From: Robert <libmesh@ro...> - 2014-07-14 16:29:34
I want to use the norm of the energy (scalar product of old residual and
current solution increment) for testing the convergence of the newton
I am now wondering how to implement this into the current NewtonSolver.
I came up with the following ideas:
1) implement a absolute_energy_tolerance directly as it exists for the
residual and step
2) (ab)use LinearSolutionMonitor and change the return value from void
to boolean and defer the actual check to usercode
3) implement a new functional UserConvergenceTest to defer the test
(with this it should be possible to move the functionality up to
DiffSolver and implement it for PetscDiffSolver as well)
Would this feature be useful for anyone else? Any further comments?
Get latest updates about Open Source Projects, Conferences and News.