I am trying to do MC simulations with the Martini ff, which involves electrostatic calculations. I tried towhee and Cassandra. But both are too slow and I have to consider other software. I wonder if EMC is superior to them in terms of computational efficiency.
Dear Xian,
Calculation speed is directly coupled to how energy is evaluated. EMC uses a cell list, which discretizes the simulation domain in order to speed up neighbor searches. Additionally, EMC can use a Verlet-like list, which gives an extra speed up. Using MARTINI with electrostatics typically means, that only part of your beads is charged. A speed up is accomplished by only considering the charged beads when applying Ewald summations as the latter uses imaginary vectors of a user-defined length. Electrostatics can also be sped up by using a cutoff rather Ewald summations, assuming the user is willing to deal with the trade of of applying cutoffs. Additionally, multi threading can be applied to decrease calculation wall time. EMC has multi threading on board, but application of such does not result in a large gain in speed. This lack in speed up gain for the latter stems from the limited amount work of a per bead energy calculation, since these are already highly optimized.
To my knowledge, Towhee does not use cell lists, nor does it treat electrostatics in any efficient way. I am not familiar with the inner workings of Cassandra. It needs to be mentioned, that both codes support configurational bias (CB) approaches. EMC does not have these algorithms incorporated due the scope for with EMC is currently developed, i.e. a Monte Carlo code for building input structures, for which EMC has been optimized. However, this does not mean, that adding CB routines would be a huge effort.