On Wed, May 23, 2012 at 6:08 AM, Klaij, Christiaan <C.Klaij at marin.nl> wrote:
> > > Yes, only the norms. As I understand now, GMRES will compute the > > > norm of the total system residual [U,V,W,P], but I would like the > > > norms of [U], [V], [W] and [P] residuals separated. Based on > > > ex42, I got the impression that I have to use KSPBuildResidual. > > > > > > > Yes, the cost of building the residual (check -log_summary to see how big > > it is) is a reason why people often avoid GMRES if they need the full > > residual. It's directly available with GCR, for example (which is > otherwise > > similar in work to FGMRES) and with most of the non-residual-minimizing > > methods. > > I find GCR to be less robust than FGMRES, Right, it can break down in some cases that FGMRES does not. > but ok, suppose I switch > to GCR, how would I get the residual vector? KSPBuildResidual(), but it's trivially fast with GCR. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20120523/8075eb10/attachment.html>
