<div dir="ltr">Okay, I ran it. Something seems off with the mesh. First, I cannot simply explain the partition. The number of shared vertices and edges<div>does not seem to come from a straight cut. Second, the mesh look scrambled on output.</div><div><br></div><div>  Thanks,</div><div><br></div><div>    Matt</div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Sun, Oct 24, 2021 at 11:49 PM Eric Chamberland <<a href="mailto:Eric.Chamberland@giref.ulaval.ca">Eric.Chamberland@giref.ulaval.ca</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>
    <p>Hi Matthew,</p>
    <p>ok, I started back from your ex44.c example and added the global
      array of coordinates.  I just have to code the creation of the
      local coordinates now.</p>
    <p>Eric<br>
    </p>
    <div>On 2021-10-20 6:55 p.m., Matthew
      Knepley wrote:<br>
    </div>
    <blockquote type="cite">
      
      <div dir="ltr">
        <div dir="ltr">On Wed, Oct 20, 2021 at 3:06 PM Eric Chamberland
          <<a href="mailto:Eric.Chamberland@giref.ulaval.ca" target="_blank">Eric.Chamberland@giref.ulaval.ca</a>>
          wrote:<br>
        </div>
        <div class="gmail_quote">
          <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
            <div>
              <p>Hi Matthew,</p>
              <p>we tried to reproduce the error in a simple example.</p>
              <p>The context is the following: We hard coded the mesh
                and initial partition into the code (see sConnectivity
                and sInitialPartition) for 2 ranks and try to create a
                section in order to use the DMPlexNaturalToGlobalBegin
                function to retreive our initial element numbers.</p>
              <p>Now the call to DMPlexDistribute give different errors
                depending on what type of component we ask the field to
                be created.  For our objective, we would like a global
                field to be created on elements only (like a P0
                interpolation).<br>
              </p>
              <p>We now have the following error generated:</p>
              <p>[0]PETSC ERROR: --------------------- Error Message
                --------------------------------------------------------------<br>
                [0]PETSC ERROR: Petsc has generated inconsistent data<br>
                [0]PETSC ERROR: Inconsistency in indices, 18 should be
                17<br>
                [0]PETSC ERROR: See <a href="https://www.mcs.anl.gov/petsc/documentation/faq.html" target="_blank">https://www.mcs.anl.gov/petsc/documentation/faq.html</a>
                for trouble shooting.<br>
                [0]PETSC ERROR: Petsc Release Version 3.15.0, Mar 30,
                2021 <br>
                [0]PETSC ERROR: ./bug on a  named rohan by ericc Wed Oct
                20 14:52:36 2021<br>
                [0]PETSC ERROR: Configure options
                --prefix=/opt/petsc-3.15.0_debug_openmpi-4.1.0_gcc7
                --with-mpi-compilers=1
                --with-mpi-dir=/opt/openmpi-4.1.0_gcc7
                --with-cxx-dialect=C++14 --with-make-np=12
                --with-shared-libraries=1 --with-debugging=yes
                --with-memalign=64 --with-visibility=0
                --with-64-bit-indices=0 --download-ml=yes
                --download-mumps=yes --download-superlu=yes
                --download-hpddm=yes --download-slepc=yes
                --download-superlu_dist=yes --download-parmetis=yes
                --download-ptscotch=yes --download-metis=yes
                --download-strumpack=yes --download-suitesparse=yes
                --download-hypre=yes
                --with-blaslapack-dir=/opt/intel/oneapi/mkl/2021.1.1/env/../lib/intel64
--with-mkl_pardiso-dir=/opt/intel/oneapi/mkl/2021.1.1/env/..
                --with-mkl_cpardiso-dir=/opt/intel/oneapi/mkl/2021.1.1/env/..
                --with-scalapack=1
                --with-scalapack-include=/opt/intel/oneapi/mkl/2021.1.1/env/../include
--with-scalapack-lib="-L/opt/intel/oneapi/mkl/2021.1.1/env/../lib/intel64
                -lmkl_scalapack_lp64 -lmkl_blacs_openmpi_lp64"<br>
                [0]PETSC ERROR: #1 PetscSFCreateSectionSF() at
                /tmp/ompi-opt/petsc-3.15.0-debug/src/vec/is/sf/utils/sfutils.c:409<br>
                [0]PETSC ERROR: #2 DMPlexCreateGlobalToNaturalSF() at
                /tmp/ompi-opt/petsc-3.15.0-debug/src/dm/impls/plex/plexnatural.c:184<br>
                [0]PETSC ERROR: #3 DMPlexDistribute() at
                /tmp/ompi-opt/petsc-3.15.0-debug/src/dm/impls/plex/plexdistribute.c:1733<br>
                [0]PETSC ERROR: #4 main() at bug_section.cc:159<br>
                [0]PETSC ERROR: No PETSc Option Table entries<br>
                [0]PETSC ERROR: ----------------End of Error Message
                -------send entire error message to <a href="mailto:petsc-maint@mcs.anl.gov" target="_blank">petsc-maint@mcs.anl.gov</a>----------<br>
                <br>
              </p>
              <p>Hope the attached code is self-explaining, note that to
                make it short, we have not included the final part of
                it, just the buggy part we are encountering right now...</p>
              <p>Thanks for your insights,</p>
            </div>
          </blockquote>
          <div>Thanks for making the example. I tweaked it slightly. I
            put in a test case that just makes a parallel 7 x 10 quad
            mesh. This works</div>
          <div>fine. Thus I think it must be something connected with
            the original mesh. It is hard to get a handle on it without
            the coordinates.</div>
          <div>Do you think you could put the coordinate array in? I
            have added the code to load them (see attached file).</div>
          <div><br>
          </div>
          <div>  Thanks,</div>
          <div><br>
          </div>
          <div>     Matt </div>
          <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
            <div>
              <p>Eric<br>
              </p>
              <div>On 2021-10-06 9:23 p.m., Matthew Knepley wrote:<br>
              </div>
              <blockquote type="cite">
                <div dir="ltr">
                  <div dir="ltr">On Wed, Oct 6, 2021 at 5:43 PM Eric
                    Chamberland <<a href="mailto:Eric.Chamberland@giref.ulaval.ca" target="_blank">Eric.Chamberland@giref.ulaval.ca</a>>
                    wrote:<br>
                  </div>
                  <div class="gmail_quote">
                    <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                      <div>
                        <p>Hi Matthew,</p>
                        <p>we tried to use that.  Now, we discovered
                          that:</p>
                        <p>1- even if we "ask" for sfNatural creation
                          with DMSetUseNatural, it is not created
                          because DMPlexCreateGlobalToNaturalSF looks
                          for a "section": this is not documented in
                          DMSetUseNaturalso we are asking ourselfs: "is
                          this a permanent feature or a temporary
                          situation?"<br>
                        </p>
                      </div>
                    </blockquote>
                    <div>I think explaining this will help clear up a
                      lot.</div>
                    <div><br>
                    </div>
                    <div>What the Natural2Global map does is permute a
                      solution vector into the ordering that it would
                      have had prior to mesh distribution.</div>
                    <div>Now, in order to do this permutation, I need to
                      know the original (global) data layout. If it is
                      not specified _before_ distribution, we</div>
                    <div>cannot build the permutation.  The section
                      describes the data layout, so I need it before
                      distribution.</div>
                    <div><br>
                    </div>
                    <div>I cannot think of another way that you would
                      implement this, but if you want something else,
                      let me know.</div>
                    <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                      <div>
                        <p>2- We then tried to create a "section" in
                          different manners: we took the code into the
                          example petsc/src/dm/impls/plex/tests/ex15.c. 
                          However, we ended up with a segfault:</p>
                        <p>corrupted size vs. prev_size<br>
                          [rohan:07297] *** Process received signal ***<br>
                          [rohan:07297] Signal: Aborted (6)<br>
                          [rohan:07297] Signal code:  (-6)<br>
                          [rohan:07297] [ 0]
                          /lib64/libpthread.so.0(+0x13f80)[0x7f6f13be3f80]<br>
                          [rohan:07297] [ 1]
                          /lib64/libc.so.6(gsignal+0x10b)[0x7f6f109b718b]<br>
                          [rohan:07297] [ 2]
                          /lib64/libc.so.6(abort+0x175)[0x7f6f109b8585]<br>
                          [rohan:07297] [ 3]
                          /lib64/libc.so.6(+0x7e2f7)[0x7f6f109fb2f7]<br>
                          [rohan:07297] [ 4]
                          /lib64/libc.so.6(+0x857ea)[0x7f6f10a027ea]<br>
                          [rohan:07297] [ 5]
                          /lib64/libc.so.6(+0x86036)[0x7f6f10a03036]<br>
                          [rohan:07297] [ 6]
                          /lib64/libc.so.6(+0x861a3)[0x7f6f10a031a3]<br>
                          [rohan:07297] [ 7]
                          /lib64/libc.so.6(+0x88740)[0x7f6f10a05740]<br>
                          [rohan:07297] [ 8]
                          /lib64/libc.so.6(__libc_malloc+0x1b8)[0x7f6f10a070c8]<br>
                          [rohan:07297] [ 9]
                          /lib64/libc.so.6(__backtrace_symbols+0x134)[0x7f6f10a8b064]<br>
                          [rohan:07297] [10]
/home/mefpp_ericc/GIREF/bin/MEF++.dev(_Z12reqBacktraceRNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE+0x4e)[0x4538ce]<br>
                          [rohan:07297] [11]
/home/mefpp_ericc/GIREF/bin/MEF++.dev(_Z15attacheDebuggerv+0x120)[0x4523c0]<br>
                          [rohan:07297] [12]
/home/mefpp_ericc/GIREF/lib/libgiref_dev_Util.so(traitementSignal+0x612)[0x7f6f28f503a2]<br>
                          [rohan:07297] [13]
                          /lib64/libc.so.6(+0x3a210)[0x7f6f109b7210]</p>
                        <p>[rohan:07297] [14]
/opt/petsc-3.15.0_debug_openmpi-4.1.0_gcc7/lib/libpetsc.so.3.15(PetscTrMallocDefault+0x6fd)[0x7f6f22f1b8ed]<br>
                          [rohan:07297] [15]
/opt/petsc-3.15.0_debug_openmpi-4.1.0_gcc7/lib/libpetsc.so.3.15(PetscMallocA+0x5cd)[0x7f6f22f19c2d]<br>
                          [rohan:07297] [16]
/opt/petsc-3.15.0_debug_openmpi-4.1.0_gcc7/lib/libpetsc.so.3.15(PetscSFCreateSectionSF+0xb48)[0x7f6f23268e18]<br>
                          [rohan:07297] [17]
/opt/petsc-3.15.0_debug_openmpi-4.1.0_gcc7/lib/libpetsc.so.3.15(DMPlexCreateGlobalToNaturalSF+0x13b2)[0x7f6f241a5602]<br>
                          [rohan:07297] [18]
/opt/petsc-3.15.0_debug_openmpi-4.1.0_gcc7/lib/libpetsc.so.3.15(DMPlexDistribute+0x39b1)[0x7f6f23fdca21]<br>
                        </p>
                      </div>
                    </blockquote>
                    <div>I am not sure what happened here, but if you
                      could send a sample code, I will figure it out. </div>
                    <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                      <div>
                        <p>If we do not create a section, the call to
                          DMPlexDistribute is successful, but
                          DMPlexGetGlobalToNaturalSF return a null SF
                          pointer...<br>
                        </p>
                      </div>
                    </blockquote>
                    <div>Yes, it just ignores it in this case because it
                      does not have a global layout. </div>
                    <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                      <div>
                        <p> </p>
                        <p>Here are the operations we are calling ( this
                          is almost the code we are using, I just
                          removed verifications and creation of the
                          connectivity which use our parallel structure
                          and code):<br>
                        </p>
                        <p>===========</p>
                        <p>  PetscInt* lCells      = 0;<br>
                            PetscInt  lNumCorners = 0;<br>
                            PetscInt  lDimMail    = 0;<br>
                            PetscInt  lnumCells   = 0;<br>
                          <br>
                            //At this point we create the cells for
                          PETSc expected input for
                          DMPlexBuildFromCellListParallel and set
                          lNumCorners, lDimMail and lnumCells to correct
                          values.<br>
                            ...<br>
                            <br>
                            DM       lDMBete = 0<br>
                            DMPlexCreate(lMPIComm,&lDMBete);<br>
                          <br>
                            DMSetDimension(lDMBete, lDimMail);<br>
                          <br>
                            DMPlexBuildFromCellListParallel(lDMBete,<br>
                                                            lnumCells,<br>
                                                           
                          PETSC_DECIDE,<br>
                                                           
                          pLectureElementsLocaux.reqNbTotalSommets(),<br>
                                                            lNumCorners,<br>
                                                            lCells,<br>
                                                            PETSC_NULL);<br>
                          <br>
                            DM lDMBeteInterp = 0;<br>
                            DMPlexInterpolate(lDMBete,
                          &lDMBeteInterp);<br>
                            DMDestroy(&lDMBete);<br>
                            lDMBete = lDMBeteInterp;<br>
                          <br>
                            DMSetUseNatural(lDMBete,PETSC_TRUE);<br>
                          <br>
                            PetscSF lSFMigrationSansOvl = 0;<br>
                            PetscSF lSFMigrationOvl = 0;<br>
                            DM lDMDistribueSansOvl = 0;<br>
                            DM lDMAvecOverlap = 0;<br>
                          <br>
                            PetscPartitioner lPart;<br>
                            DMPlexGetPartitioner(lDMBete, &lPart);<br>
                            PetscPartitionerSetFromOptions(lPart);<br>
                          <br>
                            PetscSection   section;<br>
                            PetscInt       numFields   = 1;<br>
                            PetscInt       numBC       = 0;<br>
                            PetscInt       numComp[1]  = {1};<br>
                            PetscInt       numDof[4]   = {1, 0, 0, 0};<br>
                            PetscInt       bcFields[1] = {0};<br>
                            IS             bcPoints[1] = {NULL};<br>
                          <br>
                            DMSetNumFields(lDMBete, numFields);<br>
                          <br>
                            DMPlexCreateSection(lDMBete, NULL, numComp,
                          numDof, numBC, bcFields, bcPoints, NULL, NULL,
                          &section);<br>
                            DMSetLocalSection(lDMBete, section);<br>
                          <br>
                            DMPlexDistribute(lDMBete, 0,
                          &lSFMigrationSansOvl,
                          &lDMDistribueSansOvl); // segfault!<br>
                        </p>
                        <p>===========<br>
                        </p>
                        <p>So we have other question/remarks:</p>
                        <p>3- Maybe PETSc expect something specific that
                          is missing/not verified: for example, we
                          didn't gave any coordinates since we just want
                          to partition and compute overlap for the
                          mesh... and then recover our element numbers
                          in a "simple way"</p>
                        <p>4- We are telling ourselves it is somewhat a
                          "big price to pay" to have to build an unused
                          section to have the global to natural ordering
                          set ?  Could this requirement be avoided?</p>
                      </div>
                    </blockquote>
                    <div>I don't think so. There would have to be _some_
                      way of describing your data layout in terms of
                      mesh points, and I do not see how you could use
                      less memory doing that. </div>
                    <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                      <div>
                        <p>5- Are there any improvement towards our
                          usages in 3.16 release?</p>
                      </div>
                    </blockquote>
                    <div>Let me try and run the code above.</div>
                    <div><br>
                    </div>
                    <div>  Thanks,</div>
                    <div><br>
                    </div>
                    <div>     Matt </div>
                    <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                      <div>
                        <p>Thanks,</p>
                        <p>Eric</p>
                        <p><br>
                        </p>
                        <div>On 2021-09-29 7:39 p.m., Matthew Knepley
                          wrote:<br>
                        </div>
                        <blockquote type="cite">
                          <div dir="ltr">
                            <div dir="ltr">On Wed, Sep 29, 2021 at 5:18
                              PM Eric Chamberland <<a href="mailto:Eric.Chamberland@giref.ulaval.ca" target="_blank">Eric.Chamberland@giref.ulaval.ca</a>>
                              wrote:<br>
                            </div>
                            <div class="gmail_quote">
                              <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Hi,<br>
                                <br>
                                I come back with _almost_ the original
                                question:<br>
                                <br>
                                I would like to add an integer
                                information (*our* original element <br>
                                number, not petsc one) on each element
                                of the DMPlex I create with <br>
                                DMPlexBuildFromCellListParallel.<br>
                                <br>
                                I would like this interger to be
                                distribruted by or the same way <br>
                                DMPlexDistribute distribute the mesh.<br>
                                <br>
                                Is it possible to do this?<br>
                              </blockquote>
                              <div><br>
                              </div>
                              <div>I think we already have support for
                                what you want. If you call</div>
                              <div><br>
                              </div>
                              <div>  <a href="https://petsc.org/main/docs/manualpages/DM/DMSetUseNatural.html" target="_blank">https://petsc.org/main/docs/manualpages/DM/DMSetUseNatural.html</a></div>
                              <div><br>
                              </div>
                              <div>before DMPlexDistribute(), it will
                                compute a PetscSF encoding the global to
                                natural map. You</div>
                              <div>can get it with</div>
                              <div><br>
                              </div>
                              <div>  <a href="https://petsc.org/main/docs/manualpages/DMPLEX/DMPlexGetGlobalToNaturalSF.html" target="_blank">https://petsc.org/main/docs/manualpages/DMPLEX/DMPlexGetGlobalToNaturalSF.html</a></div>
                              <div><br>
                              </div>
                              <div>and use it with</div>
                              <div><br>
                              </div>
                              <div>  <a href="https://petsc.org/main/docs/manualpages/DMPLEX/DMPlexGlobalToNaturalBegin.html" target="_blank">https://petsc.org/main/docs/manualpages/DMPLEX/DMPlexGlobalToNaturalBegin.html</a></div>
                              <div><br>
                              </div>
                              <div>Is this sufficient?</div>
                              <div><br>
                              </div>
                              <div>  Thanks,</div>
                              <div><br>
                              </div>
                              <div>     Matt</div>
                              <div> </div>
                              <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                                Thanks,<br>
                                <br>
                                Eric<br>
                                <br>
                                On 2021-07-14 1:18 p.m., Eric
                                Chamberland wrote:<br>
                                > Hi,<br>
                                ><br>
                                > I want to use DMPlexDistribute from
                                PETSc for computing overlapping <br>
                                > and play with the different
                                partitioners supported.<br>
                                ><br>
                                > However, after calling
                                DMPlexDistribute, I noticed the elements
                                are <br>
                                > renumbered and then the original
                                number is lost.<br>
                                ><br>
                                > What would be the best way to keep
                                track of the element renumbering?<br>
                                ><br>
                                > a) Adding an optional parameter to
                                let the user retrieve a vector or <br>
                                > "IS" giving the old number?<br>
                                ><br>
                                > b) Adding a DMLabel (seems a wrong
                                good solution)<br>
                                ><br>
                                > c) Other idea?<br>
                                ><br>
                                > Of course, I don't want to loose
                                performances with the need of this <br>
                                > "mapping"...<br>
                                ><br>
                                > Thanks,<br>
                                ><br>
                                > Eric<br>
                                ><br>
                                -- <br>
                                Eric Chamberland, ing., M. Ing<br>
                                Professionnel de recherche<br>
                                GIREF/Université Laval<br>
                                (418) 656-2131 poste 41 22 42<br>
                                <br>
                              </blockquote>
                            </div>
                            <br clear="all">
                            <div><br>
                            </div>
                            -- <br>
                            <div dir="ltr">
                              <div dir="ltr">
                                <div>
                                  <div dir="ltr">
                                    <div>
                                      <div dir="ltr">
                                        <div>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</div>
                                        <div><br>
                                        </div>
                                        <div><a href="http://www.cse.buffalo.edu/~knepley/" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
                                        </div>
                                      </div>
                                    </div>
                                  </div>
                                </div>
                              </div>
                            </div>
                          </div>
                        </blockquote>
                        <pre cols="72">-- 
Eric Chamberland, ing., M. Ing
Professionnel de recherche
GIREF/Université Laval
(418) 656-2131 poste 41 22 42</pre>
                      </div>
                    </blockquote>
                  </div>
                  <br clear="all">
                  <div><br>
                  </div>
                  -- <br>
                  <div dir="ltr">
                    <div dir="ltr">
                      <div>
                        <div dir="ltr">
                          <div>
                            <div dir="ltr">
                              <div>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</div>
                              <div><br>
                              </div>
                              <div><a href="http://www.cse.buffalo.edu/~knepley/" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
                              </div>
                            </div>
                          </div>
                        </div>
                      </div>
                    </div>
                  </div>
                </div>
              </blockquote>
              <pre cols="72">-- 
Eric Chamberland, ing., M. Ing
Professionnel de recherche
GIREF/Université Laval
(418) 656-2131 poste 41 22 42</pre>
            </div>
          </blockquote>
        </div>
        <br clear="all">
        <div><br>
        </div>
        -- <br>
        <div dir="ltr">
          <div dir="ltr">
            <div>
              <div dir="ltr">
                <div>
                  <div dir="ltr">
                    <div>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</div>
                    <div><br>
                    </div>
                    <div><a href="http://www.cse.buffalo.edu/~knepley/" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
                    </div>
                  </div>
                </div>
              </div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    <pre cols="72">-- 
Eric Chamberland, ing., M. Ing
Professionnel de recherche
GIREF/Université Laval
(418) 656-2131 poste 41 22 42</pre>
  </div>

</blockquote></div><br clear="all"><div><br></div>-- <br><div dir="ltr" class="gmail_signature"><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div>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</div><div><br></div><div><a href="http://www.cse.buffalo.edu/~knepley/" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br></div></div></div></div></div></div></div>