DOLFIN

Dolfin 1.1.0 and scitools

Asked by Bur

Hi,

I've updated Dolfin this morning on my ubuntu 12.04 distro. By now I've got weird results (more exactly strange pictures) when using the scitools utility as mentionned in the Fenics Book.

To confirm this behaviour I've ran the vcp2D.py tutorial from http://fenicsproject.org/_static/tutorial/fenics_tutorial_examples.tar.gz.

Maybe there is something in the numbering of cells/vertices...?

Is this a bug, or simply a new way I have to adapt? I'd be glad if you can tell me. As I really need of that for my PhD I've downgraded my Dolfin installation back to 1.0 (via the ppa)... Unfortunately I don't remember the version of scitools I had when the 'bug' appeared.

Thanks in advance,
Nicolas.

Question information

Language:
English Edit question
Status:
Solved
For:
DOLFIN Edit question
Assignee:
No assignee Edit question
Solved by:
Johan Hake
Solved:
Last query:
Last reply:
Revision history for this message
Best Johan Hake (johan-hake) said : 		#1

This is expected behavior. dofs are now reordered so they do not
corresponds to any geometrical structurs such as Cells or Vertices. If
you are running serial runs you can turn this reordering off.

from dolfin import *
parameters.reorder_dofs_serial = False

Johan

On 01/15/2013 12:11 PM, Bur wrote:
> New question #219220 on DOLFIN:
> https://answers.launchpad.net/dolfin/+question/219220
>
> Hi,
>
> I've updated Dolfin this morning on my ubuntu 12.04 distro. By now I've got weird results (more exactly strange pictures) when using the scitools utility as mentionned in the Fenics Book.
>
> To confirm this behaviour I've ran the vcp2D.py tutorial from http://fenicsproject.org/_static/tutorial/fenics_tutorial_examples.tar.gz.
>
> Maybe there is something in the numbering of cells/vertices...?
>
> Is this a bug, or simply a new way I have to adapt? I'd be glad if you can tell me. As I really need of that for my PhD I've downgraded my Dolfin installation back to 1.0 (via the ppa)... Unfortunately I don't remember the version of scitools I had when the 'bug' appeared.
>
> Thanks in advance,
> Nicolas.
>

Revision history for this message
Bur (n-bur) said : 		#2

Thank you Johan for the quick answer.

Just to continue on this topic:
What do you mean by 'running serial runs'?
The turning off of the parameter will always work? It's just to improve the computation time, isnt't it?

With the reordering, does it exist a simple way to get the values of a solution on (part of) the boundary in order to plot them (what was (almost) easy with scitool)?

Best regards,
Nicolas.

Revision history for this message
Bur (n-bur) said : 		#3

Thanks Johan Hake, that solved my question.

Revision history for this message
Johan Hake (johan-hake) said : 		#4

On 01/15/2013 02:01 PM, Bur wrote:
> Question #219220 on DOLFIN changed:
> https://answers.launchpad.net/dolfin/+question/219220
>
> Bur posted a new comment:
> Thank you Johan for the quick answer.
>
> Just to continue on this topic:
> What do you mean by 'running serial runs'?

Not in parallel using MPI.

> The turning off of the parameter will always work? It's just to improve the computation time, isnt't it?

For some iterative algorithms the time is reduced with reordered dofs.

> With the reordering, does it exist a simple way to get the values of a
> solution on (part of) the boundary in order to plot them (what was
> (almost) easy with scitool)?

You can always use the dofmap of a FunctionSpace to tabulate how dofs
correlates with geometric properities.

Johan

> Best regards,
> Nicolas.
>

Revision history for this message
Bur (n-bur) said : 		#5

Very clear, once again :)
Thanks a lot.