[petsc-users] Seeking help for analyzing non-convergence (KSP)

[abdullahasivas at gmail.com]

Hello Florian,
I usually apply following steps before going to preconditioners;
- Construct a smaller matrix and examine the nonzeros, try to use permutations to get a smaller bandwidth (minimum degree or symrcm).
- Use symmetric diagonal scaling, which usually helps a lot but mostly not enough by itself.
- I have found out that (as a part of my master thesis), ParaSails is a good general purpose preconditioner and you can have symmetric ParaSails preconditioners easily.
I hope this helps. 
Regards,Abdullah Ali Sivas
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20161104/d4682e11/attachment.html>