like this : "impure function Chirp" instead of "function chirp", and in the equation part you must call it this way

equation
u1 = Chirp();
u2 = mafonction(Chirp());
end ChirpSignal;

//instead of :
/*
u1=Chirp();
u2 = mafonction(u1);
*/

that's it.

We do evaluate the functions during compilation if they are not declared impure and they have constant inputs.
An alternative is to use +d=nogen,noevalfunc but is better to declare them impure as you did if you don't want them to be evaluated during compile time.

This should not be a problem at all:

u1=Chirp(); 

u2 = mafonction(u1); 

It does work fine in the zip i attached and not have things in reverse.

Thank you for your answer ! It was very helpul.