<div dir="ltr"><div>A change in the distribution of plane waves for parallel FFT, aimed towards increasing the scalability, may cause an increase of the execution time under usual conditions (the default behavior has been reverted to the previous one in the development version, though). Some other changes in the default behavior may also affect the execution time. In particular, the default for Davidson diagonalization is now to use a set of correction vectors of minimal dimension (twice the number of bands) instead of the previous one (four times). This spares some memory but may sometimes increase the execution time. The default number of processors for parallel linear algebra has also changed. Under some conditions, parallel linear algebra may however be slower than the serial one. None of the above justifies more than say a 25% difference or so. For a recent (this morning) comparison: <a href="https://gitlab.com/QEF/q-e/-/issues/294">https://gitlab.com/QEF/q-e/-/issues/294</a></div><div><br></div><div>Paolo<br>
</div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Tue, Feb 16, 2021 at 1:22 AM Stephen Zhang <<a href="mailto:lolzen@berkeley.edu">lolzen@berkeley.edu</a>> wrote:<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>Hi Everyone,</div><div><br></div><div>I'm currently using both the latest qe-6.7 and qe-6.4.1 qmcpack in my school computing HPC cluster and i've noticed that qe-6.4.1-qmcpack's version of pw.x is several times faster than qe-6.7 when performing scf and nscf runs. Both were configured for parallel runs using openmpi. Another difference is that while qe-6.4.1-qmcpack's pw.x is much faster during the run, it is also much slower when writing the data into prefix.save compared to qe-6.7, however this may be due to the fact that the 6.4.1-qmcpack's pw.x is configured to write the wavefunctions in the hdf5 format. The final difference I noticed was that qe-6.4.1-qmcpack's pw.x's filesize is much larger at 20mb compared to 4mb for qe-6.7's pw.x. <br></div><div><br></div><div>Thanks,</div><div>Stephen<br></div><br clear="all"><br>-- <br><div dir="ltr"><div dir="ltr"><i>University of California, Berkeley</i><div><i>Department of Letter and Sciences</i></div></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" class="gmail_signature"><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>