It's strange that the iteration breaks at step 0. If I try right
preconditioning, it breaks at step 1 with nan. Besides, if there is only on
multigrid level, the problem is solved correctly with 1 iteration, so matrices
should be correct. The coarse solution is good enough for smoothing.
The question is, in which case may a CG or GMRES solver breaks at step 0 with
nan?
NaN errors, like segmentation faults, are relatively easy to find because you
know that every NaN you get is wrong. So you check that the matrix and right
hand side and solution vector you give to the solver don't have any NaNs. If
they don't, and you get NaNs out, then the problem likely lies with the
preconditioner, so you single step through the solver around the place where
the preconditioner is applied and check in which operation the NaN appears.
In your case, you seem to have already found out that it's the Chebyshev
preconditioner. The question is why. Is it because the matrix the
preconditioner uses has NaNs in it? Is it because the preconditioner makes an
assumption that it can divide by certain matrix entries, but they are zero?
Single stepping in a debugger will give you the answer :-)
Best
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].
To view this discussion on the web visit
https://groups.google.com/d/msgid/dealii/b46276e6-5427-16ef-5f56-846a56cf216f%40colostate.edu.
