On 12/01/2017 06:50 AM, Timo Heister wrote:
They're always going to be there because we keep constrained nodes in the
linear system.
If we modify the way ConstraintMatrix operates, we could work around
this though:
1. Without rescaling those equations (as we do inside ConstraintMatrix
right now) the spurious EV would all be equal to 1 and so ignoring
them is easier.
Correct. But we re-scale for a good reason, namely so that all entries
in the matrix are of comparable size.
2. One could also zero out the constrained rows after the fact.
But only in one of the two matrices of a generalized eigenvalue problem.
At least one of the two matrices needs to remain invertible.
3. Did we rip out the support for removing the constrained entries
from the matrix completely?
Yes.
W.
--
------------------------------------------------------------------------
Wolfgang Bangerth email: [email protected]
www: http://www.math.colostate.edu/~bangerth/
--
The deal.II project is located at http://www.dealii.org/
For mailing list/forum options, see
https://groups.google.com/d/forum/dealii?hl=en
---
You received this message because you are subscribed to the Google Groups "deal.II User Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
