On Tue, 02 Mar 2010 20:37:02 +0100, Jed Brown <jed at 59A2.org> wrote: > I think these should be run before and after *each* step, not each call > to TSStep.
Note that the distinction between "step" and "stage" is somewhat subtle as far as the user is concerned. From an implementation perspective, doing things at intermediate stages is somewhat tricky since the state is more complex, the error estimate has not been computed, and the whole step may be rejected. I envisioned TSSetPostStep being called immediately after the integrator has accepted the step. However, I could imagine there may be things the user would like to do each time the nonlinear solve converges, therefore a TSSetPostStage might also be desirable. Jed
