Re: [Help-gsl] convergence check in odeiv_rk4imp ?

[Tuomo Keskitalo]

Hello,

My guess is that rk4imp can't make Newton iteration converge when you 
use a large step size. You should be able to use a small fixed step size 
with any stepper, if that is really needed.

If fixed step sizes are not mandatory, I suggest you use the driver or 
evolve functions to traverse from t0 to t1. Please see the example codes 
in the documentation.

Regards,
Tuomo

On 01/06/2010 01:34 AM, Forest Yang wrote:
> Hi
>
>   I was trying to evaluate the rk4imp solver at large step, but not
> being able to do it on a fixed stepsize way.
> Can I truely do rk4imp with a fixed time step ?
>
> I used the following two lines, and also put a print line in f, the
> time printed out have many steps between [t,t+h], Is it normal in the
> internal stage when solving the fixed point or still an adaptive
> stepsize control ?
>
>
>  GSL_ODEIV_FN_EVAL(&sys, t, y, dydt_in);
>  int status = odeiv2_step_apply (s, t, h, y, y_err, dydt_in, dydt_out, &sys);
>
> Lingyun
>
>
>
> On Thu, Dec 31, 2009 at 1:24 AM, Tuomo Keskitalo <address@hidden> wrote:
> > Hello,
> >
> > take a look at my ode-initval2 extension to GSL, it's rk4imp iterates to
> > convergence. Note: I'm currently working on finetuning the ode-initval2
> > framework, it will soon change slightly from version 0.9.
> >
> > http://iki.fi/tuomo.keskitalo/gsl/ode-initval2/
> >
> > On 12/29/2009 08:35 PM, Forest Yang wrote:
> >
> > > Hi All,
> > >
> > >    Is there any plan for implementing convergence check in implicit
> > > RK4 method ?
> > > It seems the 2 stages Gauss points are only iterate 3 steps, without
> > > checking whether it is converged or not.
> > > Another possible way could be using the Jacobian to solve for "Y_i",
> > > the fixed point.
> > >
> > >   The reason I care about this is, Implicit Gauss RK4 is symplectic,
> > > which for Hamiltonian Dynamics simulation is a very important
> > > property,
> > > it has better conservation of the first integral.
> > >
> > >
> > > Best regards,
> > > Forest.
> > >
> > >
> > > _______________________________________________
> > > Help-gsl mailing list
> > > address@hidden
> > > http://lists.gnu.org/mailman/listinfo/help-gsl
> >
> > --
> > address@hidden
> > http://iki.fi/tuomo.keskitalo

--
address@hidden
http://iki.fi/tuomo.keskitalo