<html><head></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space; ">Dear all,<div><br></div><div>I'm doing some test runs on bulk silicon, using QE 4.0.5. I need to calculate the eigenvalues on dense k-points grids,</div><div>to test convergence of some properties. So, I did a simple scf + nscf runs with with increasing (automatic) k-point grid</div><div>(from 8 8 8 1 1 1 to 24 24 24 1 1 1). The calculation is parallel and uses 8 CPUs.</div><div><br></div><div>All the grids with a number of k-points > 804 show a strange behaviour, namely, that once reached the 804th k-point</div><div>(so I see in the output the "Computing kpt #: 804" line) the calculation remains running but no update of output files is done.</div><div><br></div><div>Has anybody experienced such kind of problem?</div><div><br></div><div><br></div><div>Just to help (me and you) to understand what is going on, I did some debug:</div><div><br></div><div>i) the problem disappears if I use -ndiag 1</div><div>(consider that if not specified, by default the parallel algorithm is used, the output header contains</div><div><div>Iterative solution of the eigenvalue problem</div><div><br></div><div> a parallel distributed memory algorithm will be used,</div><div> eigenstates matrixes will be distributed block like on</div><div> ortho sub-group = 2* 2 procs</div><div>)</div><div>Of course, I don't know if in this case (that is, with -ndiag 1) the problem could appear with much denser grids (e.g. 64 64 64 1 1 1), but</div><div>I would say not</div><div><br></div><div>ii) the problem is maybe a memory (allocation/deallocation?????) issue because (see also point iv), if the parallel algorithm is used (which is what the code</div><div>chooses by default), the problem also disappears on decreasing the cutoff from 30 Ry to 15 Ry</div><div><br></div><div>iii) the code stops (in the sense that keeps running without doing anything) in SUBROUTINE pcegterg (PW/cegterg.f90)</div><div>at the point </div><div>CALL zsqmred( nbase, vl, desc_old( nlax_ ), desc_old, nbase+notcnv, hl, nx, desc )</div><div>(this means that the calculation of the 804th k-point executes the line immediately before, but not the one immediately after)</div><div>Of course, all this provided I did my debug correctly.</div><div><br></div><div>iv) using QE 4.1.2 the code runs 25% slower and stops exactly at the same k-point but in this case it doesn't keep running</div><div>(the code stops with a segmentation fault error)</div><div><br></div><div><br></div><div>Another issue I would like to point out is that, in the cases when the calculation finishes correctly, the one with "-ndiag 1" runs much</div><div>faster (not sure but maybe half the time), so it could be better to set the code in such a way that, in these "not-expensive" cases the</div><div>parallel diagonalization is disabled by default.</div><div><br></div><div>Giovanni</div></div><div><div><br class="webkit-block-placeholder"></div><div><br class="webkit-block-placeholder"></div><div>
<div><br class="Apple-interchange-newline">PS these are my input files:</div><div><br></div><div>>>>>>>>>>>>>>>>> Si.scf.in</div><div><div>&CONTROL</div><div> calculation = 'scf'</div><div> title = 'Si'</div><div> restart_mode = 'from_scratch'</div><div> outdir = '/scratch/cantele/prova'</div><div> prefix = 'Si'</div><div> pseudo_dir = '/home/nm_settings/software/CODES/Quantum-ESPRESSO/pseudo'</div><div> wf_collect = .true.</div><div> verbosity = 'high'</div><div>/</div><div>&SYSTEM</div><div> ibrav = 2</div><div> celldm(1) = 10.20927</div><div> nat = 2</div><div> ntyp = 1</div><div> ecutwfc = 30.0</div><div>/</div><div>&ELECTRONS</div><div> conv_thr = 1.0d-8</div><div> mixing_beta = 0.7</div><div>/</div><div>ATOMIC_SPECIES</div><div> Si 28.0855 Si.pz-vbc.UPF</div><div>ATOMIC_POSITIONS { alat }</div><div> Si 0.00 0.00 0.00</div><div> Si 0.25 0.25 0.25</div><div>K_POINTS { automatic }</div><div>8 8 8 0 0 0</div><div><br></div><div><br></div><div><br></div><div>>>>>>>>>>>>>>>>> Si.nscf.in</div><div><div>&CONTROL</div><div> calculation = 'nscf'</div><div> title = 'Si'</div><div> restart_mode = 'from_scratch'</div><div> outdir = '/scratch/cantele/prova'</div><div> prefix = 'Si'</div><div> pseudo_dir = '/home/nm_settings/software/CODES/Quantum-ESPRESSO/pseudo'</div><div> wf_collect = .true.</div><div> verbosity = 'high'</div><div>/</div><div>&SYSTEM</div><div> ibrav = 2</div><div> celldm(1) = 10.20927</div><div> nat = 2</div><div> ntyp = 1</div><div> ecutwfc = 30.0</div><div> nbnd = 60</div><div>/</div><div>&ELECTRONS</div><div> diago_full_acc = .true.</div><div> diago_thr_init = 1.0d-6</div><div>/</div><div>ATOMIC_SPECIES</div><div> Si 28.0855 Si.pz-vbc.UPF</div><div>ATOMIC_POSITIONS { alat }</div><div> Si 0.00 0.00 0.00</div><div> Si 0.25 0.25 0.25</div><div>K_POINTS { automatic }</div><div>24 24 24 1 1 1</div></div></div><div><br></div><div>--</div><div><br></div><div>Dr. Giovanni Cantele<br>Coherentia CNR-INFM and Dipartimento di Scienze Fisiche<br>Universita' di Napoli "Federico II"<br>Complesso Universitario di Monte S. Angelo - Ed. 6<br>Via Cintia, I-80126, Napoli, Italy<br>Phone: +39 081 676910<br>Fax: +39 081 676346<br>E-mail: <a href="mailto:giovanni.cantele@cnr.it">giovanni.cantele@cnr.it</a><br> <a href="mailto:giovanni.cantele@na.infn.it">giovanni.cantele@na.infn.it</a><br>Web: <a href="http://people.na.infn.it/~cantele">http://people.na.infn.it/~cantele</a><br>Research Group: <a href="http://www.nanomat.unina.it/">http://www.nanomat.unina.it</a><br>Skype contact: giocan74</div>
</div>
<br></div></body></html>