[DFTB-Plus-User] DFTB+ 21.2 + nodes

[Bálint Aradi]

Dear Toufik,

Scalability depends very strongly on the interconnect between the nodes 
and the *size of the system* you treat. As DFTB uses rather small 
matrices, you might need several thousands of atoms to scale between 
nodes properly. You may parallelise over spin/k-points (if you have any) 
building groups, that can improve scaling a lot.

Best regards,

Bálint

On 05.01.22 11:01, toufik esssakhri wrote:
> Dear all,
> 
> I am using DFTB+ 21.2 for my calculations. However, I notice that for 
> more nodes more running time, which is somehow "strange" for me; ie
>   1 node     4mn 34sec
>   3 nodes   5mn 16sec
>   6 nodes  12mn 15sec
> 10 nodes 29mn 26sec
> 12 nodes 36mn 44sec
> Compilation with MPI=1 
> https://dftbplus-recipes.readthedocs.io/en/latest/parallel/compiling.html
> 
> I know it is maybe more related to the architecture of our 
> supercomputing system but I didn't notice such behavior for other DFT 
> codes I am using.
> 
> Best regards,
> Toufik
> 
> _______________________________________________
> 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
> 


-- 
Dr. Bálint Aradi
Bremen Center for Computational Materials Science, University of Bremen
http://www.bccms.uni-bremen.de/cms/people/b-aradi/


-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature
Type: application/pgp-signature
Size: 840 bytes
Desc: OpenPGP digital signature
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20220105/a34d8e09/attachment.sig>