<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <div class="moz-cite-prefix">Hi Chris <br>
      it might it be happening exactly the opposite. <br>
      if you don't specify anything the configure tries all the options
      from the best to the worse and the usage for mkl is tested as
      first guess if  I am not wrong. If you pass it a specific path
      just tries that one and deals  with it as expecting ordinary fftw
      library, so it may be failing in finding a working fft and turns
      on the internal one. <br>
      <br>
      Could you send the make.inc files in the 2 cases or the config log
      ? <br>
      Pietro <br>
      <br>
      On 03/01/2019 11:13 AM, Christoph Wolf wrote:<br>
    </div>
    <blockquote type="cite"
cite="mid:CAMC_G_44XRJtUKinXWm-k-p8bq=Lc2P1Ls+TArX2v_GOp8CDhA@mail.gmail.com">
      <meta http-equiv="content-type" content="text/html; charset=utf-8">
      <div dir="ltr">
        <div dir="ltr">
          <div dir="ltr">Dear all,
            <div><br>
            </div>
            <div>please forgive this "beginner" question but I am facing
              a weird problem. When compiling qe-6.4 (intel compiler,
              intel MPI+OpenMP) with or without intel's fftw libs I find
              that in openMP with 2 threads per core the intel fftw
              version is roughly "twice as slow" as the internal one</div>
            <div><br>
            </div>
            <div>"internal"</div>
            <div>
              <div>     General routines</div>
              <div>     calbec       :      2.69s CPU      2.70s WALL ( 
                   382 calls)</div>
              <div>     fft          :      0.47s CPU      0.47s WALL ( 
                   122 calls)</div>
              <div>     ffts         :      0.05s CPU      0.05s WALL ( 
                    12 calls)</div>
              <div>     fftw         :     49.97s CPU     50.12s WALL ( 
                 14648 calls)</div>
              <div> </div>
              <div>     Parallel routines</div>
              <div> </div>
              <div>     PWSCF        :  1m45.03s CPU     1m46.59s WALL</div>
              <div><br>
              </div>
              <div>"intel fftw"</div>
              <div>
                <div>     General routines</div>
                <div>     calbec       :      6.36s CPU      3.20s WALL
                  (     382 calls)</div>
                <div>     fft          :      0.93s CPU      0.47s WALL
                  (     121 calls)</div>
                <div>     ffts         :      0.10s CPU      0.05s WALL
                  (      12 calls)</div>
                <div>     fftw         :    109.63s CPU     55.23s WALL
                  (   14648 calls)</div>
                <div> </div>
                <div>     Parallel routines</div>
                <div> </div>
                <div>     PWSCF        :   3m18.32s CPU   1m41.01s WALL</div>
              </div>
              <div><br>
              </div>
              <div>as a benchmark I am running a perovskite with 120
                k-points on 30 processors (one node); There is no
                (noticeable) difference if I export OMP_NUM_THREADS=1
                (only MPI) so I guess I made some mistake during the
                build with regards to the libraries.</div>
              <div><br>
              </div>
              <div>Build process is as below</div>
              <div><br>
              </div>
              <div>
                <p
class="gmail-m_-8828478814826450334gmail-m_436820312386234918MsoListParagraph"
                  style="margin:0cm 0cm 0.0001pt
                  56pt;font-size:12pt;font-family:굴림;text-align:justify"><span
style="font-size:10pt;font-family:"\00b9d1\00c740
                    \00ace0\00b515"" lang="EN-US">module load
                    intel19/compiler-19</span></p>
                <p
class="gmail-m_-8828478814826450334gmail-m_436820312386234918MsoListParagraph"
                  style="margin:0cm 0cm 0.0001pt
                  56pt;font-size:12pt;font-family:굴림;text-align:justify"><span
style="font-size:10pt;font-family:"\00b9d1\00c740
                    \00ace0\00b515"" lang="EN-US">module load
                    intel19/impi-19</span><br>
                </p>
                <p
class="gmail-m_-8828478814826450334gmail-m_436820312386234918MsoListParagraph"
                  style="margin:0cm 0cm 0.0001pt
                  56pt;font-size:12pt;font-family:굴림;text-align:justify"><span
style="font-size:10pt;font-family:"\00b9d1\00c740
                    \00ace0\00b515"" lang="EN-US"><br>
                  </span></p>
                <p
class="gmail-m_-8828478814826450334gmail-m_436820312386234918MsoListParagraph"
                  style="margin:0cm 0cm 0.0001pt
                  56pt;text-align:justify"><span lang="EN-US"><font
                      face="맑은 고딕"><span style="font-size:13.3333px">export
                        FFT_LIBS="-L$MKLROOT/intel64"</span></font><br>
                  </span></p>
                <p
class="gmail-m_-8828478814826450334gmail-m_436820312386234918MsoListParagraph"
                  style="margin:0cm 0cm 0.0001pt
                  56pt;text-align:justify"><span lang="EN-US"><font
                      face="맑은 고딕"><span style="font-size:13.3333px">export
                        LAPACK_LIBS="-lmkl_blacs_intelmpi_lp64"</span></font><br>
                  </span></p>
                <div>
                  <p
class="gmail-m_-8828478814826450334gmail-m_436820312386234918MsoListParagraph"
                    style="margin:0cm 0cm 0.0001pt
                    56pt;font-size:12pt;font-family:굴림;text-align:justify"><span
style="font-size:10pt;font-family:"\00b9d1\00c740
                      \00ace0\00b515"" lang="EN-US"><span
                        class="gmail-m_-8828478814826450334gmail-il">export</span> CC=icc <span
                        class="gmail-m_-8828478814826450334gmail-il">FC</span>=ifort
                      F77=ifort MPIF90=mpiifort MPICC=mpiicc</span></p>
                  <p
class="gmail-m_-8828478814826450334gmail-m_436820312386234918MsoListParagraph"
                    style="margin:0cm 0cm 0.0001pt
                    56pt;font-size:12pt;font-family:굴림;text-align:justify"><span
style="font-size:10pt;font-family:"\00b9d1\00c740
                      \00ace0\00b515"" lang="EN-US"><br>
                    </span></p>
                  <p
class="gmail-m_-8828478814826450334gmail-m_436820312386234918MsoListParagraph"
                    style="margin:0cm 0cm 0.0001pt
                    56pt;text-align:justify"><font face="맑은 고딕"><span
                        style="font-size:13.3333px">./configure
                        --enable-parallel --with-scalapack=intel
                        --enable-openmp</span></font><br>
                  </p>
                  <p
class="gmail-m_-8828478814826450334gmail-m_436820312386234918MsoListParagraph"
                    style="margin:0cm 0cm 0.0001pt
                    56pt;text-align:justify"><font face="맑은 고딕"><span
                        style="font-size:13.3333px"><br>
                      </span></font></p>
                  This detects BLAS_LIBS, LAPACK_LIBS, SCALAPACK_LIBS
                  and FFT_LIBS.</div>
              </div>
              <div><br>
              </div>
              <div>I am not experienced with benchmarking so if my
                benchmark is garbage please suggest a suitable system!</div>
              <div><br>
              </div>
              <div>Thanks in advance!</div>
              <div>Chris </div>
              <div><br>
              </div>
              -- <br>
              <div dir="ltr" class="gmail_signature">
                <div dir="ltr">Postdoctoral Researcher<br>
                  Center for Quantum Nanoscience, Institute for Basic
                  Science<br>
                  Ewha Womans University, Seoul, South Korea</div>
              </div>
            </div>
          </div>
        </div>
      </div>
      <br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <br>
      <pre wrap="">_______________________________________________
users mailing list
<a class="moz-txt-link-abbreviated" href="mailto:users@lists.quantum-espresso.org">users@lists.quantum-espresso.org</a>
<a class="moz-txt-link-freetext" href="https://lists.quantum-espresso.org/mailman/listinfo/users">https://lists.quantum-espresso.org/mailman/listinfo/users</a></pre>
    </blockquote>
    <p><br>
    </p>
  </body>
</html>