Hi,
> Just as a rough estimate, how much time of a computation does each of
> these
> terms take? Which is the most costly/cheap?
To give a one sentence answer, the non-bonded terms (Coulomb&LJ) take up
by far (>90%) the largest part of a (non-polarizable, non QM/MM)
calculation. These further split into direct sum (LJ and Coulomb up to
cutoff) and reciprocal (the Fourier space part of PME), of which direct
sum takes very roughly about 60%. In extreme cases, depending on the write
frequency and number of MPI threads, your job speed may depend more on
interconnect and disk speed than on actually doing the claculations.
But note that any amber calculation will contain precisely this
information at the end of mdout, so you can check for you own system and
parameter settings...
Kind Regards,
Dr. Thomas Steinbrecher
formerly at the
BioMaps Institute
Rutgers University
610 Taylor Rd.
Piscataway, NJ 08854
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Sep 16 2011 - 16:30:02 PDT
