<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
  <head>
    <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#ffffff" text="#000000">
    Sorry, I don't quite understand what you mean by first error was
    earlier?<br>
    <br>
    Yes, symbols are on and I found the crash-related part in the log
    file. See it attached. <br>
    <br>
    Regards,<br>
    Alexander<br>
    <br>
    On 09.01.2012 12:51, Jed Brown wrote:
    <blockquote
cite="mid:CAM9tzSkik2jPhcKuMO4S06Z8d14i-wWMWx8+3FtfLQhycGE2eg@mail.gmail.com"
      type="cite">
      <p>The first error was likely earlier.</p>
      <p>Do you have debugging symbols turned on?</p>
      <p>Valgrind has options to write exception files for false
        positives. MPICH is usually valgrind-clean while Open MPI is
        quite noisy.</p>
      <p>I will look at the log when I have a computer.</p>
      <div class="gmail_quote">On Jan 9, 2012 5:47 AM, "Alexander
        Grayver" <<a moz-do-not-send="true"
          href="mailto:agrayver@gfz-potsdam.de">agrayver@gfz-potsdam.de</a>>
        wrote:<br type="attribution">
        <blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt
          0.8ex; border-left: 1px solid rgb(204, 204, 204);
          padding-left: 1ex;">
          <div bgcolor="#ffffff" text="#000000"> Hi Jed,<br>
            <br>
            On 09.01.2012 12:34, Jed Brown wrote:
            <blockquote type="cite">
              <div class="gmail_quote">On Mon, Jan 9, 2012 at 05:23,
                Alexander Grayver <span dir="ltr"><<a
                    moz-do-not-send="true"
                    href="mailto:agrayver@gfz-potsdam.de"
                    target="_blank">agrayver@gfz-potsdam.de</a>></span>
                wrote:<br>
                <blockquote class="gmail_quote" style="margin: 0pt 0pt
                  0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204);
                  padding-left: 1ex;">
                  <div>I run petsc-dev like that:<br>
                    mpirun -n 2 /home/solveTest -ksp_view -ksp_monitor
                    -ksp_converged_reason -ksp_monitor_true_residual
                    -log_summary -mat_type mpiaij -ksp_rtol 1.0e-12<br>
                    <br>
                  </div>
                </blockquote>
                <div>[...] </div>
                <blockquote class="gmail_quote" style="margin: 0pt 0pt
                  0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204);
                  padding-left: 1ex;">
                  <div> But just switching to one process (-n 1) makes
                    it crashing:<br>
                  </div>
                </blockquote>
                <div><br>
                </div>
                <div>Try removing the "-mat_type mpiaij" part. Note that
                  you can still use "-mat_type aij". Also note that you
                  should usually use a prefix so that you are setting
                  the type of a specific matrix instead of setting the
                  type of all matrices (for which MatSetFromOptions() is
                  called).</div>
              </div>
            </blockquote>
            <br>
            There is only one matrix in application. It is very simple
            one, the test just loads system matrix and rhs and solves
            system. I want to use it to test different
            solvers/preconditioners.<br>
            <br>
            <blockquote type="cite">
              <div class="gmail_quote">
                <div> </div>
                <blockquote class="gmail_quote" style="margin: 0pt 0pt
                  0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204);
                  padding-left: 1ex;">
                  <div> <br>
                    [0]PETSC ERROR:
                    ------------------------------------------------------------------------<br>
                    [0]PETSC ERROR: Caught signal number 11 SEGV:
                    Segmentation Violation, probably memory access out
                    of range</div>
                </blockquote>
              </div>
              <br>
              <div>Please run with valgrind --db-attach=yes (or at least
                a debugger) to get a stack trace.</div>
            </blockquote>
            <br>
            This is the stack trace:<br>
            <br>
                 zdotc,               FP=7fffa35afce0<br>
            VecNorm_Seq,         FP=7fffa35afe40<br>
            VecNorm,             FP=7fffa35aff80<br>
            VecNormalize,        FP=7fffa35b0080<br>
            GMREScycle,          FP=7fffa35b02e0<br>
            KSPSolve_GMRES,      FP=7fffa35b0390<br>
            KSPSolve,            FP=7fffa35b1ef0<br>
            main,                FP=7fffa35b31d0<br>
            <br>
            It crashes here:<br>
            <br>
               PetscFunctionBegin;<br>
               if (type == NORM_2 || type == NORM_FROBENIUS) {<br>
                 ierr = VecGetArrayRead(xin,&xx);CHKERRQ(ierr);<br>
                 *z = BLASdot_(&bn,xx,&one,xx,&one);  //
            <<<< crash<br>
            <br>
            The log from valgrind is attached, but it is completely
            filled with stupid MPI messages. How could one filter out
            all that stuff? <br>
            <br>
            Regards,<br>
            Alexander<br>
          </div>
        </blockquote>
      </div>
    </blockquote>
    <br>
  </body>
</html>