[CP2K-user] [CP2K:18682] DFTB calculations slower in development version cp2k

[Roel Van de Ven]

Dear all,

I was running some tests with the latest cp2k version as I wanted to try 
out the new implementations for the dispersion within DFTB. During these 
test I found out that the DFTB calculations in general were slower when 
compared to the 9.1 version I typically use. 
Some more testing showed that this was not a general issue, since regular 
DFT (BLYP) and the semi-empirical PM6 method had no slowdown. 
I did this testing on the same system and installed cp2k with the same flags
./install_cp2k_toolchain.sh --with-sirius=no --with-spla=no --with-elpa=no 
--with-plumed=install

Additionally the input files are exactly the same. In the last tests I 
added a line to change images to infinite as the default setting was 
changed between the versions, but this had no impact as far as I know.
&DBCSR
     AVG_ELEMENTS_IMAGES -1
&END DBCSR

I looked a bit into the outputfile and the timings it gives and it seems 
like the time loss is due to calls to rebuild_ks_matrix (and subsequently 
to build_dftb_ks_matrix), but I have no idea what I can do about this. Does 
anyone know what I can do? Or what is wrong?

here is a link to google drive with all the files I used to do the 
calculations https://drive.google.com/drive/folders/1rUq-CwZygGr1GYs9pdMvf98r9YRa0Xjg?usp=sharing 
(both input and output)

Kind regards,
Roel v/d Ven

-- 
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/61a4ae93-20f5-4e69-940a-36aa4390ced8n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20230419/96150e38/attachment.htm>