<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    Dear Dr. Knepley, <br>
    &nbsp;&nbsp;&nbsp;&nbsp; It is very nice to hear that. I will read the manual. Do we
    have any examples showing its functions?<br>
    <br>
    thank you so much,<br>
    Alan<br>
    On 4/19/2012 9:35 AM, Matthew Knepley wrote:
    <blockquote
cite="mid:CAMYG4GmMBoTo2igMhwZOhYUsi9w6Nn+RW1Ss5cSOuhspMhxvNg@mail.gmail.com"
      type="cite">On Thu, Apr 19, 2012 at 10:18 AM, Zhenglun (Alan) Wei
      <span dir="ltr">&lt;<a moz-do-not-send="true"
          href="mailto:zhenglun.wei@gmail.com">zhenglun.wei@gmail.com</a>&gt;</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">
          <div bgcolor="#FFFFFF" text="#000000">
            <div>
              <div>
                <div>"</div>
                <div>&nbsp; &nbsp; TESTVAR ***a, ***b, ***c;</div>
                <div>&nbsp; &nbsp; TESTVAR **aa, **bb, **cc;&nbsp;</div>
                <div>&nbsp; &nbsp; TESTVAR *arraya, *arrayb, *arrayc;</div>
              </div>
              <div><br>
              </div>
              <div>&nbsp; &nbsp; arraya = (TESTVAR*) calloc(SIZE*SIZE*SIZE,
                sizeof(TESTVAR));</div>
              <div>&nbsp; &nbsp; arrayb = (TESTVAR*) calloc(SIZE*SIZE*SIZE,
                sizeof(TESTVAR));</div>
              <div>&nbsp; &nbsp; arrayc = (TESTVAR*) calloc(SIZE*SIZE*SIZE,
                sizeof(TESTVAR));</div>
              <div><br>
              </div>
              <div>&nbsp; &nbsp; aa =(TESTVAR**) calloc(SIZE*SIZE,
                sizeof(TESTVAR*));</div>
              <div>&nbsp; &nbsp; bb =(TESTVAR**) calloc(SIZE*SIZE,
                sizeof(TESTVAR*));</div>
              <div>&nbsp; &nbsp; cc =(TESTVAR**) calloc(SIZE*SIZE,
                sizeof(TESTVAR*));</div>
              <div>&nbsp; &nbsp;&nbsp;</div>
              <div>&nbsp; &nbsp; for(i = 0; i &lt; SIZE*SIZE; i++) {</div>
              <div>&nbsp; &nbsp; &nbsp; aa[i] = &amp;arraya[i*SIZE];</div>
              <div>&nbsp; &nbsp; &nbsp; bb[i] = &amp;arrayb[i*SIZE];</div>
              <div>&nbsp; &nbsp; &nbsp; cc[i] = &amp;arrayc[i*SIZE];&nbsp;</div>
              <div>&nbsp; &nbsp; }</div>
              <div><br>
              </div>
              <div>&nbsp; &nbsp; a =(TESTVAR***) calloc(SIZE*SIZE,
                sizeof(TESTVAR**));</div>
              <div>&nbsp; &nbsp; b =(TESTVAR***) calloc(SIZE*SIZE,
                sizeof(TESTVAR**));</div>
              <div>&nbsp; &nbsp; c =(TESTVAR***) calloc(SIZE*SIZE,
                sizeof(TESTVAR**));</div>
              <div>&nbsp; &nbsp;&nbsp;</div>
              <div>&nbsp; &nbsp; for(i = 0; i &lt; SIZE; i++) {</div>
              <div>&nbsp; &nbsp; &nbsp; a[i] = &amp;aa[i*SIZE];</div>
              <div>&nbsp; &nbsp; &nbsp; b[i] = &amp;bb[i*SIZE];</div>
              <div>&nbsp; &nbsp; &nbsp; c[i] = &amp;cc[i*SIZE];</div>
              <div>&nbsp; &nbsp; }</div>
              <div>"</div>
              <div>&nbsp; It works. However, I wonder if there is any other
                good ideas for 3D problem other than this kinda of
                'two-layer' approach.</div>
            </div>
            <div><br>
            </div>
            <div><b><u>What is the reason for not using DMDA?</u><br>
              </b>In 2D, I established a 2D array for data communication
              between nodes by using MPI derived data type. It allows me
              to easily communicate both contiguous (i.e.
              MPI_TYPE_CONTIGUOUS) and non-contiguous (i.e.
              MPI_TYPE_VECTOR) data. That is why I use this similar
              approach in 3D, though an additional data type, i.e.
              MPI_TYPE_INDEXED, need to be used. Does DMDA have those
              type of function or derived data type?</div>
          </div>
        </blockquote>
        <div><br>
        </div>
        <div>It definitely does communication between the local pieces.
          Do you want something else?</div>
        <div>&nbsp;</div>
        <blockquote class="gmail_quote" style="margin:0 0 0
          .8ex;border-left:1px #ccc solid;padding-left:1ex">
          <div bgcolor="#FFFFFF" text="#000000">
            <blockquote class="gmail_quote" style="margin:0 0 0
              .8ex;border-left:1px #ccc solid;padding-left:1ex">
              <div>
                <div>"2, I have a little question on PETSc about 3D
                  processor ordering. Does PETSc have any function
                  giving me the nodes/rank number of neighboring
                  nodes/ranks? Are those 'Application Ordering'
                  functions applicable for my case?"</div>
              </div>
            </blockquote>
            <div><br>
            </div>
            <div><u><b>What do you mean by neighboring? If it is jsut
                  stencil neighbors, then use a local vector.</b></u></div>
            <div>When I send and receive data with MPI_Send and
              MPI_RECV, I need provide the 'destination' (in MPI_Send
              refer to'<a moz-do-not-send="true"
href="http://www.mcs.anl.gov/research/projects/mpi/www/www3/MPI_Send.html"
                target="_blank">http://www.mcs.anl.gov/research/projects/mpi/www/www3/MPI_Send.html</a>')

              and 'source' (in MPI_RECV refer to'<a
                moz-do-not-send="true"
href="http://www.mcs.anl.gov/research/projects/mpi/www/www3/MPI_Recv.html"
                target="_blank">http://www.mcs.anl.gov/research/projects/mpi/www/www3/MPI_Recv.html</a>').

              In a 2D problem with Cartesian grid, 4 processes divide
              the whole domain to 4 sub-domain. <br>
              ---------------------------- <br>
              &nbsp;&nbsp;&nbsp;&nbsp;&nbsp; 2&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; |&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; 3&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; |<br>
              ----------------------------<br>
              &nbsp;&nbsp;&nbsp;&nbsp;&nbsp; 0&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; |&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; 1&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; |<br>
              ---------------------------<br>
              &nbsp;Then, for node 1, the neighboring nodes are '0' and '3',
              which '0' is the left node and '3' is the top node. I
              wonder if PETSc has any function that I can call to obtain
              those neighboring nodes so that I do not need to construct
              my function. <br>
            </div>
          </div>
        </blockquote>
        <div><br>
        </div>
        <div>Yes, it looks like you should just use a DMDA. See the
          manual section.</div>
        <div><br>
        </div>
        <div>&nbsp; &nbsp;Matt</div>
        <div>&nbsp;</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> I'm sorry for confusing you. <br>
              <br>
              thanks in advance,<br>
              Alan <br>
            </div>
            <br>
            On 4/19/2012 4:52 AM, Matthew Knepley wrote:
            <blockquote type="cite">On Wed, Apr 18, 2012 at 3:52 PM,
              Alan Wei <span dir="ltr">&lt;<a moz-do-not-send="true"
                  href="mailto:zhenglun.wei@gmail.com" target="_blank">zhenglun.wei@gmail.com</a>&gt;</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">
                  Dear all,
                  <div>&nbsp; &nbsp; I hope you're having a nice day. I have a
                    further question on this issue in 3D.</div>
                  <div>1, Following the idea of Dr. Brown and Dr.
                    Knepley, I finished a 2D test, which works very
                    fine. Here, I did it in 3D by</div>
                  <div>
                    <div>
                      <div>"</div>
                      <div>&nbsp; &nbsp; TESTVAR ***a, ***b, ***c;</div>
                      <div>&nbsp; &nbsp; TESTVAR **aa, **bb, **cc;&nbsp;</div>
                      <div>&nbsp; &nbsp; TESTVAR *arraya, *arrayb, *arrayc;</div>
                    </div>
                    <div><br>
                    </div>
                    <div>&nbsp; &nbsp; arraya = (TESTVAR*) calloc(SIZE*SIZE*SIZE,
                      sizeof(TESTVAR));</div>
                    <div>&nbsp; &nbsp; arrayb = (TESTVAR*) calloc(SIZE*SIZE*SIZE,
                      sizeof(TESTVAR));</div>
                    <div>&nbsp; &nbsp; arrayc = (TESTVAR*) calloc(SIZE*SIZE*SIZE,
                      sizeof(TESTVAR));</div>
                    <div><br>
                    </div>
                    <div>&nbsp; &nbsp; aa =(TESTVAR**) calloc(SIZE*SIZE,
                      sizeof(TESTVAR*));</div>
                    <div>&nbsp; &nbsp; bb =(TESTVAR**) calloc(SIZE*SIZE,
                      sizeof(TESTVAR*));</div>
                    <div>&nbsp; &nbsp; cc =(TESTVAR**) calloc(SIZE*SIZE,
                      sizeof(TESTVAR*));</div>
                    <div>&nbsp; &nbsp;&nbsp;</div>
                    <div>&nbsp; &nbsp; for(i = 0; i &lt; SIZE*SIZE; i++) {</div>
                    <div>&nbsp; &nbsp; &nbsp; aa[i] = &amp;arraya[i*SIZE];</div>
                    <div>&nbsp; &nbsp; &nbsp; bb[i] = &amp;arrayb[i*SIZE];</div>
                    <div>&nbsp; &nbsp; &nbsp; cc[i] = &amp;arrayc[i*SIZE];&nbsp;</div>
                    <div>&nbsp; &nbsp; }</div>
                    <div><br>
                    </div>
                    <div>&nbsp; &nbsp; a =(TESTVAR***) calloc(SIZE*SIZE,
                      sizeof(TESTVAR**));</div>
                    <div>&nbsp; &nbsp; b =(TESTVAR***) calloc(SIZE*SIZE,
                      sizeof(TESTVAR**));</div>
                    <div>&nbsp; &nbsp; c =(TESTVAR***) calloc(SIZE*SIZE,
                      sizeof(TESTVAR**));</div>
                    <div>&nbsp; &nbsp;&nbsp;</div>
                    <div>&nbsp; &nbsp; for(i = 0; i &lt; SIZE; i++) {</div>
                    <div>&nbsp; &nbsp; &nbsp; a[i] = &amp;aa[i*SIZE];</div>
                    <div>&nbsp; &nbsp; &nbsp; b[i] = &amp;bb[i*SIZE];</div>
                    <div>&nbsp; &nbsp; &nbsp; c[i] = &amp;cc[i*SIZE];</div>
                    <div>&nbsp; &nbsp; }</div>
                    <div>"</div>
                    <div>&nbsp; It works. However, I wonder if there is any
                      other good ideas for 3D problem other than this
                      kinda of 'two-layer' approach.</div>
                  </div>
                </blockquote>
                <div><br>
                </div>
                <div>What is the reason for not using DMDA?</div>
                <div>&nbsp;</div>
                <blockquote class="gmail_quote" style="margin:0 0 0
                  .8ex;border-left:1px #ccc solid;padding-left:1ex">
                  <div>
                    <div>2, I have a little question on PETSc about 3D
                      processor ordering. Does PETSc have any function
                      giving me the nodes/rank number of neighboring
                      nodes/ranks? Are those 'Application Ordering'
                      functions applicable for my case?</div>
                  </div>
                </blockquote>
                <div><br>
                </div>
                <div>What do you mean by neighboring? If it is jsut
                  stencil neighbors, then use a local vector.</div>
                <div><br>
                </div>
                <div>&nbsp; &nbsp;Matt</div>
                <div>&nbsp;</div>
                <blockquote class="gmail_quote" style="margin:0 0 0
                  .8ex;border-left:1px #ccc solid;padding-left:1ex">
                  <div>
                    <div>thanks,</div>
                    <div>Alan</div>
                    <br>
                    <div class="gmail_quote">On Fri, Apr 13, 2012 at
                      5:41 PM, Jed Brown <span dir="ltr">&lt;<a
                          moz-do-not-send="true"
                          href="mailto:jedbrown@mcs.anl.gov"
                          target="_blank">jedbrown@mcs.anl.gov</a>&gt;</span>
                      wrote:<br>
                      <blockquote class="gmail_quote" style="margin:0 0
                        0 .8ex;border-left:1px #ccc
                        solid;padding-left:1ex">
                        <div>
                          <div class="gmail_quote">On Fri, Apr 13, 2012
                            at 17:38, Zhenglun (Alan) Wei <span
                              dir="ltr">&lt;<a moz-do-not-send="true"
                                href="mailto:zhenglun.wei@gmail.com"
                                target="_blank">zhenglun.wei@gmail.com</a>&gt;</span>
                            wrote:<br>
                            <blockquote class="gmail_quote"
                              style="margin:0 0 0 .8ex;border-left:1px
                              #ccc solid;padding-left:1ex"> &nbsp;&nbsp;&nbsp; I have a
                              final question on it. Is it taken a lot of
                              memory for doing this? As I understand,
                              pointers won't occupy many memories and it
                              works like an alias. It will not, to my
                              limit knowledge, take much extra memory by
                              doing this. </blockquote>
                          </div>
                          <br>
                        </div>
                        <div>A pointer takes about as much space as a
                          floating point value, so that array of
                          pointers costs about 1*N compared to the N*N
                          matrix.</div>
                      </blockquote>
                    </div>
                    <br>
                  </div>
                </blockquote>
              </div>
              <br>
              <br clear="all">
              <span class="HOEnZb"><font color="#888888">
                  <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>
                </font></span></blockquote>
            <br>
          </div>
        </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>
  </body>
</html>