<html>
  <head>
    <meta content="text/html; charset=windows-1252"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    Hi Lorenzo, I had a little problem with the 2016 version as well
    with intel compiler.<br>
    <i>If you compile the following snippet</i><br>
    <br>
    <p>program prova<br>
        use elpa1<br>
      end program prova<br>
    </p>
    <p><i>the compiler complains with:</i><br>
    </p>
    <p>[nvarini@deneb2 elpa-2016.05.003]$ mpiifort
      -I$PWD/include/elpa-2016.05.003/modules prova.f90 -L$PWD/lib<br>
      prova.f90(2): error #7002: Error in opening the compiled module
      file.  Check INCLUDE paths.   [PRECISION]<br>
        use elpa1<br>
      ------^<br>
      compilation aborted for prova.f90 (code 1)<br>
      <br>
    </p>
    <p>With the 2016 version there is a new directory called
      private_modules, which contain precision.mod:<br>
      [nvarini@deneb2 elpa-2016.05.003]$ ls
      private_modules/precision.mod <br>
      private_modules/precision.mod<br>
    </p>
    <p>This bug of the installation procedure should be fixed in the
      next release. The workaround is to copy manually
      private_modules/precision.mod inside the
      include/elpa-2016.05.003/modules</p>
    <p>Also, if you find a bug by using mkl scalapack during the
      diagonalization with the error:<br>
      <br>
    </p>
    <p>Fatal error in PMPI_Cart_sub: Other MPI error, error stack:<br>
       PMPI_Cart_sub(242)...................:
      MPI_Cart_sub(comm=0xc400fced, remain_dims=0x7ffe8e4e1f68,
      comm_new=0x7ffe8e4e1ec0) failed<br>
       PMPI_Cart_sub(178)...................:<br>
       MPIR_Comm_split_impl(270)............:<br>
       MPIR_Get_contextid_sparse_group(1330): Too many communicators
      (5/16384 free on this process; ignore_id=0)<br>
       Fatal error in PMPI_Cart_sub: Other MPI error, error stack:</p>
    <p>It is likely to fix this by using elpa.<br>
    </p>
    <p>Nic<br>
    </p>
    <p><br>
    </p>
    <br>
    <div class="moz-cite-prefix">On 09/22/2016 02:17 PM, Lorenzo
      Paulatto wrote:<br>
    </div>
    <blockquote
cite="mid:CAG+GtJf7M7Hc3ZkPM0G5iVzxPtXr84oGA6r0KUqF3Lg3sv3XjQ@mail.gmail.com"
      type="cite">
      <meta http-equiv="Content-Type" content="text/html;
        charset=windows-1252">
      <div dir="ltr">By the way, when I finished managed to compile elpa
        2016 (thanks to Andreas Marek for helping) I noticed that the
        __ELPA_2016 interface does not match a couple of calls, notably
        get_elpa_row_col_comms returns an integer. Here is the patch,
        I'm not uploading it on SVN, in case it does not match your
        findings.<br>
        <br>
        Index: dspev_drv.f90<br>
===================================================================<br>
        --- dspev_drv.f90       (revision 12981)<br>
        +++ dspev_drv.f90       (working copy)<br>
        @@ -720,7 +720,7 @@<br>
              CALL BLACS_Gridinfo(ortho_cntx,nprow, npcol,
        my_prow,my_pcol)<br>
         <br>
         #if defined(__ELPA_2016)<br>
        -     success = get_elpa_row_col_comms(ortho_comm, my_prow,
        my_pcol,mpi_comm_rows, mpi_comm_cols)<br>
        +     ierr    = get_elpa_row_col_comms(ortho_comm, my_prow,
        my_pcol,mpi_comm_rows, mpi_comm_cols)<br>
              success = solve_evp_real_1stage(n,  n,   s, lds,    w, 
        vv, lds,SIZE(s,2),nb  ,mpi_comm_rows, mpi_comm_cols)<br>
         #elif defined(__ELPA_2015)<br>
              ierr = get_elpa_row_col_comms(ortho_comm, my_prow,
        my_pcol,mpi_comm_rows, mpi_comm_cols)<br>
        Index: zhpev_drv.f90<br>
===================================================================<br>
        --- zhpev_drv.f90       (revision 12981)<br>
        +++ zhpev_drv.f90       (working copy)<br>
        @@ -1522,7 +1522,7 @@<br>
         #if defined(__ELPA) || defined(__ELPA_2016) ||
        defined(__ELPA_2015)<br>
              CALL BLACS_Gridinfo(ortho_cntx,nprow, npcol,
        my_prow,my_pcol)<br>
         #if defined(__ELPA_2016)<br>
        -     success = get_elpa_row_col_comms(ortho_comm, my_prow,
        my_pcol,mpi_comm_rows, mpi_comm_cols)<br>
        +     ierr    = get_elpa_row_col_comms(ortho_comm, my_prow,
        my_pcol,mpi_comm_rows, mpi_comm_cols)<br>
              success = solve_evp_complex(n, n, h, size(h,1), w,  v,
        size(h,1), size(h,2), nb, &<br>
                                    mpi_comm_rows, mpi_comm_cols)<br>
         #elif defined(__ELPA_2015)<br>
        <br>
        <br>
        <br>
        <br>
      </div>
      <div class="gmail_extra"><br>
        <div class="gmail_quote">On 20 September 2016 at 17:42, Lorenzo
          Paulatto <span dir="ltr"><<a moz-do-not-send="true"
              href="mailto:lorenzo.paulatto@impmc.upmc.fr"
              target="_blank">lorenzo.paulatto@impmc.upmc.fr</a>></span>
          wrote:<br>
          <blockquote class="gmail_quote" style="margin:0 0 0
            .8ex;border-left:1px #ccc solid;padding-left:1ex">
            <div dir="ltr">
              <div>Hello,<br>
              </div>
              I've just spent several hours trying to compile elpa 2016
              on BlueGENE (and I did not manage...). I'm pretty
              skeptical about the ability of the average user to compile
              it at all on anything else than intel compiler/mpi/mkl,
              but this is a common problem, hence I'm not against the
              change in principle.<br>
            </div>
            <div class="gmail_extra">
              <div>
                <div class="h5"><br>
                  <div class="gmail_quote">On 20 September 2016 at
                    00:01, Filippo SPIGA <span dir="ltr"><<a
                        moz-do-not-send="true"
                        href="mailto:filippo.spiga@quantum-espresso.org"
                        target="_blank">filippo.spiga@quantum-<wbr>espresso.org</a>></span>
                    wrote:<br>
                    <blockquote class="gmail_quote" style="margin:0 0 0
                      .8ex;border-left:1px #ccc solid;padding-left:1ex">Dear
                      everyone,<br>
                      <br>
                      I did several tests in the past days to verify
                      results match between SCALAPACK, ELPA 2015, ELPA
                      2016 and the "custom diagonalization". I generated
                      a lot of outputs, I did not compare every single
                      one of them but all of those I have been inspected
                      are all good and consistent.<br>
                      <br>
                      Because the current ELPA library structure,
                      self-detect the ELPA version using autoconf is
                      quite complex and I cannot guardantee it works
                      every time on every machine. Things need to be
                      simplified a bit on QE side (and as consequence
                      rely more on the ability of the user to compile a
                      library for his/her own environment).<br>
                      <br>
                      Instead of support both "2015" API and "2016" API,
                      I suggest to support via configure only the 2016
                      version. I propose the following changes in the
                      configure:<br>
                      - deprecate generic "--with-elpa"<br>
                      - add "--with-elpa-lib=<path>" which point
                      to the directory of the ELPA library<br>
                      - add "--with-elpa-include=<path>" which
                      point to the directory of the ELPA Fortran module<br>
                      <br>
                      ELPA is enabled ("__ELPA_2016" in DFLAGS) iff
                      both  "--with-elpa-lib=<path>" (".so" or
                      ".a" library present in <path>) and
                      "--with-elpa-include=<path>" ("elpa1.mod"
                      present in <path>) are specified. If someone
                      wants to use the "2015" API then it is always
                      possible to edit manually <a
                        moz-do-not-send="true" href="http://make.in"
                        rel="noreferrer" target="_blank">make.in</a>
                      ("__ELPA_2016" will continue to exhist in the
                      code).<br>
                      <br>
                      I am planning to commit changes around Wed or Thu
                      unless there are particular objections.<br>
                      <br>
                      Cheers<br>
                      <div>
                        <div><br>
                          <br>
                          On Sep 11, 2016, at 2:37 PM, Filippo SPIGA
                          <<a moz-do-not-send="true"
                            href="mailto:filippo.spiga@quantum-espresso.org"
                            target="_blank">filippo.spiga@quantum-espress<wbr>o.org</a>>
                          wrote:<br>
                          > Dear all,<br>
                          ><br>
                          > ELPA (Eigenvalue SoLvers for
                          Petaflop-Applications) is a parallel
                          eigen-solver library working with QE and
                          provided natively via a self-hosted package
                          under "archive/". This package contains a old
                          version of ELPA. The latest official ELPA
                          release available on the web adopts different
                          routine API (small changes but enought to
                          require a change in the code).<br>
                          ><br>
                          > After some changes in the code, we now
                          support two versions: ELPA 2015.11.001
                          (__ELPA_2015) and ELPA 2016.05.003
                          (__ELPA_2016). Again, these two versions have
                          different routine API and these discrepancies
                          force us to maintain two API in the code. It
                          is ugly but it is nothing we can do about.<br>
                          ><br>
                          > The old ELPA in the code will be
                          deprecated by v6.0 is going to be released. I
                          am working on a configure step that
                          transparently detects ELPA 2015 versus ELPA
                          2016. If I fails, I will work around a
                          mechanism that allows to pass to the configure
                          both the location of the shared library and
                          the location of the Fortran modules associated
                          with it.<br>
                          ><br>
                          > Stay tuned<br>
                          ><br>
                          > Cheers<br>
                          <br>
                          --<br>
                          Filippo SPIGA ~ Quantum ESPRESSO Foundation ~
                          <a moz-do-not-send="true"
                            href="http://www.quantum-espresso.org"
                            rel="noreferrer" target="_blank">http://www.quantum-espresso.or<wbr>g</a><br>
                          <br>
                          <br>
                          ______________________________<wbr>_________________<br>
                          Q-e-developers mailing list<br>
                          <a moz-do-not-send="true"
                            href="mailto:Q-e-developers@qe-forge.org"
                            target="_blank">Q-e-developers@qe-forge.org</a><br>
                          <a moz-do-not-send="true"
                            href="http://qe-forge.org/mailman/listinfo/q-e-developers"
                            rel="noreferrer" target="_blank">http://qe-forge.org/mailman/li<wbr>stinfo/q-e-developers</a><br>
                        </div>
                      </div>
                    </blockquote>
                  </div>
                  <br>
                  <br clear="all">
                  <br>
                </div>
              </div>
              <span class="HOEnZb"><font color="#888888">-- <br>
                  <div data-smartmail="gmail_signature">
                    <div>Dr. Lorenzo Paulatto </div>
                    <div>IdR @ IMPMC -- CNRS & Université Paris 6</div>
                    <div>phone: +33 (0)1 44275 084 / skype: paulatz</div>
                    <div>www:   <a moz-do-not-send="true"
                        href="http://www-int.impmc.upmc.fr/%7Epaulatto/"
                        target="_blank">http://www-int.impmc.upmc.fr/~<wbr>paulatto/</a></div>
                    <div>mail:  23-24/4é16 Boîte courrier 115, 4 place
                      Jussieu 75252 Paris Cédex 05</div>
                  </div>
                </font></span></div>
          </blockquote>
        </div>
        <br>
        <br clear="all">
        <br>
        -- <br>
        <div class="gmail_signature" data-smartmail="gmail_signature">
          <div>Dr. Lorenzo Paulatto </div>
          <div>IdR @ IMPMC -- CNRS & Université Paris 6</div>
          <div>phone: +33 (0)1 44275 084 / skype: paulatz</div>
          <div>www:   <a moz-do-not-send="true"
              href="http://www-int.impmc.upmc.fr/%7Epaulatto/"
              target="_blank">http://www-int.impmc.upmc.fr/~paulatto/</a></div>
          <div>mail:  23-24/4é16 Boîte courrier 115, 4 place Jussieu
            75252 Paris Cédex 05</div>
        </div>
      </div>
      <br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <br>
      <pre wrap="">_______________________________________________
Q-e-developers mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Q-e-developers@qe-forge.org">Q-e-developers@qe-forge.org</a>
<a class="moz-txt-link-freetext" href="http://qe-forge.org/mailman/listinfo/q-e-developers">http://qe-forge.org/mailman/listinfo/q-e-developers</a>
</pre>
    </blockquote>
    <br>
    <pre class="moz-signature" cols="72">-- 
Nicola Varini, PhD

Scientific IT and Application Support (SCITAS)
Theory and simulation of materials (THEOS) 
ME B2 464 (Bâtiment ME)
Station 1
CH-1015 Lausanne
+41 21 69 31332
<a class="moz-txt-link-freetext" href="http://scitas.epfl.ch">http://scitas.epfl.ch</a>

Nicola Varini </pre>
  </body>
</html>