-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hello David-- > I've noticed that in the anneal loop as ensemble size is increased by > about an order of magnitude, the delta_t decreases by an order of > magnitude and the number of steps at a given temperature increases by > an order of magnitude. This is greatly increasing my runtime for a > fixed total number of structures even though I am also increasing my > number of threads. The number of steps is increasing *because* the stepsize is decreasing. > I am not changing any annealing parameters to account for increase in > ensemble size. Should I, and if so, which parameters should I change > and how? You might be able to decrease final time a bit. However, configuration space is an order of magnitude larger, so (effectively) longer dynamics runs are warranted. We currently keep endT constant for different sized ensembles. Note that you should not make the number of threads larger than the number of processors on a node. The bottom line: large ensembles can be very expensive to compute. regards-- Charles -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.3 (GNU/Linux) Comment: Processed by Mailcrypt 3.5.8 <http://mailcrypt.sourceforge.net/> iD8DBQFE0Lx6PK2zrJwS/lYRAqUqAJ90S+PgjhhcZCCdze2Audf0cu5/4gCdHlwQ OnwJgfTe+QdRxXnn2iEaqZ0= =xj6Q -----END PGP SIGNATURE-----
