Dear All
I wonder if it is possible to use functions rather than arrays as an input when call compiled dlls in GAUSS?
5 Answers
0
accepted
Many optimization codes have a reverse communication interface where the compiled code (FORTRAN/C/C++) will return a specific return value when it would like the objective function to be evaluated along with the values at which it would like them evaluated at.
Below is a simple example of the concept. For this example, we will assume that we have a FORTRAN function that will assign ret to be equal to 1, assign the values at which it would like the function evaluated to parmout and then return whenever it would like a function evaluation. You could then set up your GAUSS code something like this:
ret = 1; parmin = zeros(4, 1); parmout = zeros(4, 1); ans = zeros(4,1); do while (ret); dllcall fortranOptRoutine(parmin, parmout, ans, ret); if (ret == 1); parmin = myFunc(parmout); endif; endo; proc (1) = myFunc(x); retp(x*x); endp;
0
Functions that are called with dllcall can only accept GAUSS variables as inputs.
Are you wanting to save the step of assigning the output of a GAUSS procedure to a variable before passing it to the function you are calling with dllcall? Or are you wanting to try to pass a GAUSS function in to a compiled (C/C++/Fortran) function that takes a function pointer as an argument?
If you provide a little more information, I may be able to help you accomplish your end goal.
0
Thanks.
I want to know how to pass a GAUSS function into a compiled Fortran function that takes a function pointer as an argument.
0
You cannot directly pass a GAUSS procedure into a Fortran function that takes a function pointer as an argument. There may be a work around. Could you provide some information about the Fortan code? Do you have the source code?
0
I do have the FORTRAN source code which does optimization, in this optimization subroutine, there is a call to another FORTRAN subroutine, say FCN(N, X, F):
subroutine FCN(N,X,F) double precision X, F F = x*x return end
which defines the objective function.
I want to compile the optimization subroutine and call it within GAUSS, while the definition of the objective function FCN is written in GAUSS rather than Fortran.
Your Answer
5 Answers
Many optimization codes have a reverse communication interface where the compiled code (FORTRAN/C/C++) will return a specific return value when it would like the objective function to be evaluated along with the values at which it would like them evaluated at.
Below is a simple example of the concept. For this example, we will assume that we have a FORTRAN function that will assign ret to be equal to 1, assign the values at which it would like the function evaluated to parmout and then return whenever it would like a function evaluation. You could then set up your GAUSS code something like this:
ret = 1; parmin = zeros(4, 1); parmout = zeros(4, 1); ans = zeros(4,1); do while (ret); dllcall fortranOptRoutine(parmin, parmout, ans, ret); if (ret == 1); parmin = myFunc(parmout); endif; endo; proc (1) = myFunc(x); retp(x*x); endp;
Functions that are called with dllcall can only accept GAUSS variables as inputs.
Are you wanting to save the step of assigning the output of a GAUSS procedure to a variable before passing it to the function you are calling with dllcall? Or are you wanting to try to pass a GAUSS function in to a compiled (C/C++/Fortran) function that takes a function pointer as an argument?
If you provide a little more information, I may be able to help you accomplish your end goal.
Thanks.
I want to know how to pass a GAUSS function into a compiled Fortran function that takes a function pointer as an argument.
You cannot directly pass a GAUSS procedure into a Fortran function that takes a function pointer as an argument. There may be a work around. Could you provide some information about the Fortan code? Do you have the source code?
I do have the FORTRAN source code which does optimization, in this optimization subroutine, there is a call to another FORTRAN subroutine, say FCN(N, X, F):
subroutine FCN(N,X,F) double precision X, F F = x*x return end
which defines the objective function.
I want to compile the optimization subroutine and call it within GAUSS, while the definition of the objective function FCN is written in GAUSS rather than Fortran.