I'm currently getting very long propagation times when attempting to use the Time Stepping propagators in Petsc 3.2, and in the interest of speeding things up, I'm curious how I can reduce the accuracy (or see what it is in the first place) to speed it up.

  • $\begingroup$ Are you solving a linear or nonlinear system of equations at each time step? $\endgroup$ – Aron Ahmadia May 2 '12 at 6:22
  • $\begingroup$ Linear, is TS for linear propagation? or just for non-linear propagation. $\endgroup$ – Andrew Spott May 2 '12 at 19:46

nonlinear systems of equations

In general, the PETSc TS propagators solve a nonlinear system of equations at each time step, so you would need to access the SNES object and modify its accuracy tolerance there. See the documentation on SNESSetFromOptions for a list of the command line options available to you for adjusting the solver tolerances.

linear system of equations

PETSc doesn't expose the KSP object belonging to TS in the same way at the command-line, @JedBrown or @MattKnepley may be able to comment on why. Programmatically, you can:

TSGetKSP(ts, &ksp);

KSPSetTolerances(ksp, rtol, abstol, dtol, maxits;

| cite | improve this answer | |
  • $\begingroup$ What about for a linear system of equations? $\endgroup$ – Andrew Spott May 2 '12 at 20:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.