[DFTB-Plus-User] Release 20.2

[Carlo Nervi]

Dear Bálint, and all dftb+ developers,
thank you very much for the efforts! I really appreciate it and the new
stable release 20.2 with many news.
I'd like to contribute for what concerns the solid state systems, periodic
calculations, and in particular the MBD and TS treatment of weak
interaction, pulling out a problem that I was mentioning in the recent
"Calculation of Kpoints" thread. The problem arises whenever a periodic
cell containing several atoms is computed for geometry optimization.
I started with the development version of dftb+, and I also tested the last
20.2 stable release. Up to now I found the following problems:

First of all I'm working with relatively large periodic systems. 128 atoms
per unit cell.
*D4* works apparently smoothly (setting s9=0), with relative fast
calculations, as expected.
*MBD *is very slow, but *TS *is extremely slow.
The average time requested to perform a single geometry step in D4 is
around 0.55 s, in MBD is 9.4 s, but in TS is 732 s.
This makes no sense to me. TS should cost approximately as D4, as far as I
know.

Are there any library problems? I'm using openblas 0.3.12, openmpi and I
set OMP_NUM_THREADS=1 to avoid OpenMP. I tried also to compile without
OpenMP, but I got the same results.

Other programs run fine in parallel (Quantum Espresso for instance for
openmpi and Gaussian16 for OpenMP).
Thank you for your time,
Carlo

-- 

------------------------------------------------------------
Prof. Carlo Nervi carlo.nervi at unito.it  Tel:+39 0116707507/8
Fax: +39 0116707855      -      Dipartimento di Chimica, via
P. Giuria 7, 10125 Torino, Italy.    http://lem.ch.unito.it/

*ICCC2020 has been postponed at 2022*

ICCC 2022 28 August - 2 September 2022, Rimini, Italy: http://www.iccc2020.com
International Conference on Coordination Chemistry (ICCC 2022)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20201119/87a2741d/attachment.htm>