The biggest reason that the BLAS/LAPACK configure changes so rapidly among patchlevels<br>is that MKL is the worst library for maintenance. They change the content AND the names<br>of the libraries, as well as the link order, in almost every version. I am beginning to really<br>
dislike the developers personally.<br><br>I suspect the difference is the exact patchlevel of PETSc used on the different machines. However,<br>as always with configure problems, we cannot tell you anything at all unless you send configure.log.<br>
<br> Matt<br><br><div class="gmail_quote">On Tue, Aug 25, 2009 at 5:47 AM, Thomas Witkowski <span dir="ltr"><<a href="mailto:thomas.witkowski@tu-dresden.de">thomas.witkowski@tu-dresden.de</a>></span> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Hi,<br>
<br>
I want to compile petsc on two larger HPC system, on both we use Intels MKL for blas and lapack. The first one is a pc farm, and I've successfully compiled petsc with the parameter<br>
<br>
--with-blas-lapack-dir=/work/licsoft/libraries/mkl/<a href="http://10.0.3.020/lib/em64t" target="_blank">10.0.3.020/lib/em64t</a><br>
<br>
This works fine. On the other system, its an SGI Altix, I've tried to do the same. But it fails in configure, because petsc tries to link the test program with "-lblas". The same test program on the pc farm was linked with the correct library parameter "-lmkl". May question is, why this can happen and what I can do to avoid the problem? Thanks for answering!<br>
<font color="#888888">
<br>
Thomas<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>