On Tue, Jan 3, 2012 at 12:12 PM, David Lonie <lonieda...@gmail.com> wrote: > I'm trying to track down a bug somewhere in the forcefields that is > causing this to happen: > > C O N J U G A T E G R A D I E N T S > > STEPS = 250 > > STEP n E(n) E(n-1) > -------------------------------- > 1 757722.040 5890899.081 > 2 179656.470 179656.470 > CONJUGATE GRADIENTS HAS CONVERGED > > The geometry is not reasonable, though it claims to be converged. Also > for step n=2, E(n-1) != E(1).
For the ML archive in case anyone else runs into this problem: I've traced the problem down to the line search. The (default) Simple line-search method seems to be unable to find a lower energy configuration in the direction of the calculated gradient. It's probably skipping over the minimum, or the function is increasing beyond the trust radius. I didn't track down which is happening, but switching to the Newton2Num line-search method fixes the problem for me (see OBForceField::SetLineSearchType()). I've fixed the other memcpy bug and made some trivial optimizations to the Simple LineSearch method in commit r4696. Dave ------------------------------------------------------------------------------ Write once. Port to many. Get the SDK and tools to simplify cross-platform app development. Create new or port existing apps to sell to consumers worldwide. Explore the Intel AppUpSM program developer opportunity. appdeveloper.intel.com/join http://p.sf.net/sfu/intel-appdev _______________________________________________ OpenBabel-Devel mailing list OpenBabel-Devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/openbabel-devel
