On Fri, Feb 13, 2009 at 4:37 PM, Yixun Liu <span dir="ltr"><<a href="mailto:enjoywm@cs.wm.edu" target="_blank">enjoywm@cs.wm.edu</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;">
Hi,<br>
Assume Ku=F is a linear system assembled from element stiffness<br>
matrix. I want to use KSP to solve Ku=F. Assume the size of K is 100x100<br>
and the number of processors is 4.<br>
The question is does the efficiency of KSP depend on the structure of K?<br>
For example if each process owns 25 rows does the efficiency of the<br>
solver depend on the structure of K?</blockquote><div><br>1) Strictly, yes it does.<br><br>2) Practically, not very much. You want to have at least a decent partition, but<br> after that the communication load is small since its a PDE, and scales well.<br>
<br>3) Everything depends on the type of PDE. Krylov solvers will not work well for<br> everything, but for elliptic things they are good.<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;">
<br>
Thanks.<br>
<font color="#888888"><br>
Yixun<br>
</font></blockquote></div>-- <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>