<div dir="ltr"><div><div>If you don't have any prefix set on the outer KSP, you should be able to simply do<br><br>-mg_levels_ksp_type chebychev<br>-mg_levels_ksp_chebychev_estimate_eigenvalues 0,0.2,0,1.1<br>-mg_levels_est_ksp_type cg<br>
-mg_levels_est_ksp_max_it X<br><br></div>Cheers,<br></div> Dave<br><div><div><br><br><br></div></div></div><div class="gmail_extra"><br><br><div class="gmail_quote">On 23 January 2014 15:49, Garth N. Wells <span dir="ltr"><<a href="mailto:gnw20@cam.ac.uk" target="_blank">gnw20@cam.ac.uk</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="im">On 2014-01-23 14:32, Matthew Knepley wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
On Thu, Jan 23, 2014 at 5:35 AM, Garth N. Wells <<a href="mailto:gnw20@cam.ac.uk" target="_blank">gnw20@cam.ac.uk</a>><br>
wrote:<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Is there are command line option to set the number of iterations<br>
used for the eigenvalue estimation inside the Chebyshev<br>
preconditioner? "-gamg_est_ksp_max_it" does the trick with GAMG,<br>
but I'd also like to change the number of iterations when using<br>
Chebyshev smoothing with ML.<br>
</blockquote>
<br>
In GAMG, we use PETSc itself to do the estimate, and give it the<br>
prefix gamg_est_. In ML, as far as I can see from the code, we use the<br>
internal<br>
estimator, and I don't think they give us this knob.<br>
<br>
</blockquote>
<br></div>
I'm not so interested in the what happens inside ML, but the Chebyshev smoother that PETSc applies. With the default 10 GMRES iterations, the eigenvalue estimate is not good enough for my problem, which leads to the outer CG breaking down with an indefinite operator error.<br>
<br>
I can get it working by changing<br>
<br>
-mg_levels_ksp_chebyshev_<u></u>estimate_eigenvalues<br>
<br>
but I'd rather do a few more iterations to get a better eigenvalue estimate. It would also be nice to be able to switch from GMRES to CG. I can do this with GAMG via "-gamg_est_ksp_type"<span class="HOEnZb"><font color="#888888"><br>
<br>
Garth</font></span><div class="HOEnZb"><div class="h5"><br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Matt<br>
<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Garth<br>
</blockquote>
<br>
--<br>
What most experimenters take for granted before they begin their<br>
experiments is infinitely more interesting than any results to which<br>
their experiments lead.<br>
-- Norbert Wiener<br>
</blockquote>
</div></div></blockquote></div><br></div>