On Thu, Oct 21, 2010 at 5:01 PM, Jed Brown <span dir="ltr"><<a href="mailto:jed@59a2.org">jed@59a2.org</a>></span> wrote:<br><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div class="im">On Thu, Oct 21, 2010 at 23:56, Matthew Knepley <<a href="mailto:knepley@gmail.com">knepley@gmail.com</a>> wrote:<br>
> You SNES is not converging. You solve 163 systems, but do 4300 function<br>
> evaluations.<br>
<br>
</div>That is because the Jacobian is assembled using coloring.<br></blockquote><div><br></div><div>With that many colors, it seems like you should sit down and work out the Jacobian.</div><div><br></div><div> Matt</div>
<div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
Sundials is probably doing fewer steps than BEuler for this problem,<br>
and it's also lagging the preconditioner and just using matrix-free<br>
differencing to define the Krylov operator. These options are not<br>
default with PETSc's TS implementations.<br>
<font color="#888888"><br>
Jed<br>
</font></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>