Dear LIANG Xiongyi,
> Have you run DFPT+U by ph.x? Yes, of course. I am one of the coauthors of this paper: https://journals.aps.org/prb/abstract/10.1103/PhysRevB.101.064305 > How many day it will take? One week? One month? It depends on many factors (size of your system, k-mesh, q-mesh, cutoff, number of cores that you use,...) > It there any strategy to accelerate it? Using NCPP? Or USPP? In order to speed up the DFPT+U calculation you should use more core and parallelize efficiently your calculation (over plane waves, k points, q points, ...) Greetings, Iurii P.S.: My first name is Iurii, not Lurii. Just copy-paste it in order to avoid making mistakes. -- Dr. Iurii TIMROV Postdoctoral Researcher STI - IMX - THEOS and NCCR - MARVEL Swiss Federal Institute of Technology Lausanne (EPFL) CH-1015 Lausanne, Switzerland +41 21 69 34 881 http://people.epfl.ch/265334 ________________________________ From: users <[email protected]> on behalf of LEUNG Clarence <[email protected]> Sent: Friday, October 23, 2020 3:33:02 PM To: Quantum Espresso users Forum Subject: Re: [QE-users] ph.x calculation with DPT+U Dear Lurii, Have you run DFPT+U by ph.x? How many day it will take? One week? One month? It there any strategy to accelerate it? Using NCPP? Or USPP? Thanks! Best regards, LIANG Xiongyi Postdoctoral Fellow City University of Hong Kong From: Timrov Iurii<mailto:[email protected]> Sent: 2020年10月19日 5:33 To: [email protected]<mailto:[email protected]> Subject: Re: [QE-users] ph.x calculation with DPT+U Dear Liang, > It is very slow, it is normal? Yes > My system consists of 80 atoms... For such a large system DFPT+U is expected to be very slow, in particular because the calculation of d2ns_bare matrix (second "bare" derivative of the occupation matrix ns) is computationally very expensive. I suggest to use the frozen-phonon approach (Phonopy). Greetings, Iurii -- Dr. Iurii TIMROV Postdoctoral Researcher STI - IMX - THEOS and NCCR - MARVEL Swiss Federal Institute of Technology Lausanne (EPFL) CH-1015 Lausanne, Switzerland +41 21 69 34 881 http://people.epfl.ch/265334 From: users <[email protected]> on behalf of LEUNG Clarence <[email protected]> Sent: Saturday, October 17, 2020 8:53:18 AM To: [email protected] Subject: [QE-users] ph.x calculation with DPT+U Dear QEusers, Recently, I do ph.x calculation with DPT+U in QE6.6. However, it is very very slow. The output file running for four days as follows: Atomic wfc used for the projector on the Hubbard manifold are NOT orthogonalized Calculating the dnsbare matrix... Calculating the dnsorth_cart matrix... Alpha used in Ewald sum = 2.8000 Calculating the d2ns_bare matrix. It might take a while! k point # 1 out of 2 The specification of my server is two nodes with 64 cores (Xeon e5) and 512G RAMs. My system consists of 80 atoms, and the input files of ph.x si as follows: &INPUTPH outdir='', prefix='', tr2_ph = 1d-12, ldisp = .true. , fildyn ='' , epsil=.false., trans=.true., nq1 = 1 , nq2 = 1 , nq3 = 1 , nat_todo = 1, alpha_mix(1) = 0.07, / 79 It is very slow, it is normal? It is very appreciated that give me some advices. Best regards, LIANG Postdoc City University of Hong Kong
_______________________________________________ Quantum ESPRESSO is supported by MaX (www.max-centre.eu) users mailing list [email protected] https://lists.quantum-espresso.org/mailman/listinfo/users
