I need to clarify this situation after Huiqun's response: "Siesta+PHON: 520 sec at 8CPU" was for two displacement calculations in total. If you have 16 cpu, you can run 2 displacement calculation at the same time. This reduces 520 to 260.
After this post, you and me learned that it is also possible to reduce 328 with q_point splitting..... You can examine my files at http://unam.bilkent.edu.tr/mt2/random/pw_forum/ On Mon, Aug 16, 2010 at 11:45 AM, Huiqun Zhou <hqzhou at nju.edu.cn> wrote: > Mehmet, > > Hope you are the same person who posted a very good tutorial on phonon > calculation via > small displacement method (SDM) and DFPT at Siesta forum. > > Interestingly, I noticed that in your tutorial DFPT (PW+PH: 328 sec at 8CPU) > was much faster > than SDM (Siesta+PHON: 520 sec at 8CPU, VASP+PHON: 1674 sec at 16CPU) for the > case > of graphene, at least by the criteria of getting identical phonon > dispersion relation. So, your > current recommendation means that was just a individual case for you, > right? > > My feel (no concret numbers) with qe shows that for the previous version > of qe (before 4.0 > or earllier?) PW+PH was somewhat slower than PW+PHON, but the situation is > reversed > with recent versions of qe. > > If you have further examples, please share with us. > > Thank you in advance! > > huiqun zhou > @earth sciences, nanjing university, china > > > ----- Original Message ----- > *From:* Mehmet Topsakal <metokal at gmail.com> > *To:* PWSCF Forum <pw_forum at pwscf.org> > *Sent:* Sunday, August 15, 2010 3:52 PM > *Subject:* Re: [Pw_forum] Speeding up the calculations > > Dear Bipul, > > You can try Small Displacement Method ( see > http://www.homepages.ucl.ac.uk/~ucfbdxa/ ) which has different idea than > DFPT. > It requires you to make 2 displacement in a supercell and calculate the > forces with single scf. You can run scf's at the same time. > For example, you need to calculate the forces in 5x5x1 graphene supercell > having 50 carbon atoms for 2 displacemet. > > BUT you cannot observe LO-TO splitting in polar materials (such as > honeycomb monolayer ZnO) by using SDM. > > Regards. > > On Sat, Aug 14, 2010 at 9:44 PM, Bipul Rakshit <bipulrr at gmail.com> wrote: > >> Hi PWSCF user, >> I am mainly doing Phonon calculation, which generally is time consuming ( >> even in parallel). I just want to know if there is any possible way to speed >> up the calculations. i.e. by installing certain libraries or so. Or using >> some appropirate methods >> >> The machine had many nodes, and each nodes have 8/16 processors. Some of >> the other details of the machines are as follows: >> >> Intel(R) Xeon(R) CPU E5345 @ 2.33GHz, 64 bit >> 4GB RAM >> >> compiler: mpif90 and icc >> >> Thanks >> -- >> Bipul Rakshit >> Research Fellow >> S N Bose Centre for Basic Sciences, >> Salt Lake, >> Kolkata 700 098 >> India >> >> _______________________________________________ >> Pw_forum mailing list >> Pw_forum at pwscf.org >> http://www.democritos.it/mailman/listinfo/pw_forum >> >> > > > -- > > Mehmet Topsakal (Ph.D. Student) > UNAM-Institute of Materials Science and Nanotechnology. > Bilkent University. 06800 Bilkent, Ankara/T?rkiye > Tel: 0090 312 290 3527 ; Fax: 0090 312 266 4365 > UNAM-web : www.nano.org.tr > > ------------------------------ > > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum > > > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum > > -- Mehmet Topsakal (Ph.D. Student) UNAM-Institute of Materials Science and Nanotechnology. Bilkent University. 06800 Bilkent, Ankara/T?rkiye Tel: 0090 312 290 3527 ; Fax: 0090 312 266 4365 UNAM-web : www.nano.org.tr -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20100816/50189984/attachment-0001.htm
