help-gsl
[Top][All Lists]
Advanced
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Help-gsl] problem

From: Tommy Nordgren
Subject: Re: [Help-gsl] problem
Date: Thu, 25 Aug 2005 21:30:52 +0200
I've looked at the function It have about 90 independent variables, and contains some nonelinear terms. The problem is probably stiff. That' when a diff eqn have particular solutions thats varying on vastly differnt scale. When that happens, Runge-Kutta have to reduce the step size to the shortest scale on which any 
particular solution vary, or suffer a disastrous loss of precision.
Stiff equations require special methods for stepping, such as Buerlich-Stoer. There is an implementation of it in GSL 
Aug 24, 2005 kl. 7:55 PM skrev address@hidden:


Thanks a lot. Attached is my function.

Jun Cao



Quoting Pau Cervera Badia <address@hidden>:



I think it means that gsl_odeiv_evolve_apply is trying to adjust h to
obtain the prescribed presicion in the integration of the equations, but 
the step-size value seems too large to obtain that precision, so the
routine chooses a smaller step (see [1]). It seems like if one or more 
of your variables or derivatives is unboundedly growing, so your
solutions are diverging, but I'm not sure if it is the only reason
because I don't know you're equations.

Maybe you can post your *func* function and we can take a look.

[1] http://www.gnu.org/software/gsl/manual/gsl-ref_25.html#SEC384

address@hidden wrote:



Hi Pau Cervera Badia:
I check h. I found that h is becoming smaller and smaller and finally get to
0.0. What does it mean? Does it mean my program is wrong? Thanks again. 

Jun Cao


Quoting Pau Cervera Badia <address@hidden>:





Maybe is changing only a little.
Can you monitor the h values, something as

while (t < t1) {

   int status = gsl_odeiv_evolve_apply(e,c,s,&sys,&t,t1,&h,y);

   printf("%.5e %.5e %.5e\n", t, t1, h);

   if (status != GSL_SUCCESS)
      break;

}

will do. If h is becoming smaller and smaller, something wrong is


happening


(maybe your ecuations are diverging somewhere).


address@hidden wrote:





Hi Pau Cervera Badia:

Thank you for your response. I check my program again. It stuck in




t=5.790250,
t1=6.467000. After gsl_odeiv_evolve_apply, t always is 5.790250 and does 




not
change. So the program stuck in the "while" loop. I try to change h to 


1e-2






and
1e-10. I get the same result. Do you have anoter idea? Thanks again. 

Jun Cao


Quoting Pau Cervera Badia <address@hidden>:
gsl_odeiv_evolve_apply(e,c,s,&sys,&t,t1,&h,y) will advance from time t to the next integration time with an optimum step-size, provided that the next time is less than t1. Otherwise it will integrate the system to
t1 exactly. So after some iterations the while condition will be false.
You can check that gsl_odeiv_evolve_apply is running properly with: 

while (t < t1) {

  int status = gsl_odeiv_evolve_apply(e,c,s,&sys,&t,t1,&h,y);
printf("%.5e %.5e\n", t, t1); // this should print t less than t1 

  if (status != GSL_SUCCESS)
     break;

}
printf("%.5e %.5e\n",t,t1); // this should print t equal to t1

If I'm not missing something, your program should work. Maybe the
problem is that t is always less than t1 for some other reason (maybe 
the routine couldn't reach the prescribed precision and
gsl_odeiv_evolve_apply is trying to reduce the step-size more and more). 





Maybe you can check the h values.

Maybe something will be of any help.

address@hidden wrote:







Hi:
I use ordinary differential equations from GSL. The part of my program 


is








like






example shown in the document.

const gsl_odeiv_step_type * T
 = gsl_odeiv_step_rk8pd;

gsl_odeiv_step * s
 = gsl_odeiv_step_alloc (T, 9);
gsl_odeiv_control * c
 = gsl_odeiv_control_y_new (1e-6, 0.0);
gsl_odeiv_evolve * e
 = gsl_odeiv_evolve_alloc (9);

double mu = 10;
gsl_odeiv_system sys = {func, NULL, 9, &mu};

double t = 0.0;
double h = 1e-6;

for (i = 1; i <no_of_data_points; i++) {
  t1 = mytable[i];
  while (t < t1)
  {
     int status = gsl_odeiv_evolve_apply (e, c, s,
                                        &sys,
                                        &t, t1,
                                        &h, y);

   if (status != GSL_SUCCESS)
       break;
   }
   ...
}
gsl_odeiv_evolve_free (e);
gsl_odeiv_control_free (c);
gsl_odeiv_step_free (s);
The weird thing is that at some i, the program stuck inside of "while" 




loop










and
never come out. gsl_odeiv_evolve_apply will give t=t, so t is always < 




t1.










As
result the program never stop. Dose anyone know how to solve this 


problem?




Thanks.

Jun Cao



_______________________________________________
Help-gsl mailing list
address@hidden
http://lists.gnu.org/mailman/listinfo/help-gsl










--
Pau Cervera i Badia (e-mail address@hidden)
{
Departament de Física Fonamental Martí i Franqués, 1 Universitat de Barcelona Planta 3, despatx 346 bis 08028 Barcelona tel: +34 934 921 155 Spain
"To err is human, but to really foul things up requires a computer." 
}


















--
Pau Cervera i Badia (e-mail address@hidden)
{
Departament de Física Fonamental Martí i Franqués, 1 Universitat de Barcelona Planta 3, despatx 346 bis 08028 Barcelona tel: +34 934 921 155 Spain
"To err is human, but to really foul things up requires a computer." 
}

















--
Pau Cervera i Badia (e-mail address@hidden)
{
Departament de Física Fonamental Martí i Franqués, 1 Universitat de Barcelona Planta 3, despatx 346 bis 08028 Barcelona tel: +34 934 921 155 Spain
"To err is human, but to really foul things up requires a computer." 
}








<ode_fcn.c>
_______________________________________________
Help-gsl mailing list
address@hidden
http://lists.gnu.org/mailman/listinfo/help-gsl



"Home is not where you are born, but where your heart finds peace" -
Tommy Nordgren, "The dying old crone"
reply via email to
[Prev in Thread] Current Thread [Next in Thread]