Can you remind me how this problem occurred? I should make it a switch in the code because it is not super cheap to do and you are the only ones to have had this problem.
Can you please send my the output of a bad run with -pc_gamg_verbose 2 (again?). I suspect that there is something wrong with your problem or there is something that should be fixed in PETSc. This solution just papers over the problem. Mark On Thu, Sep 18, 2014 at 6:39 AM, Rhodri Davies <[email protected]> wrote: > Hi guys, > > Have you had any further thoughts on merging the gamg-zerod branch into > master? Any update would be appreciated. > > Best wishes, > > Rhodri > > > > --------------------------------------------------------------------------------------------------- > Research Fellow > Research School of Earth Sciences > The Australian National University > Canberra, ACT, 0200, Australia > > T: +61 (0)2 6125 3643 > E: [email protected] > W: http://rses.anu.edu.au/people/dr-rhodri-davies > > --------------------------------------------------------------------------------------------------- > > > > On 23 Aug 2014, at 06:27, Mark Adams <[email protected]> wrote: > > > > > On Thu, Jul 31, 2014 at 10:44 AM, Stephan Kramer <[email protected]> > wrote: > >> An update on this. We have been using mark/gamg-zerod branch and it fixes >> for us the issue with zero diagonals in the coarsened operators (making the >> sor smoother, but bizarrely not the jacobi smoother fail). We in fact have >> some cases where cheby+jacobi does not converge (indefinite pc), but >> cheby+sor (with the mark/gamg-zerod branch) works well, so we'd be very >> much interested in getting this (or something similar) merged in master. > > > Jed: this has been a pull request for 4+ months (#162). > > >
