It seems that newer version of SUNDIALS has a different API. I have added
autoconf checks and preprocessor directives to be compatible with both new and
old versions of SUNDIALS. These changes will be released in the next version.
Thank you for your feedback and patch !
With the best regards,
Konstantin.
On Friday, June 29, 2012 12:17:13 pm Aai wrote:
> Hi,
>
> I tried to make this work on platform SUSE Linux 12.1. I (only)
> encountered the following error(s) after installing the necessary
>
> packages (e.g. sundials):
> > yacts.cpp: In memberfunctie ‘bool yacts::initTrajectory(std::string)’:
> > yacts.cpp:148:8: fout: ‘CV_SS’ was not declared in this scope
> > yacts.cpp:148:38: fout: ‘CVodeMalloc’ was not declared in this scope
> > yacts.cpp:150:39: fout: ‘CVodeSetFdata’ was not declared in this scope
> > make[3]: *** [yacts.o] Fout 1
> > make[3]: Map '/home/aai/jplus-0.4.2/src' wordt verlaten
> > make[2]: *** [all] Fout 2
> > make[2]: Map '/home/aai/jplus-0.4.2/src' wordt verlaten
> > make[1]: *** [all-recursive] Fout 1
> > make[1]: Map '/home/aai/jplus-0.4.2' wordt verlaten
> > make: *** [all] Fout 2
>
> I changed function
>
> bool yacts::initTrajectory(string prefix)
>
> of file yacts.cc into the following
>
> > bool yacts::initTrajectory(string prefix) {
> > string trjname=getTrajectoryFilenameBase(prefix)+".trj";
> > trjfile::open(trjname);
> > if (!trjfile::hasNextFrame()) {
> > // file is new, we need to create and save the initial frame
> > saveFrame();
> > } else {
> > // read the last frame
> > trjfile::toFrame(-1);
> > loadFrame();
> > };
> >
> > // now we have the first frame inside J+
> > // pull it out and feed to the solver
> > jarray_of_type<double> jtnext=get("T"); // compute next time
> > jarray_of_type<double> jt=jengine::get("T"); // first version of T,
> > curr time
> > assert(jt.rank()==0); // assert the time is scalar
> >
> > jarray_of_type<double> js=get("S");
> >
> > nvs = N_VNew_Serial(neq); assert(nvs);
> > for(int i=0;i<js.size();i++) NV_Ith_S(((N_Vector)nvs),i)=js[i];
> > cvode_mem = CVodeCreate(CV_BDF, CV_NEWTON);
> > int flag;
> > realtype abstol=1.0e-5; /*abstol*/
>
> CHANGES
>
> > flag = CVodeInit(cvode_mem, rhs, jt[0], (N_Vector)nvs); assert(!flag);
> > flag = CVodeSStolerances(cvode_mem,1.0e-3 ,abstol);
> > // pass pointer to "this" object for rhs
> > flag = CVodeSetUserData(cvode_mem, this);
>
> END OF CHANGES
>
> > // specify the solver and subspace dimension
> > flag = CVSpgmr(cvode_mem,PREC_NONE,300);
> >
> > // now we are ready to integrate
> > };
>
> Then 'make install' ran without error(s).
>
> Trying out two examples
>
> > yacts ./examples/0_constant/constant
>
> ok
>
> > yacts ./examples/1_pendulum/pendulum | gnuplot
>
> Ok
>
> On 28-06-12 18:51, Konstantin L. Metlov wrote:
> > For some years I was thinking to do some large-scale numerical
> > mathematics with J. It was clear to me from the very beginning that J
> > can be an ideal language for this, very concise and efficient. But many
> > important ingredients were lacking, like routines for numerical
> > integration, Fourier transforms, ODE solving, vector field
> > visualization. There were some, of course, but scattered and
> > unsupported.
> >
> > This time, when I needed to integrate Landau-Lifshitz-Gilbert equation
> > from magnetism on finite element mesh (to simulate dynamics of
> > nano-scale magnets), I decided to put together some bits and pieces
> > (including my own) in a form of simulation language I called J&+ ("with
> > J added", jplus or J+) and an ODE solver yacts. You can find the details
> > here:
> >
> > http://www.fti.dn.ua/~metlov/jplus/
> >
> > Have a look !
> >
> > With the best regards,
> >
> > Konstantin.
> >
> > ----------------------------------------------------------------------
> > For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
