<div dir="ltr"><div><div>Dear QE developers, <br><br></div>Using the latest version of the QE trunk (r 12199), I noticed that if I run an nscf calculation<br></div>with a number of processor smaller than the total number of k-points but not a multiple of it it crashes with<br><div><div><div><br><div> stopping ...<br> Error in routine diropn (10):<br> can't open a connected unit<br> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%<br><br></div><div>I have 216 k-points.<br></div><div><br></div><div>I did <br>mpirun -np 120 $PW -npool 120 < <a href="http://nscf.in">nscf.in</a><br></div><div><br></div><div>If I do instead <br></div><div>mpirun -np 72 $PW -npool 72 < <a href="http://nscf.in">nscf.in</a><br><br></div><div>It works perfectly. <br><br></div><div>Although I perfectly understand that 120 cores leads to a clear load imbalance, it should still be possible to run it (I guess). <br><br></div><div>If I recall correctly, this was working before (I might be wrong)?<br><br></div><div>If it is not suppose to be working, it might be nice to have a error message similar to the one you get if you use more cpu than k-points. <br><br></div><div>Best, <br><br></div><div>Samuel.<br></div><div><br>-- <br><div class="gmail_signature"><div dir="ltr"><span><font color="#888888"><pre cols="72">------------------------------------------------------------------------------------------------
Dr. Samuel Poncé
Department of Materials
University of Oxford
Parks Road
Oxford OX1 3PH, UK
Phone: +44 1865 612789<br> email: <a href="mailto:samuel.ponce@materials.ox.ac.uk" target="_blank">samuel.ponce@materials.ox.ac.uk</a> <a href="mailto:fabio.caruso@materials.ox.ac.uk" target="_blank"></a>
web: <a href="http://giustino.materials.ox.ac.uk/index.php/Site/SamuelPonc%e9" target="_blank">http://giustino.materials.ox.ac.uk/index.php/Site/SamuelPonc%e9</a>
------------------------------------------------------------------------------------------------</pre></font></span></div></div>
</div></div></div></div></div>