<div dir="ltr"><div dir="ltr">On Wed, Jan 6, 2021 at 11:05 AM Roland Richter <<a href="mailto:roland.richter@ntnu.no">roland.richter@ntnu.no</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>Hei,</p>
    <p>I ran the program in both versions using "valgrind
      --tool=memcheck --leak-check=full --show-leak-kinds=all
      <binary> -malloc_debug". I got</p>
    <p>==3059== LEAK SUMMARY:<br>
      ==3059==    definitely lost: 12,916 bytes in 32 blocks<br>
      ==3059==    indirectly lost: 2,415 bytes in 2 blocks<br>
      ==3059==      possibly lost: 0 bytes in 0 blocks<br>
      ==3059==    still reachable: 103,511 bytes in 123 blocks<br>
      ==3059==         suppressed: 0 bytes in 0 blocks<br>
    </p>
    <p>but none of the leaks is related to the scaling-function itself.</p>
    <p>Did I miss something here?</p>
    <p></p></div></blockquote><div>Here is my analysis. It is certainly the case that MatScale() does not mysteriously scale by other numbers.</div><div>It is used all over the place in tests, and in the code. Your test requires another package. Thus, it seems</div><div>reasonable to guess that a bad interaction with that package (memory overwrite, conflicting layout or format, etc.)</div><div>is responsible for the behavior you see.</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!<br>
    </p>
    <div>Am 06.01.21 um 15:26 schrieb Matthew
      Knepley:<br>
    </div>
    <blockquote type="cite">
      
      <div dir="ltr">
        <div dir="ltr">On Wed, Jan 6, 2021 at 2:41 AM Roland Richter
          <<a href="mailto:roland.richter@ntnu.no" target="_blank">roland.richter@ntnu.no</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>Hei,</p>
              <p>I added one additional function to the code:</p>
              <p><i>void test_scaling_petsc_pointer(const Mat
                  &in_mat,</i><i><br>
                </i><i>                                Mat &out_mat,</i><i><br>
                </i><i>                                const PetscScalar
                  &scaling_factor) {</i><i><br>
                </i><i>    MatCopy (in_mat, out_mat,
                  SAME_NONZERO_PATTERN);</i><i><br>
                </i><i>    PetscScalar *mat_ptr;</i><i><br>
                </i><i>    MatDenseGetArray (out_mat, &mat_ptr);</i><i><br>
                </i><i>    PetscInt r_0, r_1;</i><i><br>
                </i><i>    MatGetLocalSize (out_mat, &r_0,
                  &r_1);</i><i><br>
                </i><i>    for(int i = 0; i < r_0 * r_1; ++i)</i><i><br>
                </i><i>        *(mat_ptr + i) = (*(mat_ptr + i) *
                  scaling_factor);</i><i><br>
                </i><i><br>
                </i><i>    MatAssemblyBegin (out_mat,
                  MAT_FINAL_ASSEMBLY);</i><i><br>
                </i><i>    MatAssemblyEnd (out_mat, MAT_FINAL_ASSEMBLY);</i><i><br>
                </i><i>}</i></p>
              <p>When replacing test function <i>test_scaling_petsc()</i>
                with <i>test_scaling_petsc_pointer()</i> everything
                works as it should, but I do not understand why.</p>
              <p>Do you have any suggestions?</p>
            </div>
          </blockquote>
          <div>The easiest explanation is that you have a memory
            overwrite in the code somewhere. Barry's suggestion to use</div>
          <div>valgrind is good.</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!<br>
              </p>
              <p><br>
              </p>
              <div>Am 05.01.21 um 15:24 schrieb Roland Richter:<br>
              </div>
              <blockquote type="cite">
                <p>Hei,</p>
                <p>the code I attached to the original mail should work
                  out of the box, but requires armadillo and PETSc to
                  compile/run. Armadillo stores the data in column-major
                  order, and therefore I am transposing the matrices
                  before and after transferring using .st().</p>
                <p>Thank you for your help!</p>
                <p>Regards,</p>
                <p>Roland<br>
                </p>
                <div>Am 05.01.21 um 15:21 schrieb Matthew Knepley:<br>
                </div>
                <blockquote type="cite">
                  <div dir="ltr">
                    <div dir="ltr">On Tue, Jan 5, 2021 at 7:57 AM Roland
                      Richter <<a href="mailto:roland.richter@ntnu.no" target="_blank">roland.richter@ntnu.no</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">Hei,<br>
                        <br>
                        I would like to scale a given matrix with a
                        fixed scalar value, and<br>
                        therefore would like to use MatScale().
                        Nevertheless, I observed an<br>
                        interesting behavior depending on the size of
                        the matrix, and currently<br>
                        I am not sure why.<br>
                        <br>
                        When running the attached code, I intend to
                        divide all elements in the<br>
                        matrix by a constant factor of 10. If I have
                        three or fewer rows and<br>
                        1024 columns, I get the expected result. If I
                        have four or more rows<br>
                        (with the same number of columns), suddenly my
                        scaling factor seems to<br>
                        be 0.01 instead of 0.1 for the PETSc-matrix. The
                        armadillo-based matrix<br>
                        still behaves as expected.<br>
                      </blockquote>
                      <div><br>
                      </div>
                      <div>1) It looks like you assume the storage in
                        your armadillo matrix is row major. I would be
                        surprised if this was true.</div>
                      <div><br>
                      </div>
                      <div>2) I think it is unlikely that there is a
                        problem with MatScale, so I would guess either
                        you have a memory overwrite</div>
                      <div>or are misinterpreting your output. If you
                        send something I can run, I will figure out
                        which it is.</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"> I currently
                        do not understand that behavior, but do not see
                        any problems<br>
                        with the code either. Are there any possible
                        explanations for that behavior?<br>
                        <br>
                        Thank you very much,<br>
                        <br>
                        regards,<br>
                        <br>
                        Roland Richter<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>
              </blockquote>
            </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>
  </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></div>