[Pw_forum] Parallelism

Pascal Boulet pascal.boulet at univ-amu.fr
Thu Oct 5 17:31:22 CEST 2017

Dear all,

We have a rather big system (a=b=c= 20 angs) of 646 atoms, 2160 states. We are running it at the Gamma point with PBE functional (no hybrid). 
Here are our settings:
We use 6*12=72 processors.
Task groups = 1
We set nr1s, nr2s and nr3s to 216, so that 72 is a divisor of 216, as suggest in the manual (we are using USPP, ecut=48 Ry, ecutRho=10*ecut).
FFTs are performed on a 6*6 grid (option ndiag=36).

The job runs for 1 hour (I set this limit for the test in the queuing system) but QE cannot perform even the 1st step of the SCF cycle (last lines of output = iteration #  1     ecut=    48.00 Ry     beta=0.20
     Davidson diagonalization with overlap).

Are the above settings wrong? Or should we just use more processors? 
Could the openmp option be useful in this case?

Thank you for your hints,

Pascal Boulet - Professor - DEPARTEMENT OF CHEMISTRY
Aix-Marseille University - ST JEROME - Avenue Escadrille Normandie Niemen - F-13013 Marseille - FRANCE
Tél: +33(0)4 13 55 18 10 - Fax : +33(0)4 13 55 18 50
Site : http://madirel.univ-amu.fr/pages_web_BOULET_PASCAL/infos - Email : pascal.boulet at univ-amu.fr

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20171005/aa95b524/attachment.html>

More information about the users mailing list