[Archive] Cholesky DECOMPOSITION

Message by julien:
Initially posted the Fri Nov 26, 2004 3:34 pm by markelov, forwarded by juju.
when I slightly changed a size of a geometry, SPIS solver complains each time step
WARNING : Error in Cholesky DECOMPOSITION of preconditionnig ) = 5413
This is a warning or error? Can I neglect it?
Thank you. Regards, Gennady

Message by julien:
Initially posted Mon Dec 06, 2004 8:41 am by roussel, forwarded by juju.
You can neglect it if it does not show up again (usually it happens only the first time because of a bad definition of convergence criterion at first solver call, when it starts from 0 values) JF
Jean-François Roussel

Message by julien:
Initially posted Thu Dec 09, 2004 2:57 pm by markelov, forwarded by juju.
Thank you for the reply but I have it _each time step_
Regards, Gennady

Message by Nuno Loureiro:
I am having the same problem (at each time step) and I can’t understand why this happens. Can anyone explain what is the cause of it?

Message by Jean-Francois Roussel:
Preconditioning is used to improve the convergence speed of Poisson solver.
In some situations (bad quality mesh e.g.) the Cholesky matrix decomposition used for that, fails.
You don’t have to worry too much if that happens, the only consequence is on the computation time.