[DFTB-Plus-User] Number of external charges in DFTB+
balint.aradi at bccms.uni-bremen.de
Mon Feb 16 02:43:12 CET 2009
> I have 20271 external charges.
> The old DFTB code needs:
> real 0m3.630s
> user 0m2.332s
> sys 0m0.160s
> DFTB+ needs:
> real 9m26.533s
> user 9m10.518s
> sys 0m17.637s
Huh, that's more than horrifying! Would you mind to send me your test
system (maybe directly, not over the list), so that I can make my own
run on it? (By the way, did you use one of the binaries provided on the
web page, or did you compile your own?)
I tried on the weekend an H2O molecule in a field of 100000 point
charges (5x more than you used), and on my old Centrino laptop it took 7
seconds but by no means several minutes.
However, also those 7 seconds are by far too much, so I wrote a small
extension to the current code, which enables to read in the data
directly from a file, without piping it through the HSD/XML parser.
Using that, the entire parsing business took less than 0.3 seconds,
which is acceptable, I guess.
Once I have your system, I can test it with the new implementation. If
the results are satisfactory, I'll put the patched version of the code
to the website, so that everybody can download it.
Dr. Bálint Aradi
Bremen Center for Computational Materials Science, University of Bremen
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 260 bytes
Desc: OpenPGP digital signature
Url : http://www.dftb-plus.info/pipermail/dftb-plus-user/attachments/20090216/ecaf30f0/attachment-0001.pgp
More information about the DFTB-Plus-User