Re: [Help-gsl] Floating point exception in gsl

[caoj]

Quoting Inigo Aldazabal Mensa <address@hidden>:

> El Thursday, 1 de March de 2007 03:41, address@hidden escribió:
> > Hi:
> >
> > My mpi program use ode solver fuction in GSL. This program run OK in
> > Dell EM64T Cluster using mpicc 1.2.6 (gcc 3.2.3), and in IA-64 Linux
> > Cluster using mpicc 1.2.6 (icc 9.1.046).
> >
> > But I get the following problem in Alpha EV7 using cc (HP Tru64 C
> > compiler): MPI process 21963 died from signal 8 (Floating point
> > exception)
> >
> > Does anyone know how to solve it? Thanks.
> >
> > Jun
> 
> Hi,
> 
> I can not assure it. But I recall something similar happening to me, also 
> with a Alpha EV7, and, if I can remember correctly, it was related to the 
> IEEE floating-point arithmetic (see the related chapter in the gsl 
> manual).
> 
> Adding:
> 
> #include <gsl/gsl_ieee_utils.h>
> ...
> gsl_ieee_env_setup()
> 
> to the code solved the problem. This if I am  not mixing up things here, 
> which of course could be, knowing myself, ahemm ;-)
> 
> Iñigo
> 
> >
> > _______________________________________________
> > Help-gsl mailing list
> > address@hidden
> > http://lists.gnu.org/mailman/listinfo/help-gsl
> 
> 
> _______________________________________________
> Help-gsl mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/help-gsl
> 
> 


Hi Inigo:

Thanks for your help.

I add the following into my program:
#include <gsl/gsl_ieee_utils.h>
gsl_ieee_env_setup();

And, I also add the following into my pbs script file(csh shell): 
setenv GSL_IEEE_MODE mask-all

But I stall get the following results:
GSL_IEEE_MODE="mask-all"
MPI process 391678 died from signal 8 (Floating point exception)

Do you know why? Thanks.

Jun