<html>
  <head>
    <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <div class="moz-cite-prefix">Hi,<br>
      <br>
      I have combined the momentum linear eqns involving x,y,z into 1
      large matrix. The Poisson eqn is solved using HYPRE strcut format
      so it's not included. I run the code for 50 timesteps (hence 50
      kspsolve) using 96 procs. The log_summary is given below. I have
      some questions:<br>
      <br>
      1. After combining the matrix, I should have only 1 PETSc matrix.
      Why does it says there are 4 matrix, 12 vector etc? <br>
      <br>
      2. I'm looking at the stages which take the longest time. It seems
      that MatAssemblyBegin, VecNorm, VecAssemblyBegin, VecScatterEnd
      have very high ratios. The ratios of some others are also not too
      good (~ 1.6 - 2). So are these stages the reason why my code is
      not scaling well? What can I do to improve it?<br>
      <br>
      Btw, I insert matrix using:<br>
      <br>
      <i>do ijk=ijk_xyz_sta+1,ijk_xyz_end</i><i><br>
      </i><i><br>
      </i><i>    II = ijk - 1</i><i>    !Fortran shift to 0-based</i><i><br>
      </i><i>    </i><i><br>
      </i><i>    call
MatSetValues(A_semi_xyz,1,II,7,int_semi_xyz(ijk,1:7),semi_mat_xyz(ijk,1:7),INSERT_VALUES,ierr)</i><i><br>
      </i><i><br>
      </i><i>end do</i><br>
      <br>
      where ijk_xyz_sta/ijk_xyz_end are the starting/end index<br>
      <br>
      int_semi_xyz(ijk,1:7) stores the 7 column global indices<br>
      <br>
      semi_mat_xyz has the corresponding values.<br>
      <br>
      and I insert vectors using:<br>
      <br>
      call
VecSetValues(b_rhs_semi_xyz,ijk_xyz_end_mz-ijk_xyz_sta_mz,(/ijk_xyz_sta_mz:ijk_xyz_end_mz-1/),q_semi_vect_xyz(ijk_xyz_sta_mz+1:ijk_xyz_end_mz),INSERT_VALUES,ierr)<br>
      <br>
      Thanks!<br>
      <br>
      <i><br>
      </i><br>
      <pre class="moz-signature" cols="72">Yours sincerely,

TAY wee-beng</pre>
      On 30/9/2012 11:30 PM, Jed Brown wrote:<br>
    </div>
    <blockquote
cite="mid:CAM9tzS=D2gzSy6VENT83fRgTuU5V5c-nB5433dy8ysn1gZS42A@mail.gmail.com"
      type="cite">
      <p>You can measure the time spent in Hypre via PCApply and
        PCSetUp, but you can't get finer grained integrated profiling
        because it was not set up that way.</p>
      <div class="gmail_quote">On Sep 30, 2012 3:26 PM, "TAY wee-beng"
        <<a moz-do-not-send="true" href="mailto:zonexo@gmail.com">zonexo@gmail.com</a>>
        wrote:<br type="attribution">
        <blockquote class="gmail_quote" style="margin:0 0 0
          .8ex;border-left:1px #ccc solid;padding-left:1ex">
          <div bgcolor="#FFFFFF" text="#000000">
            <div>On 27/9/2012 1:44 PM, Matthew Knepley wrote:<br>
            </div>
            <blockquote type="cite">On Thu, Sep 27, 2012 at 3:49 AM, TAY
              wee-beng <span dir="ltr"><<a moz-do-not-send="true"
                  href="mailto:zonexo@gmail.com" target="_blank">zonexo@gmail.com</a>></span>
              wrote:<br>
              <div class="gmail_quote">
                <blockquote class="gmail_quote" style="margin:0 0 0
                  .8ex;border-left:1px #ccc solid;padding-left:1ex"> Hi,<br>
                  <br>
                  I'm doing a log summary for my 3d cfd code. I have
                  some questions:<br>
                  <br>
                  1. if I'm solving 3 linear equations using ksp, is the
                  result given in the log summary the total of the 3
                  linear eqns' performance? How can I get the
                  performance for each individual eqn?<br>
                </blockquote>
                <div><br>
                </div>
                <div>Use logging stages: <a moz-do-not-send="true"
href="http://www.mcs.anl.gov/petsc/petsc-dev/docs/manualpages/Profiling/PetscLogStagePush.html"
                    target="_blank">http://www.mcs.anl.gov/petsc/petsc-dev/docs/manualpages/Profiling/PetscLogStagePush.html</a></div>
                <div> </div>
                <blockquote class="gmail_quote" style="margin:0 0 0
                  .8ex;border-left:1px #ccc solid;padding-left:1ex"> 2.
                  If I run my code for 10 time steps, does the log
                  summary gives the total or avg performance/ratio?<br>
                </blockquote>
                <div><br>
                </div>
                <div>Total.</div>
                <div> </div>
                <blockquote class="gmail_quote" style="margin:0 0 0
                  .8ex;border-left:1px #ccc solid;padding-left:1ex"> 3.
                  Besides PETSc, I'm also using HYPRE's native geometric
                  MG (Struct) to solve my Cartesian's grid CFD poisson
                  eqn. Is there any way I can use PETSc's log summary to
                  get HYPRE's performance? If I use boomerAMG thru
                  PETSc, can I get its performance?</blockquote>
                <div><br>
                </div>
                <div>If you mean flops, only if you count them yourself
                  and tell PETSc using <a moz-do-not-send="true"
href="http://www.mcs.anl.gov/petsc/petsc-dev/docs/manualpages/Profiling/PetscLogFlops.html"
                    target="_blank">http://www.mcs.anl.gov/petsc/petsc-dev/docs/manualpages/Profiling/PetscLogFlops.html</a></div>
                <div><br>
                </div>
                <div>This is the disadvantage of using packages that do
                  not properly monitor things :)</div>
                <div><br>
                </div>
                <div>    Matt</div>
                <div> </div>
              </div>
            </blockquote>
            So u mean if I use boomerAMG thru PETSc, there is no proper
            way of evaluating its performance, beside using
            PetscLogFlops?<br>
            <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"> <span><font
                      color="#888888"><br>
                      -- <br>
                      Yours sincerely,<br>
                      <br>
                      TAY wee-beng<br>
                      <br>
                    </font></span></blockquote>
              </div>
              <br>
              <br clear="all">
              <div><br>
              </div>
              -- <br>
              What most experimenters take for granted before they begin
              their experiments is infinitely more interesting than any
              results to which their experiments lead.<br>
              -- Norbert Wiener<br>
            </blockquote>
            <br>
          </div>
        </blockquote>
      </div>
    </blockquote>
    <br>
  </body>
</html>