On Mon, Apr 2, 2012 at 11:33 AM, Xuefeng Li <span dir="ltr"><<a href="mailto:li@loyno.edu">li@loyno.edu</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">
On Mon, 2 Apr 2012, Matthew Knepley wrote:<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
On Mon, Apr 2, 2012 at 10:11 AM, Xuefeng Li <<a href="mailto:li@loyno.edu" target="_blank">li@loyno.edu</a>> wrote:<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Hi, all.<br>
<br>
I am using multigrid with DMMG (5 levels) in my petsc program.<br>
Turning on the dmmg_grid_sequence option seems<br>
to slow down the program dramatically. For instance,<br>
the program converges in 3 iterations in 6.270e+01 sec<br>
when dmmg_grid_sequence is OFF; it also converges<br>
in 3 iterations but in 1.699e+04 sec when dmmg_grid_sequence<br>
is ON.<br>
<br>
Is this typical of petsc (I'm using petsc-3.1-p8)?<br>
Are there ways to improve the performance when<br>
dmmg_grid_sequence is ON?<br>
<br>
</blockquote>
<br>
Its impossible to know what is happening from this description. There is no<br>
"generic" performance of grid sequencing. The best thing I can say to this<br>
is,<br>
it looks like the interpolated solution to your coarse problem is not a<br>
great<br>
solution to your fine problem.<br>
<br>
</blockquote>
That is possible.<br>
<br>
<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
However, you could be oversolving the coarse<br>
problems, blowing up the work, or interpolating badly for your problem.<br>
<br>
</blockquote>
What do you mean by "oversolving the coarse problems"?<br>
I am using the same stopping conditions in the two runs.<br></blockquote><div><br></div><div>Which may be too tight for the coarse runs.</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
The only difference is dmmg_grid_sequence ON and OFF.<br>
<br>
I also observed that in other situations, when dmmg_grid_sequence<br>
is ON, the execution will abort due to "Out of memory" error,<br>
whereas when OFF, the program converges without error.<br>
How much more memory does grid-sequence employ? Twice as much?<br></blockquote><div><br></div><div>If you use Galerkin MG, then t would depend on how full your Jacobian is.</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">
Regards,<br>
<br>
--Xuefeng Li, <a href="tel:%28504%29865-3340" value="+15048653340" target="_blank">(504)865-3340</a>(phone)<br>
Like floating clouds, the heart rests easy<br>
Like flowing water, the spirit stays free<br>
<a href="http://www.loyno.edu/~li/home" target="_blank">http://www.loyno.edu/~li/home</a><br>
New Orleans, Louisiana <a href="tel:%28504%29865-2051" value="+15048652051" target="_blank">(504)865-2051</a>(fax)<br>
<br>
</blockquote></div><br><br clear="all"><div><br></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>