PETSc allocation issues and keep_diagonal

Asked by Charl on 2013-04-02

Hello,

I upgrading from FEniCS 1.0 to 1.2, I came across the PETSc issue that has been previously discussed on this forum. When integrating over subdomains, it throws errors about non-preallocated entries. Using keep_diagonal=True on the assemble() function makes these errors go away, but this appears to make my simulation blow up.

Is there some necessary post-processing that I am failing to do, such as re-zeroing certain entries of the solution vector?

cheers

Question information

Language:
English Edit question
Status:
Answered
For:
DOLFIN Edit question
Assignee:
No assignee Edit question
Last query:
2013-04-02
Last reply:
2013-04-05
Jan Blechta (blechta) said : #1

Maybe it is blowing up because you have singular system. Do you try calling
A.ident_zeros()
where A is your matrix before solving the problem. It can possibly make problem regular.

Can you supply a small example demonstrating the problem?

Can you help with this problem?

Provide an answer of your own, or ask Charl for more information if necessary.

To post a message you must log in.