Dave -
I think I based this on some discussion on the either the dev or the users
list; I don't remember if you were involved, but it was some of the guys I
considered more experienced in the practical aspects; I mostly do code. So
it is folklore without analysis, but it was amber folklore (I think the
statement was something to the effect that it is good to have 0.25 psec
snapshots). Sorry if I have been propagating something that is not quite
right based on faith. Anybody else want to weigh in on this issue? It is
actually really impt in terms of the cost of simulation and how far you can
scale (bigger traj are a pain, and have a real cost at high scaling, and as
the systems get bigger the cost goes up).
Regards - Bob
----- Original Message -----
From: "David A. Case" <case.scripps.edu>
To: <amber-developers.scripps.edu>
Sent: Thursday, May 04, 2006 6:39 PM
Subject: Re: amber-developers: Troubles at PSC
> On Thu, May 04, 2006, Robert Duke wrote:
>
>> I dump trajectory every 250 steps
>
> Is this a real requirement, i.e. do you need sub-picosecond resolution in
> dynamics? The structures at this time resolution will be highly
> correlated
> with each other, so there is in effect redundant information in the
> snapshots.
>
> I haven't saved snapshots at less than 1 ps resolution (usually 2ps or
> longer)
> for a long time. In the cases where I need more fine grained information,
> I
> usually run a short, "fine-grained" simulation (with a small ntwx) to get
> the
> needed data.
>
> It should be clear that I may be way off base here, since I don't know the
> real requirements. But you might want to see if you could refine the
> requirements to require less frequent disk access.
>
> ....regards...dave
>
>
Received on Fri May 05 2006 - 08:13:57 PDT
