Dear all, I used "-snes_ksp_ew_conv" to inexactly solve the linear system with Petsc prescribed forcing term BUT I didn't catch any difference or improvement. I just screen out the "rtol" of KSP during SNES iteration and found out It is constant (1e-5) in both exact and inexact cases. Please let me know how I should implement "inexact" method in SNES and what is the relation between KSP tolerances and inexact setup?
Merry Xmas, BehZad -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20111226/617ab97f/attachment-0001.htm>
