|
Adaptivity fixes, MD by VerletForceIntegration introduced, MD molecule::Trajectories, atom Max::Order, no more recursive going down the fragmentation level
MD
==
molecule::Trajectories is now a map to a struct trajectory list of all the MD steps.
struct Trajectory contains STL vectors of coordinates, velocities and forces. Both are needed for the new VerletForceIntegration.
Parsing of coordinates, velocities and forces from the config file was completely rewritten:
- in the FastParsing case, we just scan for IonType1_1 to find the last step, set the file pointer there and scan all remaining ones
- in the other case, we create the atoms in the first step and put them in a hash for lookup on later steps and read in sequentially (with file pointer moving on).
- This is a lot faster than the old variant.
VerletForceIntegration() implemented in a working manner with force smoothing (mean of actual and last one).
OutputTrajectoriesXYZ() now concatenates the single MD steps into one xyz file, so that the animation can be viewed with e.g. jmol or vmd
config:Deltat is now public (lazy me) and set to 1 instead of 0 initially, also Deltat is parsed accordingly (if not present, defaults to 1)
MatrixContainer::ParseMatrix from parser.cpp is used during VerletForceIntegration() to parse the forces file. Consequently, we have included parser.* in the Makefile.am.
Fix: MoleculeListClass::OutputConfigForListOfFragments() stores config file under config::configpath, before it backup'd the path twice to PathBackup.
Adaptivity
==========
Adaptivity (CheckOrderAtSite()) had compared not absolute, but real values which caused sign problems and faulty behaviour.
Adapatvity (CheckOrderAtSite()) had included atoms by Order (desired one) not by FragOrder (current one) into the list of candidates, which caused faulty behaviour.
CheckOrderAtSite() did single stepping wrong as the mask bit (last in AtomMask) was checked for true not false! Also bit was not set to false initially in FragmentMolecule().
Adaptivity: FragmentMolecule now returns 1 - continue, 2 - don't ... to tell whether we still have to continue with the adaptive cycle (is given as return value from molecuilder)
introduced atom::MaxOrder
StoreOrderAtSiteFile() now also stores the MaxOrder and ParseOrderAtSiteFromFile() parses it back into atom class
Removed Fragmentation Recursion
===============================
As we switched in the prelude of the adaptivity to a monotonous increase from order 1 to the desired one, we don't need to recursively go down each level from a given current bond order, as all these fragments have already been created on the lower orders. Consequently, FragmentBOSSANOVA does not contain NumLevels or alike anymore. This simplified the code a bit, but probably is not yet completely done. (but is working, checked on heptan).
SPFragmentGenerator() does not need Labels anymore, global ones are used for checks. Consequently, PowerSetGenerator() and FragmentSearch structure don't initialise/contain them anymore. We always compare against ...->GetTrueFather()->nr.
|
