Dear CHARMM experts,
For some reasons, I have to write my own code using the CHARMM 22 force-field. I have performed many kinds of tests and all the results from my code are exactly the same from CHARMM package. However, I found that CHARMM package is about 10 times faster than my code in a large test system calculation. In my code, I have also used neighbor list and exclude list methods. I am not sure why there is a big difference in the time consumed. I have used the parallel version of CHARMM package on my Sun cluster with the cpu specified to be 1. In this way, I thoght the time would be comparable with my serial code. However, it's not. By the way, the number of atoms for the large test system is about 100,000. I used the f90 compiler on Sun cluster with the -fast flag to compile my code.
Any ideas about this will be greatly appreciated,