On Wed, Dec 9, 2009 at 12:49 PM, Dominik Szczerba <span dir="ltr"><<a href="mailto:dominik@itis.ethz.ch">dominik@itis.ethz.ch</a>></span> wrote:<br><div class="gmail_quote"><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Excellent point. Where it broke the reason was:<br>
<br>
Starting KSPSolve<br>
0 KSP Residual norm 7.841852910113e-04<br>
1 KSP Residual norm 8.927576540999e-05<br>
2 KSP Residual norm 1.694401713084e-05<br>
3 KSP Residual norm 5.069425875318e-06<br>
Linear solve converged due to CONVERGED_RTOL iterations 3<br>
<br>
which I assume means that the norm became exactly 0<br></blockquote><div><br>I don't think so. It would have printed that for the residual.<br> </div><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
HOWEVER: putting -ksp_rtol -1 still does not eliminate the problem. Any further suggestions?<br></blockquote><div><br>You -ksp_view to see the tolerances used.<br><br> Matt<br> </div><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Dominik<br>
<br>
Matthew Knepley wrote:<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><div class="im">
Did you use -ksp_converged_reason to determine why PETSc stopped iterating?<br>
<br>
Matt<br>
<br></div><div><div></div><div class="h5">
On Wed, Dec 9, 2009 at 11:45 AM, Dominik Szczerba <<a href="mailto:dominik@itis.ethz.ch" target="_blank">dominik@itis.ethz.ch</a> <mailto:<a href="mailto:dominik@itis.ethz.ch" target="_blank">dominik@itis.ethz.ch</a>>> wrote:<br>
<br>
I am using<br>
<br>
-ksp_rtol 0 -ksp_atol 1e-6 (defaults for all the rest)<br>
<br>
and observe for most of the times convergence, as expected.<br>
<br>
Starting KSPSolve<br>
0 KSP Residual norm 6.315132950112e-04<br>
1 KSP Residual norm 9.139539245601e-05<br>
2 KSP Residual norm 1.869750068013e-05<br>
3 KSP Residual norm 6.788145356909e-06<br>
4 KSP Residual norm 1.729115712274e-06<br>
5 KSP Residual norm 7.214445966641e-07<br>
solved in 2.328146e+00/4.144644e+00 s<br>
solution converged in 5 iterations<br>
||Ax-b|| = 7.21445e-07<br>
<br>
with ||...|| being residue computed on my own for a check.<br>
<br>
However, SOMETIMES, I do observe:<br>
<br>
Starting KSPSolve<br>
0 KSP Residual norm 1.228562818225e-02<br>
1 KSP Residual norm 4.045833318123e-04<br>
2 KSP Residual norm 6.589999324657e-05<br>
3 KSP Residual norm 1.494849372638e-05<br>
4 KSP Residual norm 4.311592172896e-06<br>
solved in 1.368086e+00/2.235830e+00 s<br>
solution converged in 4 iterations<br>
||Ax-b|| = 4.31159e-06<br>
<br>
which means that the iterations break sooner than they should.<br>
Again, this is not often, but worries me why this should happen...<br>
Using or not -ksp_converged_use_initial_residual_norm does not make<br>
any difference to this behavior: the last output Residual norm is<br>
sometimes higher than the requested atol.<br>
<br>
Any clarifications are highly appreciated.<br>
<br>
Dominik<br>
<br>
<br>
<br>
<br>
-- <br>
What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.<br>
-- Norbert Wiener<br>
</div></div></blockquote>
<br>
</blockquote></div><br><br clear="all"><br>-- <br>What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.<br>-- Norbert Wiener<br>