<div dir="ltr"><div dir="ltr">On Thu, Jan 21, 2021 at 11:36 PM Aaron Friesz <<a href="mailto:friesz@usc.edu" target="_blank">friesz@usc.edu</a>> wrote:</div><div dir="ltr"><br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr">1 ) In order to fix this it is stated that a change can be made in <span style="color:rgb(0,0,0);white-space:pre-wrap">FFTXlib/fft_param.f90. This implies that it must then be recompiled. Is this correct? (I'd rather avoid this as it will make me responsible for maintaining the software, instead of our computing resources division.)</span></div></blockquote><div><br></div><div>if the origin of your problem is the one descripted in the thread ( guess so), the proper fix is this one:</div><div> <a href="https://lists.quantum-espresso.org/pipermail/users/attachments/20190425/2584418a/attachment.obj">https://lists.quantum-espresso.org/pipermail/users/attachments/20190425/2584418a/attachment.obj</a></div><div>and yes, you have to recompile. Alternatively: &system ... force_symmorphic=.true. ... / will not look for symmetry operations with associated fractional translations and should solve the problem. Alternatively, install a newer version.<br></div><div><br></div><div>If you really want to allow FFT dimensions > 2048 (the hard-coded value; 5076 is what you have) you have to modify the code and recompile. Note that the typical real-space resolution is something of the order 0.1 A, so 2048 corresponds to approx. 200 A cell side: quite a big cell<br></div><div><br></div><div>The FFT is three-dimensional, so each one-dimensional FFTis very fast, but you need a bunch of them!<br></div><div><br></div><div>Paolo</div><div><br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div><div><span style="color:rgb(0,0,0);white-space:pre-wrap"></span></div><div><span style="color:rgb(0,0,0);white-space:pre-wrap">2 ) </span><span style="color:rgb(0,0,0);white-space:pre-wrap">Is the number in parentheses, (5076), the number of terms in the fft? Are there any heuristics to reducing this number short of reducing the number of atoms in the system?</span></div><div><span style="color:rgb(0,0,0);white-space:pre-wrap"><br></span></div><div><span style="color:rgb(0,0,0);white-space:pre-wrap">3 ) Why is this check made? Is the FFT computed in every loop of the algorithm, thus incurring a large computational cost? (From what I remember FFTW should be very fast on arrays of this size, but if it is calculated often that could be an issue.)</span></div><div><span style="color:rgb(0,0,0);white-space:pre-wrap"><br></span></div><div><span style="color:rgb(0,0,0);white-space:pre-wrap">Thank you for your help!</span></div><div><br></div></div></div>
_______________________________________________<br>
Quantum ESPRESSO is supported by MaX (<a href="http://www.max-centre.eu" rel="noreferrer" target="_blank">www.max-centre.eu</a>)<br>
users mailing list <a href="mailto:users@lists.quantum-espresso.org" target="_blank">users@lists.quantum-espresso.org</a><br>
<a href="https://lists.quantum-espresso.org/mailman/listinfo/users" rel="noreferrer" target="_blank">https://lists.quantum-espresso.org/mailman/listinfo/users</a></blockquote></div><br clear="all"><br>-- <br><div dir="ltr"><div dir="ltr"><div><div dir="ltr"><div>Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,<br>Univ. Udine, via delle Scienze 206, 33100 Udine, Italy<br>Phone +39-0432-558216, fax +39-0432-558222<br><br></div></div></div></div></div></div>