On Wed, Aug 17, 2011 at 07:02, Richard Katz <richard.katz at earth.ox.ac.uk>wrote:
> Is there an example showing the use of these controllers? > Simplest is to run with -ts_type alpha -ts_alpha_adapt. Run with -help |grep ts_alpha_adapt for the parameters. src/ts/examples/tutorials/ex18.c uses this in an isogeometric method for Navier-Stokes-Korteweg > > > I think having these interfaces is really undesirable and that we should > unify it, but the information provided by the error estimates for TSGL are > quite different from most methods (OTOH, they seem to me noisy which limits > their utility, but there are other possible reasons for that, including > "starting methods"). > > Is it possible for the user to push a custom controller? I.e. a call-back > that gets passed the SNES converged reason, the last SNES residual, the > current time-step size, the solution vector, etc. > That is how it was designed, except that the controller would probably pull out the SNES themselves and query the reason themselves (because it's hard to predict every piece of information that they might want to use). > > So now that I have TSBEULER working for my problem, should I try to upgrade > to TSALPHA ? > The source you sent is a DAE, which TSALPHA (not being L-stable) will not be robust for. You should still be able to "try it" by running with -ts_type alpha (or -ts_type theta which is closely related; I just tried alpha and TSInterpolate_Alpha is missing, will fix). -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20110817/2689e884/attachment.html>
