<html>
  <head>
    <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    I have seen similar behavior on my mac (works fine on Linux) -- I
    reported this to the mailing list a few weeks back.   I eventually
    tracked it down to a BLAS issue but gave up on finding the exact
    cause as I needed to move on -- moved over to MUMPS.   But the
    problem is not in your imagination.<br>
    <br>
    If I have the time, I will try to get back to it (especially since I
    finally learned that you have to use dsymutil to get the line
    numbers in the debugger/valgrind).<br>
    <br>
    -sanjay<br>
    <br>
    <br>
    <br>
    <div class="moz-cite-prefix">On 1/28/13 3:09 PM, Jed Brown wrote:<br>
    </div>
    <blockquote
cite="mid:CAM9tzS==HufkQEo3UOSYn9HweAuh=fk=xhafSZTkavG6vQCVug@mail.gmail.com"
      type="cite">
      <div dir="ltr">Vague "random answers" isn't very helpful. If there
        is a real problem, we'd like a test case so we can track it
        down.</div>
      <div class="gmail_extra"><br>
        <br>
        <div class="gmail_quote">On Mon, Jan 28, 2013 at 3:58 PM, Gaetan
          Kenway <span dir="ltr"><<a moz-do-not-send="true"
              href="mailto:kenway@utias.utoronto.ca" target="_blank">kenway@utias.utoronto.ca</a>></span>
          wrote:<br>
          <blockquote class="gmail_quote" style="margin:0 0 0
            .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi
            everyone
            <div><br>
            </div>
            <div>I have the exactly same issue actually. When I updated
              to petsc-3.3, SuperLU_dist was giving me random answers to
              KSPSolve(). Maybe half of the time you would get the same
              result as 3.2, other times it was a little off and other
              times widely differnet. I am using SuperLU_dist with a
              PREONLY ksp object. </div>
            <div><br>
            </div>
            <div>I haven't tracked down what is causing it and reverted
              back to petsc-3.2 that still works. </div>
            <div><br>
            </div>
            <div>Also, to fix the issue with the configure below, just
              drop out the download-metis. You need it for 3.3 but not
              3.2</div>
            <span class="HOEnZb"><font color="#888888">
                <div><br>
                </div>
              </font></span>
            <div><span class="HOEnZb"><font color="#888888">Gaetan</font></span>
              <div>
                <div class="h5"><br>
                  <br>
                  <div class="gmail_quote">On Mon, Jan 28, 2013 at 4:42
                    PM, Jed Brown <span dir="ltr"><<a
                        moz-do-not-send="true"
                        href="mailto:jedbrown@mcs.anl.gov"
                        target="_blank">jedbrown@mcs.anl.gov</a>></span>
                    wrote:<br>
                    <blockquote class="gmail_quote" style="margin:0 0 0
                      .8ex;border-left:1px #ccc solid;padding-left:1ex">
                      <div dir="ltr">Send -ksp_monitor_true_residual
                        -ksp_view output for both cases so we can try to
                        identify the source of the different convergence
                        behavior.
                        <div>
                          <div>
                            <div class="gmail_extra"><br>
                              <br>
                              <div class="gmail_quote">On Mon, Jan 28,
                                2013 at 3:37 PM, Brian Helenbrook <span
                                  dir="ltr"><<a
                                    moz-do-not-send="true"
                                    href="mailto:bhelenbr@clarkson.edu"
                                    target="_blank">bhelenbr@clarkson.edu</a>></span>
                                wrote:<br>
                                <blockquote class="gmail_quote"
                                  style="margin:0 0 0
                                  .8ex;border-left:1px #ccc
                                  solid;padding-left:1ex">Dear
                                  Petsc-Users-List,<br>
                                  <br>
                                  I recently upgraded to petsc3.3-p5
                                  from petsc3.2-p7 and the results from
                                  my code have changed.  I am using
                                  superLU with the following options:<br>
                                  <br>
                                  -ksp_type preonly -pc_type lu
                                  -pc_factor_mat_solver_package
                                  superlu_dist<br>
                                  <br>
                                  Everything was working with petsc3.2
                                  but now I get totally different
                                  answers and the iteration doesn't
                                  converge.   My build configuration is<br>
                                  <br>
                                  ./config/configure.py
                                  --prefix=${HOME}/Packages
                                  --with-fortran=0
                                   --download-superlu_dist=1 --with-x=0
                                  --download-parmetis=1
                                  --download-metis=1
                                  --with-mpi-dir=${HOME}/Packages
                                  --with-valgrind-dir=${HOME}/Packages<br>
                                  <br>
                                  I am running on OS X 10.8.2 with
                                  openmpi-1.6.3.<br>
                                  <br>
                                  I have run valgrind on my code and it
                                  is clean (except for start-up issues
                                  with mpi which occur before my code is
                                  entered.)<br>
                                  <br>
                                  I'm not very sure how to go about
                                  debugging this.   What I've tried is
                                  to re-install pets-3.2-p7, but now I
                                  am having trouble getting that to
                                  build:<br>
                                  <br>
                                  ./config/configure.py
                                  --prefix=${HOME}/Packages
                                  --with-fortran=0
                                   --download-superlu_dist=1 --with-x=0
                                  --download-parmetis=1
                                  --download-metis=1
                                  --with-mpi-dir=${HOME}/Packages
                                  --with-valgrind-dir=${HOME}/Packages<br>
===============================================================================<br>
                                               Configuring PETSc to
                                  compile on your system<br>
===============================================================================<br>
===============================================================================<br>
                                  Compiling & installing Metis; this
                                  may take several minutes              
                                  ===============================================================================
                                                                       
                                                                       
                                     
*******************************************************************************<br>
                                           UNABLE to CONFIGURE with
                                  GIVEN OPTIONS    (see configure.log
                                  for details):<br>
-------------------------------------------------------------------------------<br>
                                  Error running make on Metis: Could not
                                  execute "cd
                                  /Users/bhelenbr/Packages/petsc-3.2-p7/externalpackages/metis-4.0.3
                                  && make clean && make
                                  library && make minstall
                                  && make clean":<br>
                                  <br>
                                  <br>
                                  Any ideas what direction to go with
                                  this?<br>
                                  <br>
                                  Thanks,<br>
                                  <br>
                                  Brian<br>
                                  <br>
                                  <br>
                                  <br>
                                  Brian Helenbrook<br>
                                  Associate Professor<br>
                                  362 CAMP<br>
                                  Mech. and Aero. Eng. Dept.<br>
                                  Clarkson University<br>
                                  Potsdam, NY 13699-5725<br>
                                  <br>
                                  <span>
                                    work: <span title="Call with Google
                                      Voice"><a moz-do-not-send="true"
                                        href="tel:315-268-2204"
                                        value="+13152682204"
                                        target="_blank">315-268-2204</a></span></span><br>
                                  <span>
                                    fax: <span title="Call with Google
                                      Voice"><a moz-do-not-send="true"
                                        href="tel:315-268-6695"
                                        value="+13152686695"
                                        target="_blank">315-268-6695</a></span></span><br>
                                  <br>
                                  <br>
                                  <br>
                                </blockquote>
                              </div>
                              <br>
                            </div>
                          </div>
                        </div>
                      </div>
                    </blockquote>
                  </div>
                  <br>
                </div>
              </div>
            </div>
          </blockquote>
        </div>
        <br>
      </div>
    </blockquote>
    <br>
  </body>
</html>