<div dir="ltr">On Sun, Dec 23, 2012 at 8:15 PM, Sanjay Govindjee <span dir="ltr"><<a href="mailto:s_g@berkeley.edu" target="_blank">s_g@berkeley.edu</a>></span> wrote:<br><div class="gmail_extra"><div class="gmail_quote">
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
  
    
  
  <div bgcolor="#FFFFFF" text="#000000">
    Sorry for the confusion.  I thought I was clear.  Here is the make
    line I was running.<br>
    <br>
    <br>
            -@${MPIEXEC} -n 2 ./ex6 -ksp_type preonly  -pc_type lu
    -pc_factor_mat_solver_package superlu_dist -options_left no \<br>
               -f arco1 > ex6_1.tmp 2>&1; \<br>
               if (${DIFF} output/ex6_1.out ex6_1.tmp) then true; \<br>
               else echo ${PWD} ; echo "Possible problem with with
    ex6_1, diffs above \n========================================="; fi;
    \<br>
               ${RM} -f ex6_1.tmp<br>
    <br>
    If you change superlu_dist to spooles it works just fine as well as
    any other iterative methods you care to try.  The matrix arcos1 was
    downloaded as per the instructions in the makefile.<br></div></blockquote><div><br></div><div style>I cannot reproduce your problem. Do you have a build with a different compiler (like GCC)? Also, what BLAS/LAPACK is being used? (You can send configure.log to <a href="mailto:petsc-maint@mcs.anl.gov">petsc-maint@mcs.anl.gov</a>.)</div>
<div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div bgcolor="#FFFFFF" text="#000000">
    I will try reproducing the superlu_dist error with
    snes/examples/tutorials/ex5 now. </div></blockquote><div style><br>This is the file Matt suggested.<br> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div bgcolor="#FFFFFF" text="#000000">
 (fyi under
    snes/examples/tests/output the files ex5_1.out and ex5_2.out are
    missing one can not run the test out of the box).</div></blockquote><div><br></div><div style>Heh, this has been missing since the beginning of time (revision 0). I'll add it.</div><div><br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div bgcolor="#FFFFFF" text="#000000"><span class="HOEnZb"><font color="#888888"><br>
    <br>
    -sanjay</font></span><div><div class="h5"><br>
    <br>
    <br>
    <div>On 12/23/12 6:07 PM, Jed Brown wrote:<br>
    </div>
    <blockquote type="cite">
      <div dir="ltr">You didn't say what options you were running ex6
        with, but with the options used for the tests, I see<br>
        <br>
        <div>~/petsc/src/ksp/ksp/examples/tests$ mpirun.hydra -n 2 ./ex6
          -f ~/petsc/datafiles/matrices/arco1 -pc_type lu
          -pc_factor_mat_solver_package superlu_dist</div>
        <div>Number of iterations =   1</div>
        <div>Residual norm = 2.23439e-11<br>
          <br>
          <br>
          You need to give precise instructions for how to reproduce the
          behavior you are seeing.<br>
          <br>
          Also, for experimenting with matrices read from files, we
          prefer src/ksp/ksp/examples/tutorials/ex10.c because it is
          better commented and has more features.</div>
      </div>
      <div class="gmail_extra"><br>
        <br>
        <div class="gmail_quote">On Sun, Dec 23, 2012 at 7:08 PM, Sanjay
          Govindjee <span dir="ltr"><<a href="mailto:s_g@berkeley.edu" target="_blank">s_g@berkeley.edu</a>></span>
          wrote:<br>
          <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
            <div bgcolor="#FFFFFF" text="#000000"> Not sure what you
              mean by where is your matrix?  I am simply running ex6 in
              the ksp/examples/tests directory.<br>
              <br>
              The reason I ran this test is because I was seeing the
              same behavior with my finite element code (on perfectly
              benign problems).<br>
              <br>
              Is there a built-in test that you use to check that
              superlu_dist is working properly with petsc?<br>
              i.e. something you know that works with with petsc 3.3-p5?<br>
              <br>
              -sanjay
              <div>
                <div><br>
                  <br>
                  <br>
                  <div>On 12/23/12 4:56 PM, Jed Brown wrote:<br>
                  </div>
                  <blockquote type="cite">
                    <div dir="ltr">Where is your matrix? It might be
                      ending up with a very bad pivot. If the problem
                      can be reproduced, it should be reported to the
                      SuperLU_DIST developers to fix. (Note that we do
                      not see this with other matrices.) You can also
                      try MUMPS.</div>
                    <div class="gmail_extra"><br>
                      <br>
                      <div class="gmail_quote">On Sun, Dec 23, 2012 at
                        6:48 PM, Sanjay Govindjee <span dir="ltr"><<a href="mailto:s_g@berkeley.edu" target="_blank">s_g@berkeley.edu</a>></span>
                        wrote:<br>
                        <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
                          <div bgcolor="#FFFFFF" text="#000000"> I
                            wanted to use SuperLU Dist to perform a
                            direct solve but seem to be encountering<br>
                            a problem.  I was wonder if this is a know
                            issue and if there is a solution for it.<br>
                            <br>
                            The problem is easily observed using ex6.c
                            in src/ksp/ksp/examples/tests.<br>
                            <br>
                            Out of the box: make runex6 produces a
                            residual error of O(1e-11), all is well.<br>
                            <br>
                            I then changed the run to run on two
                            processors and add the flag<br>
                            -pc_factor_mat_solver_package spooles  this
                            produces a residual error of O(1e-11), all
                            is still well.<br>
                            <br>
                            I then switch over to
                            -pc_factor_mat_solver_package superlu_dist
                            and the<br>
                            residual error comes back as 22.6637! 
                            Something seems very wrong.<br>
                            <br>
                            My build is perfectly vanilla:<br>
                            <br>
                            export PETSC_DIR=/Users/sg/petsc-3.3-p5/<br>
                            export PETSC_ARCH=intel<br>
                            <br>
                            ./configure --with-cc=icc --with-fc=ifort  \
-download-{spooles,parmetis,superlu_dist,prometheus,mpich,ml,hypre,metis}<br>
                            <br>
                            make PETSC_DIR=/Users/sg/petsc-3.3-p5/
                            PETSC_ARCH=intel all<br>
                            make PETSC_DIR=/Users/sg/petsc-3.3-p5/
                            PETSC_ARCH=intel test<span><font color="#888888"><br>
                                <br>
                                -sanjay<br>
                              </font></span></div>
                        </blockquote>
                      </div>
                      <br>
                    </div>
                  </blockquote>
                  <br>
                </div>
              </div>
            </div>
          </blockquote>
        </div>
      </div>
    </blockquote>
    <br>
    <pre cols="72"></pre>
  </div></div></div>

</blockquote></div><br></div></div>