dolfin team mailing list archive

[Evan Lezar <question125019@xxxxxxxxxxxxxxxxxxxxx>]

Question #125019 on DOLFIN changed:
https://answers.launchpad.net/dolfin/+question/125019

    Status: Answered => Open

Evan Lezar is still having a problem:
Thanks to all who have responded.

One side-note that I would like to make is that I am uncertain whether
discussions like this should be continued through the Question framework
supplied by Launchpad, or whether standard replies to mailing list
emails should suffice. For now, I will continue on in the Question
thread.

Marie:
Regarding the filtering of the eigenpairs. At present I am using FEniCS to generate eigensystems that I am solving with a variety of methods, and it is the performance of these methods that I am interested at present and as such, I have implemented things such that there is no additional information available to the solvers other than the matrices (and perhaps some parameters that help me to choose the shift if a shift-invert process is used). Therefore, I can't really compare coefficients associated with specific DOFs to decide on valid/invalid eigenpairs. 

Anders:
A parameter that allows for the setting of the values inserted on the diagonal would alleviate some of the problem, as I would (as you say) be able to ensure that the eigenvalues do not fall in the region of interest in the spectrum. However, this would only be useful for problems where I have some idea of the resultant spectrum (which in this case I do).

Garth:
As I have mentioned, what I am doing here is simply a hack to get my problems working as soon as possible - I have quite a tight deadline and since the workings of FEniCS are not what my research is focussed on I have not spent much time trying to implement a proper solution. This does sound similar to what is required for the periodic bcs case, but I see that there has not been much discussion related to your blueprint. Essentially, passing a zero-valued DirichletBC to an Assembler should ignore the DOFs associated with the boundary and assemble the reduced matrix. Some time ago there was mention that this could be implemented using restrictions, but I was unable to give it more attention.

Thanks again.

-- 
You received this question notification because you are a member of
DOLFIN Team, which is an answer contact for DOLFIN.