[pbe_cfd]

So, I came up with a solution to my prototype problem. Calling an external Fortran90 subroutine from an OpenFoam application.

STEPS
1) Declare an interface C function in the main file or in .H file and include it

extern "C"
{
void fortfunc_(int *ii, double *ff); // underscore is important (gfortran linux)!
}

When mixing Fortran with C++, name mangling must be prevented. You may use "objdump -t <file>" file: object or library files. In this example the subroutine named as fortfunc in the Fortran 90 code. But it's referred as fortfunc_

2) Compile your Fortran code and obtain object files "*.o". You may link the object files or you may create a static library form object files as follows,

ar rc myPbeLib.a fortfunc.o other.o object.o files.o // specific to Linux system

3) In your solver folder Make/options append -lgfortran (runtime fortran libraries) and your object or library files. If your fortran subroutine call some other routines from other libraries, add them as well. (Last case is not tested. )

EXE_LIBS = -lfiniteVolume\
-lgfortran\
$(FOAM_USER_LIBBIN)/fortfunc.o

4) Build your executable with wmake in solver folder.

This approach worked for my prototype problem. However I don't know if this is the best way, most probably NOT. Cyp's suggestion sounds really cool. But, I had a problem with obtaining the shared object files (*.so). I had done it before for a C++ code for boundary conditions. There it had worked very nicely. But with Fortran code I had problems. It doesn't compile the fortran code. If someone can describe how to obtain shared object files for Fortran codes, it will be really cool. Then, we can close this thread with a happy end ;-)

All the best;
Evren

[Nicolas Périnet]

Dear all,

I have posted in this thread because it is the closest to the topic of my problem.

I want to code a Hookstep-Newton-Krylov algorithm to find peculiar solutions of the flow with some symmetry properties in a rotating annulus. This algorithm involves OpenFOAM for solving Navier-Stokes equations and additional routines in FORTRAN 90, especially from linear algebra libraries.

I have created a main program in C++ which calls the fortran fsubroutines (here a test case nothing.h. Some of these fortran subroutines call OpenFOAM. I am trying to make interact the main program with the fortran subroutines and OpenFOAM.

I have made a test case where:
-the main program is "icoDHRAFoamEssai.C"
-it calls the fortran function "nothing in "nothing.f90"
-which calls at its turn "integration" in "integration.C" that gathers all the OpenFOAM functions that are dedicated to solve Navier-Stokes equations in my case.

Here you can find some parts of my code.
First the main function in "icoDHRAFoamEssai.C"

#include "fvCFD.H"
#include <iostream>
#include <fstream>
#include <string>
using namespace std;

// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //

extern "C" { void nothing_(double* AA, int &lengthAA,int* argvLength, int &argc,char* argvChar, int& argv1StrLength); } 
void integration_(double* AA, int* lengthAA, int& argc);

int main(int argc, char* argv[])
{
      //commands that I didn't attach here

      nothing_(AA, lengthAA, argvLength, argc, argvChar, argv1StrLength); 

      //some other commands
}

the subroutine "nothing" is as follows:

subroutine nothing(AA, lengthAA, argvLength, argc, argvChar, argv1StrLength)
	  
integer lengthAA
real*8  AA(lengthAA)
integer argc
integer argvLength(argc)
integer argv1StrLength
character argvChar(argv1StrLength+1)

!some fortran instructions
      
call integration(AA, lengthAA, argc)

return
end

And then "integration" in "integration.C"

void Integration_(double* AA, int* lengthAA,int& argc)
{
      #   include "setRootCase.H"
      #   include "createTime.H"
      #   include "createMesh.H"
      #   include "readEnvironmentalProperties.H"
      #   include "createFields.H"
      #   include "initContinuityErrs.H"

      Info<< "\nStarting time loop\n" << endl;
	
      #   include "readPISOControls.H"
      #   include "CourantNo.H"

      //some instructions that I added
      
      for (runTime++; !runTime.end(); runTime++)
      {
             //OpenFOAM instructions
      }

      Info<< "End\n" << endl;

      //other instructions

}

I have first compiled it using the commands:

gfortran -c nothing.f90 -o nothing.o

To link "nothing" with the main code I changed the "Make/options" file in the directory of my project to:

EXE_INC = \
    -I$(LIB_SRC)/finiteVolume/lnInclude

EXE_LIBS = \
    -lfiniteVolume\
	/usr/lib64/libgfortran.so.1\	
	$(HOME)/OpenFOAM/nperinet-1.4.1/applications/solvers/icoDHRAFoamEssai/nothing.o

As stated in this thread, I just added the two last lines.
I compile with "wmake"; if I don't call the function "integration" in "nothing.f90", everything goes well. When I call it, I get the error message:

In function `nothing_':
:undefined reference to `integration_'

Then it is most likely because I don't link "integration" to "nothing" and the main program, but I can mistake (I am beginner in OpenFOAM and C++ as well). The problem is that I absolutely don't know how to do this (which commands I have to add to which file to make this work, etc...).

I would be very grateful to anyone who may supply any suggestion.

Nicolas Périnet

[anfho]

Dear All,

I am also going to call a whole list of FORTRAN 90 subroutines from OF (in my case 1.7.1 though). I do this by adding the *.o files to the "objectFiles".

Apparently, as I am reading through this really helpful thread, it seems like there are different methods to do this: static library, dynamic library, static objects, dynamic objects, shared objects, objects, ...

I would appreciate, if someone could comment on the differences of these strategies (efficiency, ease of use, ...). This question came up earlier in the thread but has not been answered.

Thanks guys. Best regards,
Andreas

[anfho]

Hi All,
did anyone try to link OF with fortran to read from an input file, something like:

integer temp
open(unit=10,file='input.dat')
read(10,*,end=100)temp
100 continue
close(10)

where the input file simply contains an integer, e.g.: "1",
and calling this from an OF application?

If I do this, I am getting a random output for "temp". It does work however, if I link a separate c++ main program and a fortran subroutine... so not sure why OF is causing trouble there...

Any help is appreciated.
Thanks, Andreas

[vahid.najafi]

Hi Dear Nima again:
I want to add surface tension(sigma) in one solver,for this reason I added :
#include ''fvCFD.H''
fvc::interpolate(interface.sigma())

in this code:
Foam::tmp<Foam::volScalarField>
Foam:haseChangeTwoPhaseMixtures::SchnerrSauer: Coeff
(
const volScalarField& p
) const
{
volScalarField limitedAlpha1(min(max(alpha1_, scalar(0)), scalar(1)));
volScalarField rho
(
limitedAlpha1*rho1() + (scalar(1) - limitedAlpha1)*rho2()
);
return

//......I want to change it( <<sigma>> surface tension multiple in it):
(3*rho1()*rho2())*sqrt(2/(3*rho1()))* (fvc::interpolate(interface.sigma()))
*rRb(limitedAlpha1)/(rho*sqrt(mag(p - pSat()) + 0.01*pSat()));
//.................................................. ......
}
dont successful wmake, and seen(was not declared ):
phaseChangeTwoPhaseMixtures/SchnerrSauer/SchnerrSauer.C:113: error: 'interface' was not declared in this scope
make: *** [Make/linux64GccDPOpt/SchnerrSauer.o] Error 1
please help me,and tell me ,How to correct this problem???

[nimasam]

dear vahid, instead of multiple posts of the same question, in different place, open new thread! with appropriate title,ask your question there!

it is an inappropriate to ask non-related issue under a post!
and about your problem, i told you here (Adding a new term in momentum equation of Interfoam).
you need a background of C++, look at here:
"http://cplus.about.com/od/learning1/ss/cppobjects.htm"

[vahid.najafi]

ok.mr nima.
Thanks

[hz283]

Hi pge_cfd,

I am reading your thread and but this solution work well for my testing. But I did not run the following command as you mentioned:

ar rc myPbeLib.a fortfunc.o other.o object.o files.o // specific to Linux system

The coupled code can run well (but this is my simple test and so there is no complex in the fortran code). Could you please tell me what is the role for the
creation of the library?

Thanks.

Originally Posted by pbe_cfd

So, I came up with a solution to my prototype problem. Calling an external Fortran90 subroutine from an OpenFoam application.

STEPS
1) Declare an interface C function in the main file or in .H file and include it

extern "C"
{
void fortfunc_(int *ii, double *ff); // underscore is important (gfortran linux)!
}

When mixing Fortran with C++, name mangling must be prevented. You may use "objdump -t <file>" file: object or library files. In this example the subroutine named as fortfunc in the Fortran 90 code. But it's referred as fortfunc_

2) Compile your Fortran code and obtain object files "*.o". You may link the object files or you may create a static library form object files as follows,

ar rc myPbeLib.a fortfunc.o other.o object.o files.o // specific to Linux system

3) In your solver folder Make/options append -lgfortran (runtime fortran libraries) and your object or library files. If your fortran subroutine call some other routines from other libraries, add them as well. (Last case is not tested. )

EXE_LIBS = -lfiniteVolume\
-lgfortran\
$(FOAM_USER_LIBBIN)/fortfunc.o

4) Build your executable with wmake in solver folder.

This approach worked for my prototype problem. However I don't know if this is the best way, most probably NOT. Cyp's suggestion sounds really cool. But, I had a problem with obtaining the shared object files (*.so). I had done it before for a C++ code for boundary conditions. There it had worked very nicely. But with Fortran code I had problems. It doesn't compile the fortran code. If someone can describe how to obtain shared object files for Fortran codes, it will be really cool. Then, we can close this thread with a happy end ;-)

All the best;
Evren