[DFTB-Plus-User] large-scale calculations_output overflow records and "cannot allocate memory error message"

khalfoun hafid h_khalfoun at yahoo.com
Fri Jun 5 14:57:48 CEST 2020


An other technical question, please , 

Is this builder of DFTB+ ( without ELSI sovers) as perform as which uses the ELSI ?
easybuilders/easybuild-easyconfigs

| 
| 
| 
|  |  |

 |

 |
| 
|  | 
easybuilders/easybuild-easyconfigs

A collection of easyconfig files that describe which software to build using which build options with EasyBuild....
 |

 |

 |


This is what the technical administrator used for DFTB+. 

 Thanks

    Le vendredi 5 juin 2020 à 13:48:53 UTC+2, Ben Hourahine <benjamin.hourahine at strath.ac.uk> a écrit :  
 
  
Hello Hafid,
 
for dense eigensolvers and an sp basis, the memory cost for the linear algebra is approximately 2 - 4 times the square of
 
 
[ nAtoms * 4 basis * ~4 bytes if a real calculation (or 8 * number of k-points for complex) ]
 
 
which for 1000 atoms is approaching 1Gb, while for your 32000 atoms would be  about 800 Gb. If you have a distributed computer with for example ~4 Gb per core that would require around 200 processors to store the problem.
 

 
 
Instead, if your calculation is suitable, the ELSI sparse solvers have much lower memory requirements (being ~ linear instead of proportional to the square of the number of atoms). You could try either the NTPoly, OMM or PEXSI solvers but with the Sparse = Yes option enabled. They each have different pre-factors for memory requirement depending on their settings. (OMM is intended for insulators, NTPoly will work for smaller gaps, and PEXSI runs for everything but has a larger time prefactor for 3D systems). See the preprint 
 
 
https://arxiv.org/abs/1912.13403
 
for a discussion of some of this.
 
 
Regards
 
Ben
 
 On 05/06/2020 07:36, khalfoun hafid via DFTB-Plus-User wrote:
  
   Dear Ben, 
  I try to calculate the electronic band structure for a large system ( 32000 atoms of  Carbon) . 
  
  Unfortunately I meet the output overflows record similarly as already been discussed 
  https://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/2018/002794.html 
  I've desactivated the writing of the dftb_pin.hsd file in the parser options ( as recommanded  in https://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/2018/002795.html ) 
  , but the problem of the non ability allocation memory still persist even I reach the limit of the required memory in the calculating machine. http://www.ceci-hpc.be/clusters.html#lemaitre3 
  what should I do  ? Using another machine with greater amount of RAM ?
  
    
  
  thanks , Hafid
  
      
  _______________________________________________
DFTB-Plus-User mailing list
DFTB-Plus-User at mailman.zfn.uni-bremen.de
https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fmailman.zfn.uni-bremen.de%2Fcgi-bin%2Fmailman%2Flistinfo%2Fdftb-plus-user&data=02%7C01%7Cbenjamin.hourahine%40strath.ac.uk%7C8f3c2e12d858443d588508d8091aec43%7C631e0763153347eba5cd0457bee5944e%7C0%7C0%7C637269358994468825&sdata=x9K%2B1%2Fn78aVGnNNg0rYVsUVS1Ia%2BBYQ1CuZ9EOe17Qc%3D&reserved=0
 
 -- 
        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

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 _______________________________________________
DFTB-Plus-User mailing list
DFTB-Plus-User at mailman.zfn.uni-bremen.de
https://mailman.zfn.uni-bremen.de/cgi-bin/mailman/listinfo/dftb-plus-user
  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20200605/db63ea2d/attachment-0001.htm>


More information about the DFTB-Plus-User mailing list