<html>
  <head>
    <meta content="text/html; charset=windows-1252"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <div class="moz-cite-prefix">On 12/20/13 3:40 PM, Que Cat wrote:<br>
    </div>
    <blockquote
cite="mid:CA+Z2Wfi38CtJ-nKSfSBHGqe=6Qn3yN9RAe5h59gY1kHE8c=S8A@mail.gmail.com"
      type="cite">
      <meta http-equiv="Content-Type" content="text/html;
        charset=windows-1252">
      <div dir="ltr"><font face="times new roman, serif">Barry,</font>
        <div><font face="times new roman, serif"><br>
          </font></div>
        <div><font face="times new roman, serif">It said that "No
            support for this operation for this object type! Matrix
            format mpiaij does not have a built-in PETSc LU!".</font></div>
      </div>
    </blockquote>
    Try with -pc_factor_mat_solver_package mumps (superlu_dist)<br>
    <br>
    If you didn't install them reconfigure with:<br>
    <br>
    --download-scalapack=1<br>
    --download-mumps=1<br>
    <br>
    or:<br>
    <br>
    --download-superlu_dist=1<br>
    <br>
    Or just run sequentially<br>
    <br>
    Anton<br>
    <blockquote
cite="mid:CA+Z2Wfi38CtJ-nKSfSBHGqe=6Qn3yN9RAe5h59gY1kHE8c=S8A@mail.gmail.com"
      type="cite">
      <div dir="ltr">
        <div><font face="times new roman, serif"><br>
          </font></div>
        <div><font face="times new roman, serif">I run it with optio<span
              style="background-color:rgb(255,255,255)">n "<code
                style="color:rgb(0,0,0)">-snes_monitor
                -ksp_monitor_true_residual -snes_converged_reason
                -ksp_converged_reason</code><span
                style="color:rgb(0,0,0)">." and received : </span></span></font></div>
        <div><span
            style="color:rgb(0,0,0);background-color:rgb(213,234,255)"><font
              face="times new roman, serif">
              <div>    0 SNES Function norm 1.414213647226e-01 </div>
              <div>    0 KSP preconditioned resid norm
                8.424084241999e+08 true resid norm 1.414213647226e-01
                ||r(i)||/||b|| 1.000000000000e+00</div>
              <div>    1 KSP preconditioned resid norm
                1.354908290122e-02 true resid norm 1.424694873074e-01
                ||r(i)||/||b|| 1.007411345428e+00</div>
              <div>Nonlinear solve did not converge due to
                DIVERGED_LINE_SEARCH iterations 0</div>
              <div>
                <br>
              </div>
              <div>It looks like that the preconditioned residual
                converges, but the true residual does not.</div>
            </font></span></div>
      </div>
      <div class="gmail_extra"><br>
        <br>
        <div class="gmail_quote">On Fri, Dec 20, 2013 at 8:30 AM, Barry
          Smith <span dir="ltr"><<a moz-do-not-send="true"
              href="mailto:bsmith@mcs.anl.gov" target="_blank">bsmith@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"><br>
              Run with<br>
            <br>
               -snes_mf_operator -pc_type lu<br>
            <br>
               I’m sorry I sent the wrong instructions before.<br>
            <span class="HOEnZb"><font color="#888888"><br>
                   Barry<br>
              </font></span>
            <div class="HOEnZb">
              <div class="h5"><br>
                On Dec 20, 2013, at 7:54 AM, Que Cat <<a
                  moz-do-not-send="true"
                  href="mailto:quecat001@gmail.com">quecat001@gmail.com</a>>
                wrote:<br>
                <br>
                > Hi Barry,<br>
                ><br>
                > Thanks for your exlanation. I used -snes_mf
                -pc_type lu and now it failed with reason -3 (/* the
                linear solve failed */). So, something is wrong with the
                function.<br>
                ><br>
                ><br>
                > On Thu, Dec 19, 2013 at 5:01 PM, Barry Smith <<a
                  moz-do-not-send="true"
                  href="mailto:bsmith@mcs.anl.gov">bsmith@mcs.anl.gov</a>>
                wrote:<br>
                ><br>
                >   ratio is || A - Afd || /||A|| where |||| is the
                frobenus norm and difference is || A - Afd ||.<br>
                ><br>
                >   Try using -snes_mf -pc_type lu<br>
                ><br>
                >    This can often converge even if the A you
                provide is wrong, so if this convergence sine then
                likely the Jacobian is wrong.<br>
                ><br>
                >    It is also possible that the function evaluation
                has a bug.<br>
                ><br>
                > <a moz-do-not-send="true"
                  href="http://www.mcs.anl.gov/petsc/documentation/faq.html#newton"
                  target="_blank">http://www.mcs.anl.gov/petsc/documentation/faq.html#newton</a><br>
                ><br>
                ><br>
                >    Barry<br>
                ><br>
                ><br>
                ><br>
                > On Dec 19, 2013, at 4:26 PM, Que Cat <<a
                  moz-do-not-send="true"
                  href="mailto:quecat001@gmail.com">quecat001@gmail.com</a>>
                wrote:<br>
                ><br>
                > > Hello,<br>
                > ><br>
                > > I use SNES to solve the non linear coupled
                equations. I have checked the hand-coded jacobian with
                -snes_type test and received the result:<br>
                > > Norm of matrix ratio 1.39606e-23 difference
                1.53372e-14 (user-defined state)<br>
                > > Norm of matrix ratio 3.03157e-10 difference
                0.33305 (constant state -1.0)<br>
                > > Norm of matrix ratio 3.03157e-10 difference
                0.33305 (constant state 1.0)<br>
                > > It said that "  if the ratio is O(1.e-8), the
                hand-coded Jacobian is probably correct." What is the
                second paramenter "difference 1.53372e-14 ...."? Does it
                play any role in checking the accuracy of hand-coded
                jacobian?<br>
                > ><br>
                > > The SNES solver for failed with the reason -6
                "the line search failed". I tried with all other line
                search method stated in the PETSC manual, but they all
                failed.<br>
                > ><br>
                > > I tried to figure out what is wrong with my
                code? Could you please give me any recommendation to
                direct me how to check my code? I appreciate your help.<br>
                > ><br>
                > > QUe<br>
                > ><br>
                ><br>
                ><br>
                <br>
              </div>
            </div>
          </blockquote>
        </div>
        <br>
      </div>
    </blockquote>
    <br>
  </body>
</html>