On Sat, Jul 7, 2012 at 11:36 AM, Ferenc Tasnadi <[email protected]> wrote:

> Dear all,
>
> I managed to install fipy parallel on supercomputers. The
> examples/Diffusion/mesh20x20.py runs fine. I use
> phi.getGlobalValue()
> to get phi at each time and what I write in ascii "x-y-phi" file. The
> figures are the same you see on fipy website.
>

Very good.


> However, if I run the examples/cahnHilliard/mesh2D.py parallel; the script
> runs, but phi with using
> phi.getGlobalValue()
> does not really change in time. The script with pysparse on 1CPU produce
> the figures with nice spinodal
> decomposition (see fipy website).
>

Hi Ferenc,

Change the default nx and ny to 50 in examples/cahnHilliard/mesh2D.py and
run

 $ python examples/cahnHilliard/mesh2D.py --trilinos

Does it evolve (separate into regions of 0 and 1 eventually)? Probably not.

I think there is a problem with the default solver when running this
example. Try switching the solver to LinearLUSolver. Does it evolve now?
Does it also work in parallel after the switch?

 $ mpirun -np 2 python examples/cahnHilliard/mesh2D.py --trilinos

Cheers

-- 
Daniel Wheeler
_______________________________________________
fipy mailing list
[email protected]
http://www.ctcms.nist.gov/fipy
  [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]

Reply via email to