<div class="gmail_quote">On Wed, Dec 14, 2011 at 02:32, Bogdan Dita <span dir="ltr"><<a href="mailto:bogdan@lmn.pub.ro">bogdan@lmn.pub.ro</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div id=":2uk">So if on the cluster the mpich2 version is optimised whatever<br>
implementation of Petsc I choose, I can't get a better time result?<br>
If this is true then this are really bad news.<br>
What are your sugestions? So far I used superlu_dist, I tried mumps a few<br>
times but I always got some strange errors and if I recall correctly from<br>
the petsc-users emails it is best to use mumps with petsc-dev?<br></div></blockquote></div><br><div>You can pass the specific optimization flags that you want. The confusion is that MPICH2 is turning optimization on by default, so even the "debug" build is using optimized compiler flags.</div>
<div><br></div><div>As for "strange errors", you have to send more information: configure.log, what you expected to happen, what actually happened, and the entire error message.</div>