<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">Danyang :</div><div class="gmail_quote">Your matrices are ill-conditioned, numerically singular with </div><div class="gmail_quote"><span style="font-size:12.8px">Recip. condition number = 6.000846e-16</span></div><div class="gmail_quote"><span style="font-size:12.8px">Recip. condition number = 2.256434e-27</span><span style="font-size:12.8px"><br></span></div><div class="gmail_quote"><span style="font-size:12.8px">Recip. condition number = 1.256452e-18</span><span style="font-size:12.8px"><br></span></div><div class="gmail_quote"><span style="font-size:12.8px">i.e., condition numbers = O(1.e16 - 1.e27), there is no accuracy in computed solution.</span></div><div class="gmail_quote"><span style="font-size:12.8px"><br></span></div><div class="gmail_quote"><span style="font-size:12.8px">I checked your </span><span style="font-size:12.8px">matrix  168 - 172, got </span><span style="font-size:12.8px">Recip. condition number = 1.548816e-12.</span></div><div class="gmail_quote"><br></div><div class="gmail_quote"><span style="font-size:12.8px">You need check your model to understand why the matrices are so ill-conditioned.</span></div><div class="gmail_quote"><br></div><div class="gmail_quote"><span style="font-size:12.8px">Hong</span></div><div class="gmail_quote"><span style="font-size:12.8px"><br></span><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">
  
    
  
  <div text="#000000" bgcolor="#FFFFFF">
    Hi Hong,<br>
    <br>
    Sorry to bother you again. The modified code works much better than
    before using both superlu or mumps. However, it still encounters
    failure. The case is similar with the previous one, ill-conditioned
    matrices.<br>
    <br>
    The code crashed after a long time simulation if I use superlu_dist,
    but will not fail if use superlu. I restart the simulation before
    the time it crashes and can reproduce the following error<br>
    <br>
     timestep:      22 time: 1.750E+04 years   delt: 2.500E+00 years  
    iter:  1 max.sia: 5.053E-03 tol.sia: 1.000E-01<br>
     Newton Iteration Convergence Summary:<br>
     Newton       maximum      maximum     solver<br>
     iteration    updatePa  updateTemp    residual    iterations 
    maxvolpa   maxvoltemp   nexvolpa   nexvoltemp<br>
         1       0.1531E+08 0.1755E+04   0.6920E-05        1    
    5585          4402        5814        5814<br>
    <br>
    *** Error in `../program_test': malloc(): memory corruption:
    0x0000000003a70d50 ***<br>
    Program received signal SIGABRT: Process abort signal.<br>
    Backtrace for this error:<br>
    <br>
    The solver failed at timestep 22, Newton iteration 2. I exported the
    matrices at timestep 1 (matrix 1) and timestep 22 (matrix 140 and
    141). Matrix 141 is where it failed.  The three matrices here are
    not ill-conditioned form the estimated value.<br>
    <br>
    I did the same using the new modified ex52f code and found pretty
    different results for matrix 141. The norm by superlu is much
    acceptable than superlu_dist. In this test, memory corruption was
    not detected. The codes and example data can be download from the
    link below.<br>
    <br>
<a href="https://www.dropbox.com/s/i1ls0bg0vt7gu0v/petsc-superlu-test2.tar.gz?dl=0" target="_blank">https://www.dropbox.com/s/i1ls0bg0vt7gu0v/petsc-superlu-test2.tar.gz?dl=0</a><br>
    <br>
    <br>
    ****************More test on matrix_and_rhs_bin2*******************<br>
    mpiexec.hydra -n 1 ./ex52f -f0
    ./matrix_and_rhs_bin2/a_flow_check_1.bin -rhs
    ./matrix_and_rhs_bin2/b_flow_check_1.bin -loop_matrices flow_check
    -loop_folder ./matrix_and_rhs_bin2 -matrix_index_start 140
    -matrix_index_end 141  -pc_type lu -pc_factor_mat_solver_package
    superlu -ksp_monitor_true_residual -mat_superlu_conditionnumber<span><br>
     -->loac matrix a<br>
     -->load rhs b<br>
     size l,m,n,mm       90000       90000       90000       90000<br></span>
      Recip. condition number = 6.000846e-16<br>
      0 KSP preconditioned resid norm 1.146871454377e+08 true resid norm
    4.711091037809e+03 ||r(i)||/||b|| 1.000000000000e+00<br>
      1 KSP preconditioned resid norm 2.071118508260e-06 true resid norm
    3.363767171515e-08 ||r(i)||/||b|| 7.140102249181e-12<br>
    Norm of error  3.3638E-08 iterations     1<br>
     -->Test for matrix          140<br>
      Recip. condition number = 2.256434e-27<br>
      0 KSP preconditioned resid norm 2.084372893355e+14 true resid norm
    4.711091037809e+03 ||r(i)||/||b|| 1.000000000000e+00<br>
      1 KSP preconditioned resid norm 4.689629276419e+00 true resid norm
    1.037236635337e-01 ||r(i)||/||b|| 2.201690918330e-05<br>
    Norm of error  1.0372E-01 iterations     1<br>
     -->Test for matrix          141<br>
      Recip. condition number = 1.256452e-18<br>
      0 KSP preconditioned resid norm 1.055488964519e+08 true resid norm
    4.711091037809e+03 ||r(i)||/||b|| 1.000000000000e+00<br>
      1 KSP preconditioned resid norm 2.998827511681e-04 true resid norm
    4.805214542776e-04 ||r(i)||/||b|| 1.019979130994e-07<br>
    Norm of error  4.8052E-04 iterations     1<span><br>
     --> End of test, bye<br>
    <br>
    <br></span>
    mpiexec.hydra -n 1 ./ex52f -f0
    ./matrix_and_rhs_bin2/a_flow_check_1.bin -rhs
    ./matrix_and_rhs_bin2/b_flow_check_1.bin -loop_matrices flow_check
    -loop_folder ./matrix_and_rhs_bin2 -matrix_index_start 140
    -matrix_index_end 141  -pc_type lu -pc_factor_mat_solver_package
    superlu_dist<span><br>
     -->loac matrix a<br>
     -->load rhs b<br>
     size l,m,n,mm       90000       90000       90000       90000<br></span>
    Norm of error  3.6752E-08 iterations     1<br>
     -->Test for matrix          140<br>
    Norm of error  1.6335E-01 iterations     1<br>
     -->Test for matrix          141<br>
    Norm of error  3.4345E+01 iterations     1<span><br>
     --> End of test, bye<br>
    <br></span>
    Thanks,<span><br>
    <br>
    Danyang<br>
    <br>
    <div>On 15-12-07 12:01 PM, Hong wrote:<br>
    </div>
    </span><div><div><blockquote type="cite">
      <div dir="ltr">
        <div class="gmail_extra">
          <div class="gmail_quote">Danyang:</div>
          <div class="gmail_quote">Add 'call MatSetFromOptions(A,ierr)'
            to your code.</div>
          <div class="gmail_quote">Attached below is ex52f.F modified
            from your ex52f.F to be compatible with petsc-dev.</div>
          <div class="gmail_quote"><br>
          </div>
          <div class="gmail_quote">Hong</div>
          <div class="gmail_quote"><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">
              <div text="#000000" bgcolor="#FFFFFF"> Hello Hong,<br>
                <br>
                Thanks for the quick reply and the option
                "-mat_superlu_dist_fact SamePattern" works like a charm,
                if I use this option from the command line. <br>
                <br>
                How can I add this option as the default. I tried using
                PetscOptionsInsertString("-mat_superlu_dist_fact
                SamePattern",ierr) in my code but this does not work. <br>
                <br>
                Thanks,<br>
                <br>
                Danyang
                <div>
                  <div><br>
                    <br>
                    <div>On 15-12-07 10:42 AM, Hong wrote:<br>
                    </div>
                    <blockquote type="cite">
                      <div dir="ltr">
                        <div class="gmail_extra">
                          <div class="gmail_quote">Danyang :</div>
                          <div class="gmail_quote"><br>
                          </div>
                          <div class="gmail_quote">Adding
                            '-mat_superlu_dist_fact SamePattern' fixed
                            the problem. Below is how I figured it out.</div>
                          <div class="gmail_quote"><br>
                          </div>
                          <div class="gmail_quote">
                            <div class="gmail_quote">1. Reading ex52f.F,
                              I see '-superlu_default' =
                              '-pc_factor_mat_solver_package
                              superlu_dist', the later enables runtime
                              options for other packages. I use
                              superlu_dist-4.2 and superlu-4.1 for the
                              tests below.</div>
                            <div class="gmail_quote"><br>
                            </div>
                            <div class="gmail_quote">2. Use the Matrix
                              168 to setup KSP solver and factorization,
                              all packages, petsc, superlu_dist and
                              mumps give same correct results:</div>
                            <div class="gmail_quote"><br>
                            </div>
                            <div class="gmail_quote">./ex52f -f0
                              matrix_and_rhs_bin/a_flow_check_168.bin
                              -rhs
                              matrix_and_rhs_bin/b_flow_check_168.bin
                              -loop_matrices flow_check -loop_folder
                              matrix_and_rhs_bin -pc_type lu
                              -pc_factor_mat_solver_package petsc</div>
                            <div class="gmail_quote"> -->loac matrix
                              a</div>
                            <div class="gmail_quote"> -->load rhs b</div>
                            <div class="gmail_quote"> size l,m,n,mm    
                                90000       90000       90000      
                              90000</div>
                            <div class="gmail_quote">Norm of error
                               7.7308E-11 iterations     1</div>
                            <div class="gmail_quote"> -->Test for
                              matrix          168</div>
                            <div class="gmail_quote">..</div>
                            <div class="gmail_quote"> -->Test for
                              matrix          172</div>
                            <div class="gmail_quote">Norm of error
                               3.8461E-11 iterations     1</div>
                            <div class="gmail_quote"><br>
                            </div>
                            <div class="gmail_quote">./ex52f -f0
                              matrix_and_rhs_bin/a_flow_check_168.bin
                              -rhs
                              matrix_and_rhs_bin/b_flow_check_168.bin
                              -loop_matrices flow_check -loop_folder
                              matrix_and_rhs_bin -pc_type lu
                              -pc_factor_mat_solver_package superlu_dist</div>
                            <div class="gmail_quote">Norm of error
                               9.4073E-11 iterations     1</div>
                            <div class="gmail_quote"> -->Test for
                              matrix          168</div>
                            <div class="gmail_quote">...</div>
                            <div class="gmail_quote"> -->Test for
                              matrix          172</div>
                            <div class="gmail_quote">Norm of error
                               3.8187E-11 iterations     1</div>
                            <div class="gmail_quote"><br>
                            </div>
                            <div class="gmail_quote">
                              <div class="gmail_quote">3. Use superlu, I
                                get</div>
                              <div class="gmail_quote">./ex52f -f0
                                matrix_and_rhs_bin/a_flow_check_168.bin
                                -rhs
                                matrix_and_rhs_bin/b_flow_check_168.bin
                                -loop_matrices flow_check -loop_folder
                                matrix_and_rhs_bin -pc_type lu
                                -pc_factor_mat_solver_package superlu</div>
                              <div class="gmail_quote">Norm of error
                                 1.0191E-06 iterations     1</div>
                              <div class="gmail_quote"> -->Test for
                                matrix          168</div>
                              <div class="gmail_quote">...</div>
                              <div class="gmail_quote"> -->Test for
                                matrix          172</div>
                              <div class="gmail_quote">Norm of error
                                 9.7858E-07 iterations     1</div>
                              <div class="gmail_quote"><br>
                              </div>
                              <div class="gmail_quote">Replacing default
                                DiagPivotThresh: 1. to 0.0, I get same
                                solutions as other packages:</div>
                              <div class="gmail_quote"><br>
                              </div>
                              <div class="gmail_quote">./ex52f -f0
                                matrix_and_rhs_bin/a_flow_check_168.bin
                                -rhs
                                matrix_and_rhs_bin/b_flow_check_168.bin
                                -loop_matrices flow_check -loop_folder
                                matrix_and_rhs_bin -pc_type lu
                                -pc_factor_mat_solver_package superlu
                                -mat_superlu_diagpivotthresh 0.0</div>
                              <div class="gmail_quote"><br>
                              </div>
                              <div class="gmail_quote">Norm of error
                                 8.3614E-11 iterations     1</div>
                              <div class="gmail_quote"> -->Test for
                                matrix          168</div>
                              <div class="gmail_quote">...</div>
                              <div class="gmail_quote"> -->Test for
                                matrix          172</div>
                              <div class="gmail_quote">Norm of error
                                 3.7098E-11 iterations     1</div>
                              <div class="gmail_quote"><br>
                              </div>
                              <div class="gmail_quote">4.</div>
                              <div class="gmail_quote">using '-mat_view
                                ascii::ascii_info', I found that
                                a_flow_check_1.bin and
                                a_flow_check_168.bin seem have same
                                structure:</div>
                              <div class="gmail_quote"><br>
                              </div>
                              <div class="gmail_quote"> -->loac
                                matrix a</div>
                              <div class="gmail_quote">Mat Object: 1 MPI
                                processes</div>
                              <div class="gmail_quote">  type: seqaij</div>
                              <div class="gmail_quote">  rows=90000,
                                cols=90000</div>
                              <div class="gmail_quote">  total:
                                nonzeros=895600, allocated
                                nonzeros=895600</div>
                              <div class="gmail_quote">  total number of
                                mallocs used during MatSetValues calls
                                =0</div>
                              <div class="gmail_quote">    using I-node
                                routines: found 45000 nodes, limit used
                                is 5</div>
                              <div class="gmail_quote"><br>
                              </div>
                              <div class="gmail_quote">
                                <div class="gmail_quote">5.</div>
                                <div class="gmail_quote">Using
                                  a_flow_check_1.bin, I am able to
                                  reproduce the error you reported: all
                                  packages give correct results except
                                  superlu_dist:</div>
                                <div class="gmail_quote">./ex52f -f0
                                  matrix_and_rhs_bin/a_flow_check_1.bin
                                  -rhs
                                  matrix_and_rhs_bin/b_flow_check_168.bin
                                  -loop_matrices flow_check -loop_folder
                                  matrix_and_rhs_bin -pc_type lu
                                  -pc_factor_mat_solver_package
                                  superlu_dist</div>
                                <div class="gmail_quote">Norm of error
                                   2.5970E-12 iterations     1</div>
                                <div class="gmail_quote"> -->Test for
                                  matrix          168</div>
                                <div class="gmail_quote">Norm of error
                                   1.3936E-01 iterations    34</div>
                                <div class="gmail_quote"> -->Test for
                                  matrix          169</div>
                                <div class="gmail_quote"><br>
                                </div>
                                <div class="gmail_quote">I guess the
                                  error might come from reuse of matrix
                                  factor. Replacing default</div>
                                <div class="gmail_quote">-mat_superlu_dist_fact

                                  <SamePattern_SameRowPerm> with</div>
                                <div class="gmail_quote">-mat_superlu_dist_fact

                                  SamePattern, I get</div>
                                <div class="gmail_quote"><br>
                                </div>
                                <div class="gmail_quote">./ex52f -f0
                                  matrix_and_rhs_bin/a_flow_check_1.bin
                                  -rhs
                                  matrix_and_rhs_bin/b_flow_check_168.bin
                                  -loop_matrices flow_check -loop_folder
                                  matrix_and_rhs_bin -pc_type lu
                                  -pc_factor_mat_solver_package
                                  superlu_dist -mat_superlu_dist_fact
                                  SamePattern</div>
                                <div class="gmail_quote"><br>
                                </div>
                                <div class="gmail_quote">Norm of error
                                   2.5970E-12 iterations     1</div>
                                <div class="gmail_quote"> -->Test for
                                  matrix          168</div>
                                <div class="gmail_quote">Norm of error
                                   9.4073E-11 iterations     1</div>
                                <div class="gmail_quote"> -->Test for
                                  matrix          169</div>
                                <div class="gmail_quote">Norm of error
                                   6.4303E-11 iterations     1</div>
                                <div class="gmail_quote"> -->Test for
                                  matrix          170</div>
                                <div class="gmail_quote">Norm of error
                                   7.4327E-11 iterations     1</div>
                                <div class="gmail_quote"> -->Test for
                                  matrix          171</div>
                                <div class="gmail_quote">Norm of error
                                   5.4162E-11 iterations     1</div>
                                <div class="gmail_quote"> -->Test for
                                  matrix          172</div>
                                <div class="gmail_quote">Norm of error
                                   3.4440E-11 iterations     1</div>
                                <div class="gmail_quote"> --> End of
                                  test, bye</div>
                                <div class="gmail_quote"><br>
                                </div>
                                <div class="gmail_quote">Sherry may tell
                                  you why SamePattern_SameRowPerm cause
                                  the difference here.</div>
                                <div class="gmail_quote">Best on the
                                  above experiments, I would set
                                  following as default</div>
                                <div class="gmail_quote">'-mat_superlu_diagpivotthresh

                                  0.0' in petsc/superlu interface.</div>
                                <div class="gmail_quote">'-mat_superlu_dist_fact

                                  SamePattern' in petsc/superlu_dist
                                  interface.</div>
                                <div class="gmail_quote"><br>
                                </div>
                                <div class="gmail_quote">Hong</div>
                              </div>
                            </div>
                          </div>
                          <div class="gmail_quote"><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">
                              <div bgcolor="#FFFFFF" text="#000000"> Hi
                                Hong,<br>
                                <br>
                                I did more test today and finally found
                                that the solution accuracy depends on
                                the initial (first) matrix quality. I
                                modified the ex52f.F to do the test.
                                There are 6 matrices and right-hand-side
                                vectors. All these matrices and rhs are
                                from my reactive transport simulation.
                                Results will be quite different
                                depending on which one you use to do
                                factorization. Results will also be
                                different if you run with different
                                options. My code is similar to the First
                                or the Second test below. When the
                                matrix is well conditioned, it works
                                fine. But if the initial matrix is well
                                conditioned, it likely to crash when the
                                matrix become ill-conditioned. Since
                                most of my case are well conditioned so
                                I didn't detect the problem before. This
                                case is a special one. <br>
                                <br>
                                <br>
                                How can I avoid this problem? Shall I
                                redo factorization? Can PETSc
                                automatically detect this prolbem or is
                                there any option available to do this?<br>
                                <br>
                                All the data and test code (modified
                                ex52f) can be found via the dropbox link
                                below. <br>
                                <u><br>
                                </u><u><a href="https://www.dropbox.com/s/4al1a60creogd8m/petsc-superlu-test.tar.gz?dl=0" target="_blank">https://www.dropbox.com/s/4al1a60creogd8m/petsc-superlu-test.tar.gz?dl=0</a></u><br>
                                <br>
                                <br>
                                Summary of my test is shown below. <br>
                                <br>
                                First, use the Matrix 1 to setup KSP
                                solver and factorization, then solve 168
                                to 172<br>
                                <br>
                                mpiexec.hydra -n 1 ./ex52f -f0
                                /home/dsu/work/petsc-superlu-test/matrix_and_rhs_bin/a_flow_check_1.bin
                                -rhs
                                /home/dsu/work/petsc-superlu-test/matrix_and_rhs_bin/b_flow_check_1.bin
                                -loop_matrices flow_check -loop_folder
                                /home/dsu/work/petsc-superlu-test/matrix_and_rhs_bin
                                -pc_type lu
                                -pc_factor_mat_solver_package
                                superlu_dist<br>
                                <br>
                                Norm of error  3.8815E-11 iterations    
                                1<br>
                                 -->Test for matrix          168<br>
                                Norm of error  4.2307E-01 iterations   
                                32<br>
                                 -->Test for matrix          169<br>
                                Norm of error  3.0528E-01 iterations   
                                32<br>
                                 -->Test for matrix          170<br>
                                Norm of error  3.1177E-01 iterations   
                                32<br>
                                 -->Test for matrix          171<br>
                                Norm of error  3.2793E-01 iterations   
                                32<br>
                                 -->Test for matrix          172<br>
                                Norm of error  3.1251E-01 iterations   
                                31<br>
                                <br>
                                Second, use the Matrix 1 to setup KSP
                                solver and factorization using the
                                implemented SuperLU relative codes. I
                                thought this will generate the same
                                results as the First test, but it
                                actually not.<br>
                                <br>
                                mpiexec.hydra -n 1 ./ex52f -f0
                                /home/dsu/work/petsc-superlu-test/matrix_and_rhs_bin/a_flow_check_1.bin
                                -rhs
                                /home/dsu/work/petsc-superlu-test/matrix_and_rhs_bin/b_flow_check_1.bin
                                -loop_matrices flow_check -loop_folder
                                /home/dsu/work/petsc-superlu-test/matrix_and_rhs_bin
                                -superlu_default<br>
                                <br>
                                Norm of error  2.2632E-12 iterations    
                                1<br>
                                 -->Test for matrix          168<br>
                                Norm of error  1.0817E+04 iterations    
                                1<br>
                                 -->Test for matrix          169<br>
                                Norm of error  1.0786E+04 iterations    
                                1<br>
                                 -->Test for matrix          170<br>
                                Norm of error  1.0792E+04 iterations    
                                1<br>
                                 -->Test for matrix          171<br>
                                Norm of error  1.0792E+04 iterations    
                                1<br>
                                 -->Test for matrix          172<br>
                                Norm of error  1.0792E+04 iterations    
                                1<br>
                                <br>
                                <br>
                                Third, use the Matrix 168 to setup KSP
                                solver and factorization, then solve 168
                                to 172<br>
                                <br>
                                mpiexec.hydra -n 1 ./ex52f -f0
                                /home/dsu/work/petsc-superlu-test/matrix_and_rhs_bin/a_flow_check_168.bin


                                -rhs
                                /home/dsu/work/petsc-superlu-test/matrix_and_rhs_bin/b_flow_check_168.bin


                                -loop_matrices flow_check -loop_folder
                                /home/dsu/work/petsc-superlu-test/matrix_and_rhs_bin
                                -pc_type lu
                                -pc_factor_mat_solver_package
                                superlu_dist<br>
                                <br>
                                Norm of error  9.5528E-10 iterations    
                                1<br>
                                 -->Test for matrix          168<br>
                                Norm of error  9.4945E-10 iterations    
                                1<br>
                                 -->Test for matrix          169<br>
                                Norm of error  6.4279E-10 iterations    
                                1<br>
                                 -->Test for matrix          170<br>
                                Norm of error  7.4633E-10 iterations    
                                1<br>
                                 -->Test for matrix          171<br>
                                Norm of error  7.4863E-10 iterations    
                                1<br>
                                 -->Test for matrix          172<br>
                                Norm of error  8.9701E-10 iterations    
                                1<br>
                                <br>
                                Fourth, use the Matrix 168 to setup KSP
                                solver and factorization using the
                                implemented SuperLU relative codes. I
                                thought this will generate the same
                                results as the Third test, but it
                                actually not.<br>
                                <br>
                                mpiexec.hydra -n 1 ./ex52f -f0
                                /home/dsu/work/petsc-superlu-test/matrix_and_rhs_bin/a_flow_check_168.bin


                                -rhs
                                /home/dsu/work/petsc-superlu-test/matrix_and_rhs_bin/b_flow_check_168.bin


                                -loop_matrices flow_check -loop_folder
                                /home/dsu/work/petsc-superlu-test/matrix_and_rhs_bin
                                -superlu_default<br>
                                <br>
                                Norm of error  3.7017E-11 iterations    
                                1<br>
                                 -->Test for matrix          168<br>
                                Norm of error  3.6420E-11 iterations    
                                1<br>
                                 -->Test for matrix          169<br>
                                Norm of error  3.7184E-11 iterations    
                                1<br>
                                 -->Test for matrix          170<br>
                                Norm of error  3.6847E-11 iterations    
                                1<br>
                                 -->Test for matrix          171<br>
                                Norm of error  3.7883E-11 iterations    
                                1<br>
                                 -->Test for matrix          172<br>
                                Norm of error  3.8805E-11 iterations    
                                1<br>
                                <br>
                                Thanks very much,<br>
                                <br>
                                Danyang<br>
                                <br>
                                <div>On 15-12-03 01:59 PM, Hong wrote:<br>
                                </div>
                                <blockquote type="cite">
                                  <div dir="ltr">
                                    <div class="gmail_extra">
                                      <div class="gmail_quote">Danyang :</div>
                                      <div class="gmail_quote">Further
                                        testing a_flow_check_168.bin,</div>
                                      <div class="gmail_quote">
                                        <div class="gmail_quote">./ex10
                                          -f0
                                          /Users/Hong/Downloads/matrix_and_rhs_bin/a_flow_check_168.bin
                                          -rhs
                                          /Users/Hong/Downloads/matrix_and_rhs_bin/x_flow_check_168.bin
                                          -pc_type lu
                                          -pc_factor_mat_solver_package
                                          superlu
                                          -ksp_monitor_true_residual
                                          -mat_superlu_conditionnumber</div>
                                        <div class="gmail_quote"> 
                                          Recip. condition number =
                                          1.610480e-12</div>
                                        <div class="gmail_quote">  0 KSP
                                          preconditioned resid norm
                                          6.873340313547e+09 true resid
                                          norm 7.295020990196e+03
                                          ||r(i)||/||b||
                                          1.000000000000e+00</div>
                                        <div class="gmail_quote">  1 KSP
                                          preconditioned resid norm
                                          2.051833296449e-02 true resid
                                          norm 2.976859070118e-02
                                          ||r(i)||/||b||
                                          4.080672384793e-06</div>
                                        <div class="gmail_quote">Number
                                          of iterations =   1</div>
                                        <div class="gmail_quote">Residual
                                          norm 0.0297686</div>
                                        <div class="gmail_quote"><br>
                                        </div>
                                        <div class="gmail_quote">condition
                                          number of this matrix =
                                          1/1.610480e-12 = 1.e+12,</div>
                                        <div class="gmail_quote">i.e.,
                                          this matrix is
                                          ill-conditioned. </div>
                                        <div class="gmail_quote"><br>
                                        </div>
                                        <div class="gmail_quote">Hong</div>
                                        <div class="gmail_quote"><br>
                                        </div>
                                        <div class="gmail_quote"><br>
                                        </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">
                                          <div text="#000000" bgcolor="#FFFFFF"> Hi Hong,<br>
                                            <br>
                                            The binary format of matrix,
                                            rhs and solution can be
                                            downloaded via the link
                                            below.<br>
                                            <br>
                                            <a href="https://www.dropbox.com/s/cl3gfi0s0kjlktf/matrix_and_rhs_bin.tar.gz?dl=0" target="_blank">https://www.dropbox.com/s/cl3gfi0s0kjlktf/matrix_and_rhs_bin.tar.gz?dl=0</a><br>
                                            <br>
                                            Thanks,<br>
                                            <br>
                                            Danyang
                                            <div>
                                              <div><br>
                                                <br>
                                                On 15-12-03 10:50 AM,
                                                Hong wrote:<br>
                                                <blockquote type="cite">
                                                  <div dir="ltr">
                                                    <div class="gmail_extra">
                                                      <div class="gmail_quote">Danyang:<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">
                                                          <div text="#000000" bgcolor="#FFFFFF"><br>
                                                          <br>
                                                          To my
                                                          surprising,
                                                          solutions from
                                                          SuperLU at
                                                          timestep 29
                                                          seems not
                                                          correct for
                                                          the first 4
                                                          Newton
                                                          iterations,
                                                          but the
                                                          solutions from
                                                          iteration
                                                          solver and
                                                          MUMPS are
                                                          correct. <br>
                                                          <br>
                                                          Please find
                                                          all the
                                                          matrices, rhs
                                                          and solutions
                                                          at timestep 29
                                                          via the link
                                                          below. The
                                                          data is a bit
                                                          large so that
                                                          I just share
                                                          it through
                                                          Dropbox. A
                                                          piece of
                                                          matlab code to
                                                          read these
                                                          data and then
                                                          computer the
                                                          norm has also
                                                          been attached.
                                                          <br>
                                                          <u><a href="https://www.dropbox.com/s/rr8ueysgflmxs7h/results-check.tar.gz?dl=0" target="_blank"></a><a href="https://www.dropbox.com/s/rr8ueysgflmxs7h/results-check.tar.gz?dl=0" target="_blank">https://www.dropbox.com/s/rr8ueysgflmxs7h/results-check.tar.gz?dl=0</a></u></div>
                                                        </blockquote>
                                                        <div><br>
                                                        </div>
                                                        <div>Can you
                                                          send us matrix
                                                          in petsc
                                                          binary format?</div>
                                                        <div><br>
                                                        </div>
                                                        <div>e.g., call
                                                          MatView(M,
                                                          PETSC_VIEWER_BINARY_(PETSC_COMM_WORLD))</div>
                                                        <div>or
                                                          '-ksp_view_mat
                                                          binary'</div>
                                                        <div><br>
                                                        </div>
                                                        <div>Hong</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">
                                                          <div text="#000000" bgcolor="#FFFFFF"><br>
                                                          <br>
                                                          Below is a
                                                          summary of the
                                                          norm from the
                                                          three solvers
                                                          at timestep
                                                          29, newton
                                                          iteration 1 to
                                                          5.<br>
                                                          <br>
                                                          Timestep 29<br>
                                                          Norm of
                                                          residual seq
                                                          1.661321e-09,
                                                          superlu
                                                          1.657103e+04,
                                                          mumps
                                                          3.731225e-11 <br>
                                                          Norm of
                                                          residual seq
                                                          1.753079e-09,
                                                          superlu
                                                          6.675467e+02,
                                                          mumps
                                                          1.509919e-13 <br>
                                                          Norm of
                                                          residual seq
                                                          4.914971e-10,
                                                          superlu
                                                          1.236362e-01,
                                                          mumps
                                                          2.139303e-17 <br>
                                                          Norm of
                                                          residual seq
                                                          3.532769e-10,
                                                          superlu
                                                          1.304670e-04,
                                                          mumps
                                                          5.387000e-20 <br>
                                                          Norm of
                                                          residual seq
                                                          3.885629e-10,
                                                          superlu
                                                          2.754876e-07,
                                                          mumps
                                                          4.108675e-21 <br>
                                                          <br>
                                                          Would anybody
                                                          please check
                                                          if SuperLU can
                                                          solve these
                                                          matrices?
                                                          Another
                                                          possibility is
                                                          that something
                                                          is wrong in my
                                                          own code. But
                                                          so far, I
                                                          cannot find
                                                          any problem in
                                                          my code since
                                                          the same code
                                                          works fine if
                                                          I using
                                                          iterative
                                                          solver or
                                                          direct solver
                                                          MUMPS. But for
                                                          other cases I
                                                          have tested, 
                                                          all these
                                                          solvers work
                                                          fine.<br>
                                                          <br>
                                                          Please let me
                                                          know if I did
                                                          not write down
                                                          the problem
                                                          clearly.<br>
                                                          <br>
                                                          Thanks,<br>
                                                          <br>
                                                          Danyang<br>
                                                          <br>
                                                          <br>
                                                          <br>
                                                          </div>
                                                        </blockquote>
                                                      </div>
                                                      <br>
                                                    </div>
                                                  </div>
                                                </blockquote>
                                                <br>
                                              </div>
                                            </div>
                                          </div>
                                        </blockquote>
                                      </div>
                                      <br>
                                    </div>
                                  </div>
                                </blockquote>
                                <br>
                              </div>
                            </blockquote>
                          </div>
                          <br>
                        </div>
                      </div>
                    </blockquote>
                    <br>
                  </div>
                </div>
              </div>
            </blockquote>
          </div>
          <br>
        </div>
      </div>
    </blockquote>
    <br>
  </div></div></div>

</blockquote></div><br></div></div>