[QE-users] QE6.8: "Slab decomposition and task groups not implemented"

[Sebastian Hütter]

Hi Paolo,

>      From the source code, I think "-ntg" now requires switching from slab to pencil decomposition using the
>     undocumented switch "-pd"?
> exactly so

Right, thanks.

>     What are the implications of this method selection? Is one "better" or "worse" than the other, whatever that might mean?
> -pd true -ntg N is the same as -ntg N in previous versions. It is convenient only when you parallelize FFTs over a large 
> number of MPI processes. Otherwise, FFTs with -pd true ("pencil decomposition") are typically slower than the default 
> ("slab decomposition").

Do you have an idea where that "large number" is? I've done a few test runs and slab is consistently 10..20% faster, so 
I guess I'm below that. Probably also requires sufficiently large FFT grids to feed enough data to each process?

> The choice will be made automatic sooner or later.

Automatically optimized defaults are always nice :)


Best,

Sebastian

> 
> Paolo
> 
> 
> 
>     Thanks for any help,
> 
>     Sebastian
> 
>     -- 
>     M.Sc. Sebastian Hütter
>     Otto-von-Guericke University Magdeburg
>     Institute of Materials and Joining Technology
>     Universitätsplatz 2
>     D-39106 Magdeburg
>     Tel.: +49 (0)391-67-54566
>     Fax: +49 (0)391-67-44569
>     _______________________________________________
>     Quantum ESPRESSO is supported by MaX (www.max-centre.eu <http://www.max-centre.eu>)
>     users mailing list users at lists.quantum-espresso.org <mailto:users at lists.quantum-espresso.org>
>     https://lists.quantum-espresso.org/mailman/listinfo/users
> 
> 
> 
> -- 
> Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
> Univ. Udine, via delle Scienze 206, 33100 Udine, Italy
> Phone +39-0432-558216, fax +39-0432-558222
> 
> 
> 
> _______________________________________________
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
> users mailing list users at lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users
>