[DFTB-Plus-User] Problems with DFTB Stable version 19.1

[Zhu, Junmian]

Dear Dr. Hourahine,

Thank you very much for your swift response! I tried to add the groups and mpirun with 8 processors, so now the DFTB+ is working.

Best regards,

Junmian Zhu
University of California, Berkeley


________________________________
From: DFTB-Plus-User <dftb-plus-user-bounces at mailman.zfn.uni-bremen.de> on behalf of Ben Hourahine <benjamin.hourahine at strath.ac.uk>
Sent: Friday, July 5, 2019 4:58:25 PM
To: dftb-plus-user at mailman.zfn.uni-bremen.de
Subject: Re: [DFTB-Plus-User] Problems with DFTB Stable version 19.1


Hello Junmian,

9 processes for such a small structure is probably spin-locking, as it will try and spread the diagonalisation across the 9 processors and work through the k-points 1 at a time. Hence any network lag or congestion will hit the performance of this very strongly.

What happens if you add this extra block to the input:

Parallel {
  Groups = 8
}


and run with

mpirun -np 8 dftb+

?

This setting will run batches of 8 k-points at a time (you have 64 in total).

Regards

Ben

On 05/07/2019 22:44, Zhu, Junmian wrote:
Dear DFTB+ Community,

I have experienced problems with MPI-enabled DFTB+ geometric optimization calculation (version: 19.1 stable release), and am able to reproduce the problem. When running the command “mpirun -np 9 dftb+” with the given input file, DFTB+ always stops at geometric step 59. Even though the DFTB+ process is still running at the backend as shown by “top” command, it stops streaming any output and got stuck there for more than 12 hours.

The behavior of DFTB+ is quite strange, because DFTB+ is able to continue to run and finish the calculation if the process is killed and restarted again, using the previous output geometry and charges. Multiple tests on the same computer give exactly the same output and get stuck exactly at step 59. (Changing number of processors only changes the step it got stuck, and OpenMP is disabled.)Therefore, I suspect this is an issue with the DFTB+ software and would like to request help on solving this problem.

Thank you very much,

Junmian Zhu
University of California, Berkeley





_______________________________________________
DFTB-Plus-User mailing list
DFTB-Plus-User at mailman.zfn.uni-bremen.de<mailto:DFTB-Plus-User at mailman.zfn.uni-bremen.de>
https://mailman.zfn.uni-bremen.de/cgi-bin/mailman/listinfo/dftb-plus-user<https://urldefense.proofpoint.com/v2/url?u=https-3A__mailman.zfn.uni-2Dbremen.de_cgi-2Dbin_mailman_listinfo_dftb-2Dplus-2Duser&d=DwMD-g&c=HUrdOLg_tCr0UMeDjWLBOM9lLDRpsndbROGxEKQRFzk&r=g6j0G2C9UjOEKslTiIDTLyzGhT7AORWuR5pp0iWgJBU&m=ivRuot4MmjdFy1Vmz4ErkeXW_myOPxOBs8CXT5MCk6U&s=LR5-vY0rIhXWC24gxilamEF81ED5shwLyITBftj4cF0&e=>

--
        Dr.  B.  Hourahine,  Senior  lecturer
            SUPA, Department  of  Physics,
             University  of  Strathclyde,
              John  Anderson  Building,
          107 Rottenrow, Glasgow G4 0NG, UK.
  +44 141 548 2325, benjamin.hourahine at strath.ac.uk<mailto:benjamin.hourahine at strath.ac.uk>

The  Department is  a partner  in SUPA,  the Scottish
            Universities Physics Alliance

The University  of Strathclyde  is a  charitable body,
       registered in Scotland, number SC015263
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20190706/89cbf458/attachment-0001.html>