Adaptive SE Eigenvalue

[Srihari Sritharan]

Hey,

It's been a long time since I've been here, but it's summer again and
I'm trying to pick up some work again.

I see your Schrodinger-Refine towards center example program in H3D,
and I would like to make an hp-adaptive version so I can do some
complex problems. Whats the best example/benchmark/etc source code I
should look at as guidance for creative an adaptive version of the
SchroEq solver?

Thanks,
Sri

[Pavel Solin]

Hi Sri,

  you can look, for example, at the Fichera corner example in H3D.

There are two adaptive eigenvalue examples in Hermes2D, 

in the tutorial part P05.

Pavel

--
hermes3d mailing list
herm...@googlegroups.com
http://groups.google.com/group/hermes3d?hl=en

--
Pavel Solin
University of Nevada, Reno
Home page: http://hpfem.org/~pavel
FEMTEC 2011: http://hpfem.org/events/femtec-2011/
Hermes: http://hpfem.org/
FEMhub: http://femhub.org/

[Srihari Sritharan]

Alright, I wrote it, but I get scrambled solutions and NaN% errors.
Yay.


I wrote it, it works (i think?). I stop the simulation before it
actually does the adapting, because it's just taking too long.

Do you have a guess as to how long it would take to adapt a mesh with
3000 DOFs initially? More than 8 minutes? Every other step took no
more than 2-3 minutes.

Thanks,
Sri

p.s. maybe I'll just run it overnight.

On Jun 13, 3:34 pm, Pavel Solin <so...@unr.edu> wrote:
> Hi Sri,
>   you can look, for example, at the Fichera corner example in H3D.
> There are two adaptive eigenvalue examples in Hermes2D,
> in the tutorial part P05.
>
> Pavel
>

[Pavel Solin]

On Wed, Jun 15, 2011 at 5:15 PM, Srihari Sritharan <sriha...@gmail.com> wrote:

Alright, I wrote it, but I get scrambled solutions and NaN% errors.
Yay.


I wrote it, it works (i think?). I stop the simulation before it
actually does the adapting, because it's just taking too long.

Do you have a guess as to how long it would take to adapt a mesh with
3000 DOFs initially? More than 8 minutes? Every other step took no
more than 2-3 minutes.

I would have to see more details before I can say anything. 

What is the criterion you use to guide the hp-adaptivity?

Is is hp- or just h-adaptivity? How many eigenfunctions 

do you have on the 3000 dof mesh?

Pavel  

[Srihari Sritharan]

Ah, turns out I was just not being patient enough, it adapts fine,
gives me hydrogen eigenvalues nicely.

However, a more troubling problem I have now is for a more difficult
domain (spherical annulus), and the VTK files take a long long time to
export.
The VTK export of the polynomial order takes a few minutes, but VTK
solution export is taking well over 3-4 hours (I didnt even let it
finish, so it could just be frozen) for a problem with maybe
6000-11000 dof. I don't get why, the same number of dof but with
hydrogen or harmonic oscillator exports significantly faster. Does the
complexity of the problem affect VTK export at all?

Sri

[Pavel Solin]

High polynomial degrees in the mesh cause many linear tetrahedra 

to be created in the visualization procedure. But it is suspicious that 

visualization should take so much time.

Pavel 

Sri

--
hermes3d mailing list
herm...@googlegroups.com
http://groups.google.com/group/hermes3d?hl=en

--
Pavel Solin
University of Nevada, Reno
Home page: http://hpfem.org/~pavel

[Srihari Sritharan]

Ok, problem apparently solved, I compiled hermes on the supercomputer,
and VTK no longer takes a century to export.

On Jun 18, 5:42 pm, Pavel Solin <so...@unr.edu> wrote:

[Pavel Solin]

On Mon, Jun 27, 2011 at 3:01 PM, Srihari Sritharan <sriha...@gmail.com> wrote:

Ok, problem apparently solved, I compiled hermes on the supercomputer,
and VTK no longer takes a century to export.

Great!

We are getting rid of Python in the basic library, so 

we will be replacing Pysparse with some C++ eigensolver.

Do you have any preference?

Pavel 

[Srihari Sritharan]

Actually, I'm not at all familiar with any C++ eigensolvers :) just
pysparse, so I guess anything works

On Jun 27, 4:54 pm, Pavel Solin <so...@unr.edu> wrote: