[gmx-developers] dihedral restraints and domain decomposition

[David Mobley]

Dear Devels,

We've been running some of our usual absolute binding free energy
calculations using distance, dihedral, and angle restraints, this time on a
larger system than normal (100k+ atoms).

What we've found in this situation is that we get crashes reliably due to
the dihedral restraints when running with domain decomposition. The errors
are typically not helpful (LINCS warnings preceding a crash, sporadic
domain decomposition errors, etc.). However, we've established that these
problems apparently go away using particle decomposition.

So, this is to ask: Are there expected issues with dihedral restraints and
domain decomposition? (We apply restraints between our ligand and reference
atoms in the protein). If what we're doing is not advised, it would be nice
to know.

Thanks!
David


-- 
David Mobley
dmobley at gmail.com
949-385-2436
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20131209/05d34c36/attachment.html>