|
From: Geoffrey H. <ge...@pi...> - 2012-08-21 21:09:36
|
> I wrote an update to the ConformerSearch class. Sorry for the delay, I've been at a pile of conferences and finally have decent network access (at the ACS meeting). > First I noticed that in the Search method, the convergence test: ... > I also noticed that the whole population was evaluated completly (the scoring method is called > for all rotorkeys), calling many time the energy scoring for the same key. So I added a mapping (C++ std::map) > that provides the energy given the key, if it was already computed. For long molecules this is quite usefull, and > save about half of energy computations, and then total search computation time is well reduced fir the same result. These are some great changes. > So I'd be glad to contribute, and now the question is how. Should I send my changes (modified conformersearch.cpp and conformersearch.h) > to a regular developper for testing, or should I register in the Sourceforge and submit through SVN? For now, we can certainly take your changes, but since it seems like you have more ideas, I'd suggest the latter. Send me your SourceForge username and I can get things set up. Thanks -- if you've followed the list, you have probably seen that people have been more extensively using the GA (great) and finding some issues (not great). So we're very interested in improvements! Thanks very much and welcome! -Geoff --- Prof. Geoffrey Hutchison Department of Chemistry University of Pittsburgh tel: (412) 648-0492 email: ge...@pi... web: http://hutchison.chem.pitt.edu/ |