Re: [AMBER-Developers] ntb vs ifbox?

From: Jason Swails <jason.swails.gmail.com>
Date: Sun, 4 Dec 2011 20:29:23 -0500

On Sun, Dec 4, 2011 at 5:54 PM, Yong Duan <duan.ucdavis.edu> wrote:

>
> But, just because there are a few special folks that request such special
> things, we should do it?


If it doesn't overly-complicate things for normal users, I think yes.

What about the great majority of other users? For
> those very few special folks, they have the freedom to edit prmtop.
>

I don't think we should start encouraging people to modify that file by
hand. A hand-edited prmtop file is a recipe for ugly, obscure errors by
even the most experienced prmtop hackers.

Force field stuff, IMO, belongs in the prmtop (IPOL, SCEE/SCNB, etc.). I
don't think PBC falls into this category.

It's so easy to mess up now. If one accidentally sets wrong ntp, you'd
>

ntb or ntp? IFBOX is only enough to tell us, usually, whether ntb == 0 or
ntb /= 0, not whether ntp == 0, ntp == 1, ntp == 2, ntb == 1, or ntb == 2,
etc.

have to restart. I bet a lot of people run into this. I did, numerous
> times. And, I've never come to the occasion of needing ntb=0 with IFBOX=1
>
in my ~20 years of using AMBER. Why one builds a periodic box and runs
> non-PBC? To study surface-tension effect to find out how a box of water is
> gradually squeezed into a sphere? I guess, there are people who do that.
> But, for great majority of our users, I bet when they build a IFBOX=1
> system, they meant to run simulations with ntb > 0 and when they build
> IFBOX=0 system, they meant to use it for ntb=0.
>

I still think the choice of periodic boundary conditions and their
parameters should be deliberate. Furthermore, the default ntb is 1, (NVT).
 If you try to set a GB model with that, you get a helpful error message.
 Likewise for trying to use PME with ntb==0.

What I could see us doing, though, is to make the default value
IFBOX-dependent. For instance, if IFBOX==0, then ntb==0 is the default,
but if IFBOX /= 0, we stick with ntb==1. If a user manually provides an
ntb value, we just use that. Anything more than that is overreaching IMO.

We would also be breaking long-standing behavior, so I think there should
be a very clear gain for such a break with tradition (such as enabling
mixed scaling for GLYCAM+AmberFF for SCEE/SCNB --> prmtop and avoiding the
ridiculous-ness of IPOL=1 for a fixed-charge FF).

Just my 2c (Fighting inflation).

All the best,
Jason

-- 
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Sun Dec 04 2011 - 17:30:02 PST
Custom Search