<p>The linear solve computed a search direction that was not a descent direction. It determined this by attempting a line search.</p>
<div class="gmail_quote">On Apr 23, 2012 8:23 AM, "Dominik Szczerba" <<a href="mailto:dominik@itis.ethz.ch">dominik@itis.ethz.ch</a>> wrote:<br type="attribution"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Hi,<br>
<br>
I am implementing both FormFunction and FormJacobian for the SNES<br>
solver. It seems to start off correctly, with the 2 functions being<br>
called one after another in a loop, and with ||Ax-b|| decreasing about<br>
3 orders of magnitude. But then only the FormFunction is being called<br>
several times, without subsequent calls to FormJacobian and obviously<br>
the residue is not being reduced any further, because there are no<br>
calls to the underlying ksp. If I inspect the solution it looks<br>
qualitatively "on the way" to the expected one, but the values are not<br>
correct, just as if the iterations were prematurely stopped. What sort<br>
of a problem can such a behavior indicate?<br>
<br>
Regards,<br>
Dominik<br>
</blockquote></div>