[scilab-Users] fsolve with external fortran subroutine

[Eduardo Torrecillas]

Hello Serge,

Thanks for your reply, the solution is really clever. Unfortunately, I am
having some troubles on implementing it. Please see below:

I tried using global variables using modules, for example, the file
mod_01.f90:

module mod01
implicit none
integer, parameter,public :: dp=8
real(kind=dp),parameter,public::pi=ATAN(1.0_dp)*4.0_dp
real(kind=dp),public::Apub

end module mod01


And then, tried using it on another f90 subroutine,

subroutine fprepara_var_equil(a,dummyout)
use mod01
implicit none
real(kind=dp),intent(in)::a
real(kind=dp),intent(out)::dummysai

Apub=a
dummysai=0.0_dp

end subroutine fprepara_var_equil

Everything goes fine during compilation, but when trying to link libraries
"exec floader.sce", I get the following message:

undefined symbol: __mod01_MOD_apub

When trying to use the same strategy using pure fortran, the code compiles
and runs perfectly. Any suggestion?

Regards,

On Mon, Jul 16, 2012 at 12:15 PM, Serge Steer <Serge.Steer at inria.fr> wrote:

> Le 16/07/2012 16:04, Eduardo Torrecillas a écrit :
>
>
>> Hi everybody,
>>
>> I am trying to use fsolve with a fortran subroutine as argument. For
>> example, consider the following calling sequence:
>>
>> X0=[0 0]
>>
>> tol=1E-15
>>
>> [X,v]=fsolve(X0,'fequil_**fortran',tol)
>>
>> In this case 'fequil_fortran' is my external fortran subroutine, given
>> below:
>>
>> ------------------------------**------------------------------**--------
>>
>> subroutine fequil_fortran(n,x,v,iflag)
>>
>> implicit none
>>
>> integer,intent(in)::n,iflag
>>
>> real(kind=8),dimension(n),**intent(in)::x
>>
>> real(kind=8),dimension(n),**intent(out)::vreal(kind=8)::a,**b
>>
>>  a=1.0_dp
>>
>> b=2.0_dp
>>
>> v(1)=a*x(1)+b*x(2)-1.0_dp
>>
>> v(2)=2.0_dp*a*x(1)-b*x(2)-5.0_**dp
>>
>> end subroutine fequil_fortran
>>
>> ------------------------------**------------------------------**---------
>>
>> It worked fine and fsolve found the solution. Now, suppose I have to pass
>> arguments to the fortran subroutine, in addition to the default
>> (n,x,v,iflag). For instance, imagine I had to pass the 'a' and 'b'
>> variables from Scilab to the fortran subroutine in the code above. I have
>> already performed something like this but with "optim". In fact, "optim"
>> gives you the possibility to pass a list of integer, real or double
>> variables to the external code. Is there anything similar to be used with
>> fsolve?
>>
>>  The answer is no, the underlying code does not handle extra arguments
> for the external function.
>
> It is however possible to declare a and b as elements of a common and to
> set their value using another fortran code called using the function call
>
> something like
> subroutine fequil_settings(a,b)
> ...
> end
>
> Also add the common declaration in fequil_fortran, compile and link both
> function with Scilab
>
> And then in Scilab you can do
> a=...
> b=...
> call("fequil_settings",a,1,"d"**,b,2,"d","out",[1 1],"d");
> [X,v]=fsolve(X0,'fequil_**fortran',tol)
>
>  Regards,
>>
>> --
>> Eduardo Torrecillas
>> AER-09
>>
>
>
>
> --
> To unsubscribe from this mailing-list, please send an empty mail to
> users-unsubscribe at lists.**scilab.org <users-unsubscribe at lists.scilab.org>
> To check the archives of this mailing list, see
> http://mailinglists.scilab.**org/ <http://mailinglists.scilab.org/>
>
>


-- 
Eduardo Torrecillas
AER-09
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scilab.org/pipermail/users/attachments/20120718/b338df75/attachment.htm>