10:26:35 AM PDT - Fri, Jul 15th 2011 |
|
The reason for introducing ncycles was simply to break free energy sampling into several stages.
During each cycle
-- Classical MD trajectory is generated for nsamples
-- The resulting MD trajectory is then processed to generate free energy differences, current values are printed out, necessary information is saved to continue sampling
The next cycle just builds up on the previous one, accumulating all the averages from the previous cycles. In that sense having N cycles with 1000 samples will lead to the same result as one cycle with N*1000 samples. If you system goes down case you can recover/continue simulations using information from N-1 cycle, but if N=1 you will loose everything.
|