<div dir="ltr">Hypre embedded LAPACK in their source so I doubt you can have hypre use MKL. Hypre supports OMP now, but I have been having a lot of numerical problems with hypre/OMP on KNL.</div><br><div class="gmail_quote"><div dir="ltr">On Mon, Jul 30, 2018 at 9:24 AM Bastian Löhrer <<a href="mailto:bastian.loehrer@tu-dresden.de">bastian.loehrer@tu-dresden.de</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
  
    
  
  <div text="#000000" bgcolor="#FFFFFF">
    On 30.07.2018 13:28, Matthew Knepley wrote:<br>
    <blockquote type="cite">
      
      <div dir="ltr">
        <div class="gmail_quote">
          <div dir="ltr">On Mon, Jul 30, 2018 at 5:48 AM Bastian Löhrer
            <<a href="mailto:bastian.loehrer@tu-dresden.de" target="_blank">bastian.loehrer@tu-dresden.de</a>>
            wrote:<br>
          </div>
          <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
            <div text="#000000" bgcolor="#FFFFFF">
              <p>Dear PETSc users,</p>
              <p>im configuring PETSc to use an intel stack including
                intel/mkl:</p>
              <tt>./configure
PETSC_ARCH=$PETSC_ARCH                                                     
                \</tt><tt><br>
              </tt><tt>--with-gnu-compilers=0
                --with-vendor-compilers=intel                                   
                \</tt><tt><br>
              </tt><tt>--with-large-file-io=1                                                                 
                \</tt><tt><br>
              </tt><tt>--CFLAGS="-L${I_MPI_ROOT}/intel64/lib
                -I${I_MPI_ROOT}/intel64/include -lmpi"            \</tt><tt><br>
              </tt><tt>--CXXFLAGS="-L${I_MPI_ROOT}/intel64/lib
                -I${I_MPI_ROOT}/intel64/include -lmpi -lmpicxx" \</tt><tt><br>
              </tt><tt>--FFLAGS="-L${I_MPI_ROOT}/intel64/lib
                -I${I_MPI_ROOT}/intel64/include -lmpi"            \</tt><tt><br>
              </tt><tt>--LDFLAGS="-L${I_MPI_ROOT}/intel64/lib
                -I${I_MPI_ROOT}/intel64/include -lmpi"           \</tt><tt><br>
              </tt><tt>--with-blas-lapack-dir="${MKLROOT}/lib/intel64"                                        
                \</tt><tt><br>
              </tt><tt>--download-hypre                                                                       
                \</tt><tt><br>
              </tt><tt>--with-debugging=yes</tt><br>
              <br>
              <b><br>
                two questions:</b><br>
              <br>
              <b><br>
                1)</b> the blas-lapack-dir option is not passed down to
              the compilation of hypre according to
              $PETSC_DIR/$PETSC_ARCH/conf/hypre<br>
              <b>Is there a way of having PETSc compile hypre with my
                intel/mkl?</b></div>
          </blockquote>
          <div><br>
          </div>
          <div>This should happen. Please send configure.log so we can
            see what went on.</div>
        </div>
      </div>
    </blockquote>
    My initial guess was that the file $PETSC_DIR/$PETSC_ARCH/conf/hypre
    lists the parameters which are used for compiling hypre. As I said,
    this file does not mention mkl anywhere.<br>
    I may be mistaken though, because having a second look I do realize
    that the mkl library is mentioned in the configure log file in line
    86661 ff. (where hypre is being configured) and at the end of the
    log.<br>
    <br>
    Here is my configure.log:<br>
       
<a class="m_7761401067452335977moz-txt-link-freetext" href="https://cloudstore.zih.tu-dresden.de/index.php/s/b6rT0WMAKEMsj8S/download" target="_blank">https://cloudstore.zih.tu-dresden.de/index.php/s/b6rT0WMAKEMsj8S/download</a><br>
    <br>
    Here is that hypre file:<br>
       
<a class="m_7761401067452335977moz-txt-link-freetext" href="https://cloudstore.zih.tu-dresden.de/index.php/s/TSfXQ2pgDw5ALZm/download" target="_blank">https://cloudstore.zih.tu-dresden.de/index.php/s/TSfXQ2pgDw5ALZm/download</a><br>
    <br>
    Thanks,<br>
    Bastian<br>
    <br>
     <br>
    <blockquote type="cite">
      <div dir="ltr">
        <div class="gmail_quote">
          <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
            <div text="#000000" bgcolor="#FFFFFF"><b><br>
                2)</b> <b>Should I omit or include any option?</b> I
              have come across a few options in previous configuration
              calls used at my department which I have removed from my
              configuration call because <br>
              <blockquote>a) I had the impression that they were of no
                additional use:<br>
                <ul>
                  <li><tt>--with-cc=mpiicc --with-cxx=mpiicpc
                      --with-fc=mpiifort --with-mpi=1</tt></li>
                  <li><tt>--with-pic</tt>, because <tt>--with-pic=1</tt>
                    is the default anyway since 2.3.1, is it not? <br>
                    (<a class="m_7761401067452335977m_3080504431895625717moz-txt-link-freetext" href="https://www.mcs.anl.gov/petsc/documentation/changes/231.html" target="_blank">https://www.mcs.anl.gov/petsc/documentation/changes/231.html</a>)</li>
                  <li><tt>--with-mpiexec=mpirun</tt></li>
                  <li><tt>--with-mpi-compilers=1</tt></li>
                  <li><tt>--known-mpi-shared=1</tt></li>
                  <li><tt>--with-mpi-dir=...</tt></li>
                  <li><tt>COPTFLAGS="-g" CXXOPTFLAGS="-g" FOPTFLAGS="-g"</tt>
                    because <tt>--with-debugging=yes</tt> adds them
                    anyway</li>
                </ul>
              </blockquote>
            </div>
          </blockquote>
          <div>Yes, getting rid of all these is fine. </div>
          <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
            <div text="#000000" bgcolor="#FFFFFF">
              <blockquote>b) because I couldn't figure out what they
                were actually for:<br>
                <ul>
                  <li><tt><tt>--configModules=PE</tt><tt>TSc.Configure</tt></tt></li>
                  <li><tt><tt>--optionsModule=PETSc.compilerOptions</tt></tt><br>
                  </li>
                </ul>
              </blockquote>
              <blockquote> </blockquote>
            </div>
          </blockquote>
          <div>Those are added automatically. These are hooks so that
            you can completely change the system without</div>
          <div>getting rid of the low-level tests. </div>
          <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
            <div text="#000000" bgcolor="#FFFFFF">
              <blockquote>
                <p>c) others:</p>
                <ul>
                  <li><tt>--known-64-bit-blas-indices</tt> I guess it
                    wouldn't hurt anyway so I guess I'll include this
                    option the next time I configure petsc<br>
                    (<a class="m_7761401067452335977m_3080504431895625717moz-txt-link-freetext" href="http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Sys/PetscBLASInt.html" target="_blank">http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Sys/PetscBLASInt.html</a>)<br>
                  </li>
                </ul>
              </blockquote>
            </div>
          </blockquote>
          <div>We have a bunch of -known-* options. They are used for
            any test that has to execute. Batch environments cannot,</div>
          <div>and thus you need to specify all of them. You do not need
            it unless you are using --with-batch.</div>
        </div>
      </div>
    </blockquote>
    Okay, fantastic, thank you!<br>
  </div>

</blockquote></div>