<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html;
      charset=windows-1252">
  </head>
  <body>
    FYI MUMPS 5.x works with IPOPT. The parallelism is decent as well,
    it's now threaded as well as multi-process.<br>
    <br>
    Damien<br>
    <br>
    <div class="moz-cite-prefix">On 1/23/2020 7:24 AM, Elias Trommer
      wrote:<br>
    </div>
    <blockquote type="cite"
      cite="mid:1579789454546.77601@rl-institut.de">
      <meta http-equiv="Content-Type" content="text/html;
        charset=windows-1252">
      <style type="text/css" style="display:none"><!--P{margin-top:0;margin-bottom:0;} --></style>
      <p>Very good advice, thank you!<br>
        Indeed I was using MUMPS4.0 and switching over to MA97 seems to
        have fixed it! I'll look into PARDISO as well.<br>
        <br>
        Best regards, Elias<br>
        <br>
      </p>
      <div style="color: rgb(33, 33, 33);">
        <hr tabindex="-1" style="display:inline-block; width:98%">
        <div id="divRplyFwdMsg" dir="ltr"><font style="font-size:11pt"
            face="Calibri, sans-serif" color="#000000"><b>Von:</b> Sean
            Hardesty <a class="moz-txt-link-rfc2396E" href="mailto:sean.hardesty@gmail.com"><sean.hardesty@gmail.com></a><br>
            <b>Gesendet:</b> Donnerstag, 16. Januar 2020 05:58<br>
            <b>An:</b> Elias Trommer<br>
            <b>Cc:</b> <a class="moz-txt-link-abbreviated" href="mailto:ipopt@list.coin-or.org">ipopt@list.coin-or.org</a><br>
            <b>Betreff:</b> Re: [Ipopt] Segmentation fault due to MUMPS
            memory allocation</font>
          <div> </div>
        </div>
        <div>
          <div dir="ltr">
            <div>Do you know what version of MUMPS you are using?
              According to my notes from experiments a couple years ago,
              MUMPS 4 does not use internal 64-bit integer indexing, so
              if you have more than 2^31 non-zeros (or maybe 2^32), it
              will crash. I think this is supported as of MUMPS 5.1.0,
              but I don't know whether Ipopt works with MUMPS 5. In
              order to solve problems requiring 64-bit indexing, I had
              to use MA86 or MA97. MA27 and MA57 don't use 64-bit
              integers either. I tried to use Pardiso, but had
              difficulty getting it to link against MKL, so I don't know
              how it behaves...</div>
            <div><br>
            </div>
            <div>    Sean<br>
            </div>
          </div>
          <br>
          <div class="gmail_quote">
            <div dir="ltr" class="gmail_attr">On Mon, Jan 13, 2020 at
              1:34 AM Elias Trommer <<a
                href="mailto:elias.trommer@rl-institut.de"
                moz-do-not-send="true">elias.trommer@rl-institut.de</a>>
              wrote:<br>
            </div>
            <blockquote class="gmail_quote" style="margin:0px 0px 0px
              0.8ex; border-left:1px solid rgb(204,204,204);
              padding-left:1ex">
              <div dir="ltr" style="font-size:12pt; color:rgb(0,0,0);
                background-color:rgb(255,255,255);
                font-family:Calibri,Arial,Helvetica,sans-serif">
                <p>Hi everyone,</p>
                <p>I'm trying to use ipopt as a solver for a non-linear
                  optimization problem in Julia. This works fine for
                  smaller instances of the problem, however for larger
                  instances MUMPS seems to fail because of a failing
                  memory allocation. The full error output is at the end
                  of this eMail.<br>
                </p>
                <p><br>
                </p>
                <p>Because I previously had problems with this, I've
                  decreased the option `mumps_mem_percent` to a value of
                  100 so that the restoration would not fail during the
                  first iteration as described in the error message.</p>
                <p>However, for this particular problem, I'm having a
                  hard time pinning down the source. It's running on an
                  Ubuntu 18.04 machine with 64Gb of RAM and 16Gb of
                  Swap, so I presume it's not a lack of actual memory,
                  but rather an allocation issue. I've also checked the
                  ulimit on the machine I'm working on and it's set to
                  unlimited.<br>
                </p>
                <p><br>
                </p>
                <p>I was hoping someone could help me pin down the cause
                  of this issue (and ideally how to fix it)!</p>
                <p><br>
                </p>
                <p>Thank you in advance,</p>
                <p>Elias Trommer<br>
                </p>
                <p><br>
                </p>
                ```
                <p>This is Ipopt version 3.12.10, running with linear
                  solver mumps.<br>
                  NOTE: Other linear solvers might be more efficient
                  (see Ipopt documentation).<br>
                  <br>
                  Number of nonzeros in equality constraint
                  Jacobian...:   863116<br>
                  Number of nonzeros in inequality constraint
                  Jacobian.:   108177<br>
                  Number of nonzeros in Lagrangian
                  Hessian.............:   358996</p>
                <p><br>
                  Total number of
                  variables............................:   145445<br>
                                       variables with only lower
                  bounds:        0<br>
                                  variables with lower and upper
                  bounds:   144771<br>
                                       variables with only upper
                  bounds:        0<br>
                  Total number of equality
                  constraints.................:   145338<br>
                  Total number of inequality
                  constraints...............:    36059<br>
                          inequality constraints with only lower
                  bounds:        0<br>
                     inequality constraints with lower and upper
                  bounds:        0<br>
                          inequality constraints with only upper
                  bounds:    36059<br>
                  <br>
                  iter    objective    inf_pr   inf_du lg(mu)  ||d|| 
                  lg(rg) alpha_du alpha_pr  ls<br>
                     0  2.3869702e+03 4.60e+01 1.00e+00  -1.0
                  0.00e+00    -  0.00e+00 0.00e+00   0<br>
                     1  2.3914541e+03 4.48e+01 2.58e+05  -1.0
                  6.23e+01    -  2.96e-05 2.61e-02F  1<br>
                     2  2.3936754e+03 3.08e+02 7.11e+07  -1.0
                  6.07e+01    -  4.74e-01 9.61e-01f  1<br>
                     3  2.3924281e+03 3.08e+02 7.10e+07  -1.0 4.05e+03 
                  -2.0 1.11e-03 6.56e-04f  1<br>
                     4  2.3924033e+03 3.08e+02 7.10e+07  -1.0 1.40e+03 
                  -1.6 5.44e-03 3.68e-05f  1<br>
                     5  2.3920925e+03 3.08e+02 7.10e+07  -1.0 3.06e+03 
                  -2.1 2.68e-03 3.03e-04f  1<br>
                     6  2.3918892e+03 3.08e+02 7.10e+07  -1.0
                  1.07e+04    -  6.52e-04 1.30e-04f  1<br>
                     7  2.3915687e+03 3.07e+02 7.09e+07  -1.0
                  1.11e+03    -  5.12e-03 7.38e-04f  1<br>
                     8  2.3908132e+03 3.07e+02 7.08e+07  -1.0
                  1.13e+03    -  1.08e-02 2.08e-03f  1<br>
                     9  2.3907401e+03 3.07e+02 9.31e+09  -1.0
                  7.76e+02    -  2.79e-02 1.84e-04f  1<br>
                  iter    objective    inf_pr   inf_du lg(mu)  ||d|| 
                  lg(rg) alpha_du alpha_pr  ls<br>
                    10  2.3892870e+03 3.05e+02 1.62e+10  -1.0
                  5.84e+02    -  2.56e-02 6.94e-03f  1<br>
                    11  2.3894049e+03 3.04e+02 3.09e+10  -1.0
                  4.25e+02    -  4.15e-02 1.21e-03f  1<br>
                    12  2.3960908e+03 2.88e+02 3.92e+10  -1.0
                  3.76e+02    -  8.17e-02 5.33e-02f  1<br>
                    13  2.4009546e+03 2.78e+02 2.85e+10  -1.0
                  3.05e+02    -  3.87e-03 3.45e-02f  1<br>
                    14  2.4012483e+03 2.77e+02 3.19e+10  -1.0
                  2.93e+02    -  1.40e-02 2.36e-03f  1<br>
                    15  2.4038152e+03 2.72e+02 3.44e+10  -1.0
                  2.93e+02    -  3.19e-02 2.10e-02f  1<br>
                    16  2.4038403e+03 2.72e+02 4.23e+10  -1.0
                  2.86e+02    -  2.55e-02 2.15e-04h  1<br>
                    17  2.4036648e+03 2.71e+02 4.17e+10  -1.0
                  4.57e+03    -  3.36e-06 3.15e-04f  1<br>
                    18  2.4069581e+03 2.67e+02 3.57e+10  -1.0
                  2.89e+02    -  2.86e-04 1.71e-02f  1<br>
                    19  2.4069888e+03 2.67e+02 6.39e+10  -1.0
                  2.83e+02    -  2.26e-02 1.75e-04h  1<br>
                  iter    objective    inf_pr   inf_du lg(mu)  ||d|| 
                  lg(rg) alpha_du alpha_pr  ls<br>
                    20  2.4068113e+03 2.67e+02 6.38e+10  -1.0
                  3.88e+04    -  1.85e-06 2.86e-05h  1<br>
                    21  2.4067828e+03 2.67e+02 4.51e+10  -1.0
                  4.20e+03    -  3.96e-06 7.14e-05h  1<br>
                  MUMPS returned INFO(1) = -9 and requires more memory,
                  reallocating.  Attempt 1<br>
                    Increasing icntl[13] from 80 to 160.<br>
                  MUMPS returned INFO(1) = -9 and requires more memory,
                  reallocating.  Attempt 1<br>
                    Increasing icntl[13] from 160 to 320.<br>
                  MUMPS returned INFO(1) = -9 and requires more memory,
                  reallocating.  Attempt 2<br>
                    Increasing icntl[13] from 320 to 640.<br>
                  MUMPS returned INFO(1) = -9 and requires more memory,
                  reallocating.  Attempt 3<br>
                    Increasing icntl[13] from 640 to 1280.<br>
                  MUMPS returned INFO(1) = -9 and requires more memory,
                  reallocating.  Attempt 4<br>
                    Increasing icntl[13] from 1280 to 2560.<br>
                    22  2.4067829e+03 2.67e+02 4.51e+10  -1.0 5.70e+00  
                  5.8 2.39e-05 1.79e-05h  1<br>
                  MUMPS returned INFO(1) = -9 and requires more memory,
                  reallocating.  Attempt 1<br>
                    Increasing icntl[13] from 2560 to 5120.<br>
                  MUMPS returned INFO(1) = -9 and requires more memory,
                  reallocating.  Attempt 2<br>
                    Increasing icntl[13] from 5120 to 10240.<br>
                  MUMPS returned INFO(1) = -9 and requires more memory,
                  reallocating.  Attempt 3<br>
                    Increasing icntl[13] from 10240 to 20480.<br>
                   Problem with integer stack size           1          
                  1           6<br>
                  <br>
                  signal (11): Segmentation Fault</p>
                ```
                <p><br>
                </p>
              </div>
              _______________________________________________<br>
              Ipopt mailing list<br>
              <a href="mailto:Ipopt@list.coin-or.org" target="_blank"
                moz-do-not-send="true">Ipopt@list.coin-or.org</a><br>
              <a href="https://list.coin-or.org/mailman/listinfo/ipopt"
                rel="noreferrer" target="_blank" moz-do-not-send="true">https://list.coin-or.org/mailman/listinfo/ipopt</a><br>
            </blockquote>
          </div>
        </div>
      </div>
      <br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <pre class="moz-quote-pre" wrap="">_______________________________________________
Ipopt mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Ipopt@list.coin-or.org">Ipopt@list.coin-or.org</a>
<a class="moz-txt-link-freetext" href="https://list.coin-or.org/mailman/listinfo/ipopt">https://list.coin-or.org/mailman/listinfo/ipopt</a>
</pre>
    </blockquote>
    <br>
  </body>
</html>