#1348 slow convergence monitors in istl iterative solvers
Metadata
Property | Value |
---|---|
Reported by | Arne Morten Kvarving (arne.morten.kvarving@sintef.no) |
Reported at | Sep 5, 2013 14:05 |
Type | Feature Request |
Version | 2.2 |
Operating System | Unspecified / All |
Description
I am in need of a mechanism to detect and handle slow convergence in the ISTL iterative solvers.
just to give an idea of how it could look; throw an exception after k iterations with ratio > alpha, for given k and alpha.
usage scenario: my users cannot be expected to have much know-how about linear solvers. i have an expensive preconditioner which should only be used when necessary (extremely anisotrophic grids). i have heuristics to try to choose one, but i don't want to make these too conservative. thus, sometimes the wrong preconditioner is selected. i need to be able to detect this, and rerun with the expensive one without user interaction.