<html>
  <head>
    <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <div class="moz-cite-prefix">Jed,<br>
      <br>
      I do see the drawbacks of this method, but it seems to have good
      scalability. <br>
      That I want to test.<br>
      <br>
      Thank you.<br>
      <br>
      On 17.08.2012 16:32, Jed Brown wrote:<br>
    </div>
    <blockquote
cite="mid:CAM9tzSkm8T5x5=u2pha-Y83oqhmOVr6tgBm+3EmaK1onH49t3Q@mail.gmail.com"
      type="cite">
      <div class="gmail_quote">On Fri, Aug 17, 2012 at 9:23 AM,
        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:0 0 0
          .8ex;border-left:1px #ccc solid;padding-left:1ex">
          <div bgcolor="#FFFFFF" text="#000000">Right. <br>
            Could you please explain what do you mean by fragile? <br>
          </div>
        </blockquote>
        <div><br>
        </div>
        <div>Fragile means unreliable. The math is not right for Krylov
          methods in general. You are relying on side-effects of
          particular methods and preconditioner side combinations.
          Fragile is the opposite of "robust". For a "robust" method,
          you should start with a compatible discretization and/or
          choose a preconditioner that is stable on the quasi-null space
          and/or filter in a way that is consistent with the Krylov
          method (e.g. as described using FGMRES).</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">
            <div class="im">
              <blockquote type="cite">
                <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">  <br>
                      Looks a bit tricky. The number of fgmres
                      iterations then defines number of the cycles
                      correction will be applied, doesn't it?</div>
                  </blockquote>
                  <div><br>
                  </div>
                  <div>Yes.</div>
                </div>
              </blockquote>
              <br>
            </div>
            If I do this command line only setup:
            <div class="im"><br>
              -pc_type composite -pc_composite_type multiplicative
              -pc_composite_pcs ksp,shell<br>
              <br>
            </div>
            How does petsc know what ShellPCApply routine to take? I
            guess I have to specify this in the code anyway?</div>
        </blockquote>
      </div>
      <br>
      <div>Yeah, to select your preconditioner in this way, you should
        name it preconditioner something meaningful and use
        PCRegisterDynamic() just like the native implementations.</div>
    </blockquote>
    <br>
    <br>
    <pre class="moz-signature" cols="72">-- 
Regards,
Alexander</pre>
  </body>
</html>