<div dir="ltr"><br><br><div class="gmail_quote">On Wed, May 13, 2015 at 10:24 PM Berk Hess &lt;<a href="mailto:hess@kth.se">hess@kth.se</a>&gt; wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
  
    
  
  <div bgcolor="#FFFFFF" text="#000000">
    <div>Hi,<br>
      <br>
      I also noticed only now that there is more logic on line 158 of
      FindFFTW.cmake that probably should have been updated (as well as
      the documentation at the top). I should have stayed out of the
      business of editing cmake files :-(</div></div></blockquote><div><br></div><div>It&#39;s not exactly your fault that nobody thought to comment anywhere that FindFFTW.cmake and the own-fftw build are supposed to fulfil the same contract...</div><div><br></div><div>Anyway, bring on the day when we can do our own FFTs! ;-)</div><div><br></div><div>Mark</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div bgcolor="#FFFFFF" text="#000000"><div><br>
      <br>
      Berk</div></div><div bgcolor="#FFFFFF" text="#000000"><div><br>
      <br>
      On 05/13/2015 10:21 PM, Mark Abraham wrote:<br>
    </div></div><div bgcolor="#FFFFFF" text="#000000">
    <blockquote type="cite">
      
      <div dir="ltr">Hi,<br>
        <br>
        <div class="gmail_quote">On Wed, May 13, 2015 at 8:54 PM Berk
          Hess &lt;<a href="mailto:hess@kth.se" target="_blank">hess@kth.se</a>&gt;
          wrote:<br>
          <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
            <div bgcolor="#FFFFFF" text="#000000">
              <div>Hi,<br>
                <br>
                I had a look at my change to the FFTW checks, but I
                don&#39;t see how something can go wrong here. As far as I
                can see the order and essential logic have not changed
                compared to 5.0.4.<br>
              </div>
            </div>
          </blockquote>
          <div><br>
          </div>
          <div>The issue is on line 91. The new AND NOT works only under
            the assumption that FFTWF_FOUND means that we ran the
            SIMD-flavour trycompile tests. But the auto-download sets
            FFTWF_FOUND *and* FFTWF_HAVE_SIMD (because it will have
            SIMD) and runs/fakes none of the tests. The previous logic
            was positive, so the issue did not arise. One fix is to
            change src/contrib/fftw/CMakeLists.txt to set a fake
            ${FFTW}_HAVE_SSE2.</div>
          <div><br>
          </div>
          <div>Ideally, we could have a sane test that the auto-download
            works, but even if we&#39;d had one, the build would have worked
            and we likely wouldn&#39;t have verified the warning-issuing
            behaviour :-(</div>
          <div><br>
          </div>
          <div>Mark</div>
          <div><br>
          </div>
          <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
            <div bgcolor="#FFFFFF" text="#000000">
              <div> Cheers,<br>
                <br>
                Berk</div>
            </div>
            <div bgcolor="#FFFFFF" text="#000000">
              <div><br>
                <br>
                On 05/13/2015 05:54 PM, Mark Abraham wrote:<br>
              </div>
            </div>
            <div bgcolor="#FFFFFF" text="#000000">
              <blockquote type="cite">
                <div dir="ltr">Hi,
                  <div><br>
                  </div>
                  <div>Thanks for the email.</div>
                  <div><br>
                    <div class="gmail_quote">On Wed, May 13, 2015 at
                      5:34 PM Hardy, Adam &lt;<a href="mailto:ah259@hw.ac.uk" target="_blank">ah259@hw.ac.uk</a>&gt;
                      wrote:<br>
                      <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Dear all,<br>
                        <br>
                        I&#39;m having trouble installing the latest
                        release. Leaving all other options default, but
                        switching GMX_BUILD_OWN_FFTW=yes results in a
                        cmake warning:<br>
                        <br>
                        &quot;<br>
                        CMake Warning at
                        cmake/gmxManageFFTLibraries.cmake:93 (message):<br>
                           The FFTW library was compiled with neither
                        --enable-sse nor --enable-sse2;<br>
                           those would have enabled SSE(2) SIMD
                        instructions.  This will give<br>
                           suboptimal performance.  You should
                        (re)compile the FFTW library with both<br>
                           SSE2 and AVX instruction support (use both
                        --enable-sse2 and --enable-avx).<br>
                           The FFTW library will determine at runtime
                        which SIMD instruction set is<br>
                           fastest for different parts of the FFTs.<br>
                         Call Stack (most recent call first):<br>
                           CMakeLists.txt:742 (include) </blockquote>
                      <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">&quot;</blockquote>
                      <div> </div>
                      <div>That&#39;s inconsistent and somewhat embarassing,
                        but not actually a problem. FFTW hasn&#39;t been
                        downloaded or built by the time that warning is
                        issued, so we have some logic that doesn&#39;t work
                        correctly in the auto-download case.</div>
                      <div><br>
                      </div>
                      <div>Later on, FFTW is built but is still compiled
                        with only --enable-sse2. We may change that
                        behaviour for 5.1, if it is clear that adding
                        --enable-avx as well is always
                        neutral-or-useful, but there&#39;s nothing wrong
                        with the resulting mdrun here.</div>
                      <div><br>
                      </div>
                      <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> This configuration
                        works as expected with 5.0.4.<br>
                        <br>
                        I get the same results with this setting as
                        &quot;no&quot;. This message doesn&#39;t make much sense in
                        either case as I do not have an FFTW library on
                        my machine.</blockquote>
                      <div><br>
                      </div>
                      <div>Hmm that could be a minor problem, but none
                        of us have tried a build on a machine without
                        FFTW, for the obvious reason. :-)</div>
                      <div> </div>
                      <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">The error I get in 5.0.4
                        for this scenario is:<br>
                        <br>
                        &quot;<br>
                        Could not find fftw3f library named libfftw3f,
                        please specify its location in CMAKE_PREFIX_PATH
                        or FFTWF_LIBRARY by hand (e.g.<br>
                         -DFFTWF_LIBRARY=&#39;/path/to/libfftw3f.so&#39;)<br>
                        <br>
                         CMake Error at
                        cmake/gmxManageFFTLibraries.cmake:76 (MESSAGE):<br>
                           Cannot find FFTW 3 (with correct precision -
                        libfftw3f for mixed-precision<br>
                           GROMACS or libfftw3 for double-precision
                        GROMACS).  Either choose the right<br>
                           precision, choose another FFT(W) library
                        (-DGMX_FFT_LIBRARY), enable the<br>
                           advanced option to let GROMACS build FFTW 3
                        for you<br>
                           (-GMX_BUILD_OWN_FFTW=ON), or use the really
                        slow GROMACS built-in fftpack<br>
                           library (-DGMX_FFT_LIBRARY=fftpack).<br>
                         Call Stack (most recent call first):<br>
                           CMakeLists.txt:738 (include)<br>
                        &quot;<br>
                        <br>
                        I see some in the release notes some changes
                        were made to the cmake warnings for AVX capable
                        processors which I belive mine is </blockquote>
                      <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">(AMD fx-8350).<br>
                      </blockquote>
                      <div><br>
                      </div>
                      <div>Yes, but we don&#39;t expect much/any improvement
                        on AMD cpus from adding --enable-avx. Use
                        --enable-sse2 always (which is what the
                        auto-download does).</div>
                      <div><br>
                      </div>
                      <div>Mark</div>
                      <div><br>
                      </div>
                      <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> Thanks,<br>
                        <br>
                        Adam Hardy<br>
                        PhD Student<br>
                        School of Engineering and Physical Sciences<br>
                        Heriot-Watt University<br>
                        Edinburgh EH14 4AS, UK<br>
                        <br>
                        -----<br>
                        We invite research leaders and ambitious early
                        career researchers to<br>
                        join us in leading and driving research in key
                        inter-disciplinary themes.<br>
                        Please see <a href="http://www.hw.ac.uk/researchleaders" target="_blank">www.hw.ac.uk/researchleaders</a>
                        for further information and how<br>
                        to apply.<br>
                        <br>
                        Heriot-Watt University is a Scottish charity<br>
                        registered under charity number SC000278.<br>
                        <br>
                        --<br>
                        Gromacs Developers mailing list<br>
                        <br>
                        * Please search the archive at <a href="http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List" target="_blank">http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List</a>
                        before posting!<br>
                        <br>
                        * Can&#39;t post? Read <a href="http://www.gromacs.org/Support/Mailing_Lists" target="_blank">http://www.gromacs.org/Support/Mailing_Lists</a><br>
                        <br>
                        * For (un)subscribe requests visit<br>
                        <a href="https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers" target="_blank">https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers</a>
                        or send a mail to <a href="mailto:gmx-developers-request@gromacs.org" target="_blank">gmx-developers-request@gromacs.org</a>.<br>
                      </blockquote>
                    </div>
                  </div>
                </div>
                <br>
                <fieldset></fieldset>
                <br>
              </blockquote>
              <br>
            </div>
            --<br>
            Gromacs Developers mailing list<br>
            <br>
            * Please search the archive at <a href="http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List" target="_blank">http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List</a>
            before posting!<br>
            <br>
            * Can&#39;t post? Read <a href="http://www.gromacs.org/Support/Mailing_Lists" target="_blank">http://www.gromacs.org/Support/Mailing_Lists</a><br>
            <br>
            * For (un)subscribe requests visit<br>
            <a href="https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers" target="_blank">https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers</a>
            or send a mail to <a href="mailto:gmx-developers-request@gromacs.org" target="_blank">gmx-developers-request@gromacs.org</a>.</blockquote>
        </div>
      </div>
      <br>
      <fieldset></fieldset>
      <br>
    </blockquote>
    <br>
  </div>

--<br>
Gromacs Developers mailing list<br>
<br>
* Please search the archive at <a href="http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List" target="_blank">http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List</a> before posting!<br>
<br>
* Can&#39;t post? Read <a href="http://www.gromacs.org/Support/Mailing_Lists" target="_blank">http://www.gromacs.org/Support/Mailing_Lists</a><br>
<br>
* For (un)subscribe requests visit<br>
<a href="https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers" target="_blank">https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers</a> or send a mail to <a href="mailto:gmx-developers-request@gromacs.org" target="_blank">gmx-developers-request@gromacs.org</a>.</blockquote></div></div>