On Thu, May 12, 2011 at 9:51 AM, Sanjay Govindjee <span dir="ltr"><<a href="mailto:s_g@berkeley.edu">s_g@berkeley.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;">
Sylvain,<br>
<br>
Is there a reason you are not using Prometheus (--download_prometheus=1) for your MG preconditioner?<br>
It was designed with 3D solid mechanics in mind.</blockquote><div><br></div><div><br></div><div>I would also point out that even for power law problems, PyLith uses ML and PCFIELDSPLIT for the elastic</div><div>solver and it works great. Have you tried this?</div>
<div><br></div><div> Thanks,</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;"><font color="#888888"><br>
-sanjay</font><div><div></div><div class="h5"><br>
<br>
<br>
On 5/11/11 11:12 PM, Sylvain Barbot wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Dear Jed,<br>
<br>
During my recent visit to ETH, I talked at length about multi-grid<br>
with Dave May who warned me about the issues of large<br>
coefficient-contrasts. Most of my problems of interest for<br>
tectonophysics and earthquake simulations are cases of relatively<br>
smooth variations in elastic moduli. So I am not too worried about<br>
this aspect of the problem. I appreciate your advice about trying<br>
simpler solutions first. I have tested at length direct solvers of 2-D<br>
and 3-D problems of elastic deformation and I am quite happy with the<br>
results. My primary concern now is computation speed, especially for<br>
3-D problems, where i have of the order 512^3 degrees of freedom. I<br>
was planning to test Jacobi and SOR smoothers. Is there another<br>
smoother you recommend for this kind of problem?<br>
<br>
Thanks,<br>
Sylvain<br>
<br>
2011/5/11 Jed Brown<<a href="mailto:jed@59a2.org" target="_blank">jed@59a2.org</a>>:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
On Wed, May 11, 2011 at 04:20, Sylvain Barbot<<a href="mailto:sylbar.vainbot@gmail.com" target="_blank">sylbar.vainbot@gmail.com</a>><br>
wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
I am still trying to design a<br>
multigrid preconditionner for the Navier's equation of elasticity.<br>
</blockquote>
I have heard, through an external source, that you have large jumps in both<br>
Young's modulus and Poisson ratio that are not grid aligned, including<br>
perhaps thin structures that span a large part of the domain. Such problems<br>
are pretty hard, so I suggest you focus on robustness and do not worry about<br>
low-memory implementation at this point. That is, you should assemble the<br>
matrices in a usual PETSc format instead of using MatShell to do everything<br>
matrix-free. This gives you access to much stronger smoothers.<br>
After you find a scheme that is robust enough for your purposes, _then_ you<br>
can make it low-memory by replacing some assembled matrices by MatShell. To<br>
realize most of the possible memory savings, it should be sufficient to do<br>
this on the finest level only.<br>
</blockquote></blockquote>
</div></div></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>