<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <div class="moz-cite-prefix">Thanks, I have corrected the shift
      amount option and I get the same error as when using the
      -st_pc_factor_shift_amount 1 <br>
      <br>
      I have been using the PETSc IO module to convert my matrix from
      sparse coordinate format into PETSc binary. I am guessing there is
      some problem with this step. I use the function:<br>
      PetscBinaryIO.PetscBinaryIO().writeMatSciPy( fh, matrix) <br>
      to convert into Petsc binary format. Is there something wrong with
      this? Do I need to manually put zeros into the diagonal elements
      of zero rows of B?<br>
      <br>
      When I solve the problem using scipy I use dense matrices. There
      are definitely entries in the matrix A at row 2395. I used python
      to confirm this before I convert it to petsc format:<br>
      <br>
                  column | value <br>
      LHS:    1795  10.9955742876j<br>
                1198    (76+0j)<br>
                1196    (72+0j)<br>
                595      0.02j<br>
      <br>
      However, the matrix is all zeros at row 2395 in B. As far as I
      understand, this doesn't make my problem singular. I am sure that
      the problem is not rank deficient when I solve it in scipy.<br>
      <br>
      Thanks again,<br>
      Toby<br>
      <br>
      <br>
      On 25/07/2013 12:51, Matthew Knepley wrote:<br>
    </div>
    <blockquote
cite="mid:CAMYG4GmkbYmfUyAo9JFJpdg+2mVeRgb=QzBtUbnR0UjQ8tnPQw@mail.gmail.com"
      type="cite">
      <div dir="ltr">On Thu, Jul 25, 2013 at 6:46 AM, Toby <span
          dir="ltr"><<a moz-do-not-send="true"
            href="mailto:T.W.Searle@sms.ed.ac.uk" target="_blank">T.W.Searle@sms.ed.ac.uk</a>></span>
        wrote:<br>
        <div class="gmail_extra">
          <div class="gmail_quote">
            <blockquote class="gmail_quote" style="margin:0px 0px 0px
0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><br>
              Ok, it looks like the -st_pc_factor_shift_type option is
              now being used, thanks. Unfortunately it hasn't fixed the
              problem.<br>
              <br>
              if I use the NONZERO type and a shift amount, the shift
              amount option is not recognised and I still get the zero
              pivot error. If I use the POSITIVE_DEFINITE type I also
              get the pivot error. Output is below<br>
              <br>
              Thanks,<br>
              Toby<br>
              <br>
              tobymac:SLEPC toby$ mpiexec ./ex7 -f1
              LHS-N7-M40-Re0.0-b0.1-Wi5.0-amp0.02.petsc  -f2
              RHS-N7-M40-Re0.0-b0.1-Wi5.0-amp0.02.petsc -st_pc_type lu
              -st_pc_factor_shift_type NONZERO -st_pc_shift_type_amount
              1<br>
              <br>
              Generalized eigenproblem stored in file.<br>
              <br>
               Reading COMPLEX matrices from binary files...<br>
              [0]PETSC ERROR: --------------------- Error Message
              ------------------------------------<br>
              [0]PETSC ERROR: Detected zero pivot in LU factorization:<br>
              see <a moz-do-not-send="true"
                href="http://www.mcs.anl.gov/petsc/documentation/faq.html#ZeroPivot"
                target="_blank">http://www.mcs.anl.gov/petsc/documentation/faq.html#ZeroPivot</a>!<br>
              [0]PETSC ERROR: Empty row in matrix: row in original
              ordering 2395 in permuted ordering 3600!<br>
            </blockquote>
            <div><br>
            </div>
            <div>1) You misspelled the option:
              -st_pc_factor_shift_amount 1</div>
            <div><br>
            </div>
            <div>2) It was used, and fixed you problem on row 0, now you
              have a problem on row 2395, namely that it is missing. No</div>
            <div>     LU factorization routine will succeed here. You
              must at least put a diagonal 0.0 on all rows.</div>
            <div><br>
            </div>
            <div>3) Same goes for the run below.</div>
            <div><br>
            </div>
            <div>4) A completely empty row means your matrix is indeed
              rank deficient. I have no idea what results are coming out
              of</div>
            <div>    scipy, but they are not using factorization.</div>
            <div><br>
            </div>
            <div>   Matt</div>
            <div> </div>
            <blockquote class="gmail_quote" style="margin:0px 0px 0px
0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">[0]PETSC
              ERROR: ------------------------------------------------------------------------<br>
              [0]PETSC ERROR: Petsc Release Version 3.3.0, Patch 5, Sat
              Dec  1 15:10:41 CST 2012<br>
              [0]PETSC ERROR: See docs/changes/index.html for recent
              updates.<br>
              [0]PETSC ERROR: See docs/faq.html for hints about trouble
              shooting.<br>
              [0]PETSC ERROR: See docs/index.html for manual pages.<br>
              [0]PETSC ERROR: ------------------------------------------------------------------------<br>
              [0]PETSC ERROR: ./ex7 on a arch-darw named tobymac by toby
              Thu Jul 25 12:41:57 2013<br>
              [0]PETSC ERROR: Libraries linked from /opt/local/lib<br>
              [0]PETSC ERROR: Configure run at Tue Jul 23 15:11:27 2013<br>
              [0]PETSC ERROR: Configure options --prefix=/opt/local
              --with-valgrind-dir=/opt/local --with-shared-libraries
              --with-scalar-type=complex --with-clanguage=C++
              --with-superlu-dir=/opt/local --with-blacs-dir=/opt/local
              --with-scalapack-dir=/opt/local
              --with-mumps-dir=/opt/local --with-metis-dir=/opt/local
              --with-parmetis-dir=/opt/local --COPTFLAGS=-O2
              --CXXOPTFLAGS=-O2 --FOPTFLAGS=-O2
              --LDFLAGS=-L/opt/local/lib --CFLAGS="-O2 -mtune=native"
              --CXXFLAGS="-O2 -mtune=native"<br>
              [0]PETSC ERROR: ------------------------------------------------------------------------<br>
              [0]PETSC ERROR: MatLUFactorSymbolic_SeqAIJ() line 334 in
              /opt/local/var/macports/build/_Users_toby_MyPorts_scienceports_math_petsc/petsc/work/petsc-3.3-p5/src/mat/impls/aij/seq/aijfact.c<br>
              [0]PETSC ERROR: MatLUFactorSymbolic() line 2750 in
              /opt/local/var/macports/build/_Users_toby_MyPorts_scienceports_math_petsc/petsc/work/petsc-3.3-p5/src/mat/interface/matrix.c<br>
              [0]PETSC ERROR: PCSetUp_LU() line 135 in
              /opt/local/var/macports/build/_Users_toby_MyPorts_scienceports_math_petsc/petsc/work/petsc-3.3-p5/src/ksp/pc/impls/factor/lu/lu.c<br>
              [0]PETSC ERROR: PCSetUp() line 832 in
              /opt/local/var/macports/build/_Users_toby_MyPorts_scienceports_math_petsc/petsc/work/petsc-3.3-p5/src/ksp/pc/interface/precon.c<br>
              [0]PETSC ERROR: KSPSetUp() line 278 in
              /opt/local/var/macports/build/_Users_toby_MyPorts_scienceports_math_petsc/petsc/work/petsc-3.3-p5/src/ksp/ksp/interface/itfunc.c<br>
              [0]PETSC ERROR: STSetUp_Shift() line 94 in
              src/st/impls/shift/shift.c<br>
              [0]PETSC ERROR: STSetUp() line 280 in
              src/st/interface/stsolve.c<br>
              [0]PETSC ERROR: EPSSetUp() line 204 in
              src/eps/interface/setup.c<br>
              [0]PETSC ERROR: EPSSolve() line 109 in
              src/eps/interface/solve.c<br>
               Number of iterations of the method: 0<br>
               Number of linear iterations of the method: 0<br>
               Number of requested eigenvalues: 1<br>
               Stopping condition: tol=1e-08, maxit=750<br>
               Number of converged approximate eigenpairs: 0<br>
              <br>
              WARNING! There are options you set that were not used!<br>
              WARNING! could be spelling mistake, etc!<br>
              Option left: name:-st_pc_shift_type_amount value: 1<br>
              <br>
              mpiexec ./ex7 -f1 LHS-N7-M40-Re0.0-b0.1-Wi5.0-amp0.02.petsc
               -f2 RHS-N7-M40-Re0.0-b0.1-Wi5.0-amp0.02.petsc -st_pc_type
              lu -st_pc_factor_shift_type POSITIVE_DEFINITE<br>
              <br>
              Generalized eigenproblem stored in file.<br>
              <br>
               Reading COMPLEX matrices from binary files...<br>
              [0]PETSC ERROR: --------------------- Error Message
              ------------------------------------<br>
              [0]PETSC ERROR: Detected zero pivot in LU factorization:<br>
              see <a moz-do-not-send="true"
                href="http://www.mcs.anl.gov/petsc/documentation/faq.html#ZeroPivot"
                target="_blank">http://www.mcs.anl.gov/petsc/documentation/faq.html#ZeroPivot</a>!<br>
              [0]PETSC ERROR: Empty row in matrix: row in original
              ordering 2395 in permuted ordering 3600!<br>
              [0]PETSC ERROR: ------------------------------------------------------------------------<br>
              [0]PETSC ERROR: Petsc Release Version 3.3.0, Patch 5, Sat
              Dec  1 15:10:41 CST 2012<br>
              [0]PETSC ERROR: See docs/changes/index.html for recent
              updates.<br>
              [0]PETSC ERROR: See docs/faq.html for hints about trouble
              shooting.<br>
              [0]PETSC ERROR: See docs/index.html for manual pages.<br>
              [0]PETSC ERROR: ------------------------------------------------------------------------<br>
              [0]PETSC ERROR: ./ex7 on a arch-darw named tobymac by toby
              Thu Jul 25 12:40:20 2013<br>
              [0]PETSC ERROR: Libraries linked from /opt/local/lib<br>
              [0]PETSC ERROR: Configure run at Tue Jul 23 15:11:27 2013<br>
              [0]PETSC ERROR: Configure options --prefix=/opt/local
              --with-valgrind-dir=/opt/local --with-shared-libraries
              --with-scalar-type=complex --with-clanguage=C++
              --with-superlu-dir=/opt/local --with-blacs-dir=/opt/local
              --with-scalapack-dir=/opt/local
              --with-mumps-dir=/opt/local --with-metis-dir=/opt/local
              --with-parmetis-dir=/opt/local --COPTFLAGS=-O2
              --CXXOPTFLAGS=-O2 --FOPTFLAGS=-O2
              --LDFLAGS=-L/opt/local/lib --CFLAGS="-O2 -mtune=native"
              --CXXFLAGS="-O2 -mtune=native"<br>
              [0]PETSC ERROR: ------------------------------------------------------------------------<br>
              [0]PETSC ERROR: MatLUFactorSymbolic_SeqAIJ() line 334 in
              /opt/local/var/macports/build/_Users_toby_MyPorts_scienceports_math_petsc/petsc/work/petsc-3.3-p5/src/mat/impls/aij/seq/aijfact.c<br>
              [0]PETSC ERROR: MatLUFactorSymbolic() line 2750 in
              /opt/local/var/macports/build/_Users_toby_MyPorts_scienceports_math_petsc/petsc/work/petsc-3.3-p5/src/mat/interface/matrix.c<br>
              [0]PETSC ERROR: PCSetUp_LU() line 135 in
              /opt/local/var/macports/build/_Users_toby_MyPorts_scienceports_math_petsc/petsc/work/petsc-3.3-p5/src/ksp/pc/impls/factor/lu/lu.c<br>
              [0]PETSC ERROR: PCSetUp() line 832 in
              /opt/local/var/macports/build/_Users_toby_MyPorts_scienceports_math_petsc/petsc/work/petsc-3.3-p5/src/ksp/pc/interface/precon.c<br>
              [0]PETSC ERROR: KSPSetUp() line 278 in
              /opt/local/var/macports/build/_Users_toby_MyPorts_scienceports_math_petsc/petsc/work/petsc-3.3-p5/src/ksp/ksp/interface/itfunc.c<br>
              [0]PETSC ERROR: STSetUp_Shift() line 94 in
              src/st/impls/shift/shift.c<br>
              [0]PETSC ERROR: STSetUp() line 280 in
              src/st/interface/stsolve.c<br>
              [0]PETSC ERROR: EPSSetUp() line 204 in
              src/eps/interface/setup.c<br>
              [0]PETSC ERROR: EPSSolve() line 109 in
              src/eps/interface/solve.c<br>
               Number of iterations of the method: 0<br>
               Number of linear iterations of the method: 0<br>
               Number of requested eigenvalues: 1<br>
               Stopping condition: tol=1e-08, maxit=750<br>
               Number of converged approximate eigenpairs: 0<br>
              <br>
              <br>
              <br>
              <br>
              On 25/07/2013 12:34, Jose E. Roman wrote:<br>
              <blockquote class="gmail_quote" style="margin:0px 0px 0px
0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">
                El 25/07/2013, a las 13:30, Toby escribió:<br>
                <br>
                <blockquote class="gmail_quote" style="margin:0px 0px
                  0px
0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">Hi
                  Matt,<br>
                  <br>
                  Thanks for the speedy reply. I tried using the options
                  you suggested:<br>
                  <br>
                  mpiexec ./ex7 -f1 LHS-N7-M40-Re0.0-b0.1-Wi5.0-amp0.02.petsc
                   -f2 RHS-N7-M40-Re0.0-b0.1-Wi5.0-amp0.02.petsc
                  -st_pc_factor_shift_type NONZERO
                  -st_pc_shift_type_amount 1<br>
                  <br>
                  But I still get the warnings:<br>
                  Option left: name:-st_pc_factor_shift_type value:
                  NONZERO<br>
                  Option left: name:-st_pc_shift_type_amount value: 1<br>
                </blockquote>
                By default, ST is using PCREDUNDANT. In order to use the
                above options you must change it to e.g. PCLU.<br>
                That is, add -st_pc_type lu<br>
                <br>
                Jose<br>
                <br>
                <br>
                <blockquote class="gmail_quote" style="margin:0px 0px
                  0px
0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">I
                  have tried it with just the -st_pc_type jacobi and
                  -st_ksp_view. This gives me some eigenvalues, but I
                  don't believe them (they do not appear on the spectrum
                  which I solve using LAPACK where all eigenvalues have
                  a real part less than 0). The output was very large,
                  but consists of repetitions of this:<br>
                  <br>
                  mpiexec ./ex7 -f1 LHS-N7-M40-Re0.0-b0.1-Wi5.0-amp0.02.petsc
                   -f2 RHS-N7-M40-Re0.0-b0.1-Wi5.0-amp0.02.petsc
                  -st_pc_type jacobi -st_ksp_view<br>
                  <br>
                  Generalized eigenproblem stored in file.<br>
                  <br>
                    Reading COMPLEX matrices from binary files...<br>
                  KSP Object:(st_) 1 MPI processes<br>
                     type: preonly<br>
                     maximum iterations=10000, initial guess is zero<br>
                     tolerances:  relative=1e-08, absolute=1e-50,
                  divergence=10000<br>
                     left preconditioning<br>
                     using NONE norm type for convergence test<br>
                  PC Object:(st_) 1 MPI processes<br>
                     type: jacobi<br>
                     linear system matrix = precond matrix:<br>
                     Matrix Object:   1 MPI processes<br>
                       type: seqaij<br>
                       rows=6000, cols=6000<br>
                       total: nonzeros=3600, allocated nonzeros=3600<br>
                       total number of mallocs used during MatSetValues
                  calls =0<br>
                         using I-node routines: found 4080 nodes, limit
                  used is 5<br>
                  <br>
                  ...<br>
                  <br>
                    Number of iterations of the method: 189<br>
                    Number of linear iterations of the method: 1520<br>
                    Number of requested eigenvalues: 1<br>
                    Stopping condition: tol=1e-08, maxit=750<br>
                    Number of converged approximate eigenpairs: 2<br>
                  <br>
                              k          ||Ax-kBx||/||kx||<br>
                      ----------------- ------------------<br>
                     1388.774454+0.001823 i     0.751726<br>
                     1388.774441+0.001820 i     0.912665<br>
                  <br>
                  <br>
                  Thanks again,<br>
                  Toby<br>
                </blockquote>
                <br>
                <span class=""><font color="#888888">
                  </font></span></blockquote>
              <span class=""><font color="#888888">
                  <br>
                  <br>
                  -- <br>
                  The University of Edinburgh is a charitable body,
                  registered in<br>
                  Scotland, with registration number SC005336.<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
        </div>
      </div>
    </blockquote>
    <br>
  </body>
</html>