s2.results.timing.

Timing results for the ten systems are shown in Table . The times represent the average of 100 dynamics steps run on an Iris Indigo workstation. Times are given for both the NEIMO calculations and the nonbond calculations, the latter of which consumes the vast majority of cpu time, even when a very fast method such as CMM is used. The NEIMO timing is shown to be rigorously proportional to for Crambin and the large proteins. The times for the small peptides apparently are shorter, but the resolution of the timing routine used is 0.01 s, so the results may be off by as much as 50%. The nonbond calculations are less consistent, even for CMM, which is proportional to , the number of atoms in the system. The lack of perfect proportionality is due to the asymmetry of the protein conformations. The Cell-Multipole algorithm[25] creates a cubic cell around the system being simulated and divides this cell into a hierarchy of smaller cubic cells, for which dipole and quadrupole terms are calculated. An oblong protein molecule would require a particularly large outer cell and would have many of its smaller cells empty. These are the least efficient conditions for CMM. Considering this limitation, the method is very nearly proportional to , and is much faster than the method of calculating all possible nonbond pairs.

Sat Jun 18 14:06:11 PDT 1994