Date: Fri, 12 Jul 2002 08:55:51 -0700
From: "David A. Case"
Subject: Re: Question about error message about "lastist" and "virtual box"

On Fri, Jul 12, 2002, Masaki Tomimoto wrote:
>
> Even if two systems were quite similar, one got the following error
> messages in MD simulation, but another did not.
>
> (1)
> The system has extended beyond
> the extent of the virtual box.
> Restarting sander will recalculate
> a new virtual box with 30 Angstroms
> extra on each side, if there is a
> restart file for this configuration.

Even for non-periodic simulations, the code creates a "virtual" bounding
box, used for constructing the non-bonded list. We have struggled over
how big to make this virtual box: if it is too big, the list construction
is slowed down; if it is too small, a conformational change could make
the system leave the box.

We currently allow 30 Ang. around the initial position of the solute, but
this can clearly not be enough in some cases. In two similar calculations,
one could go outside the "virtual box" and the other not.

If you need more "space" to allow you system to change shape, change the
"extraboxdim" variable in nonper.h to a larger value. But you might also
want to think if something is "wrong", e.g. if you did not expect such a
large conformational change.

>
> Exceeding lastist in get_istack
> lastist = 3500000
> top_stk= 2134860
> isize = 2052750
> request= 4187610
> Increase lastist in the &cntrl namelist
>
> I already enlarge lastist parameter from 2000000 to 3500000, because I got
> the similar error message in the previous run. At that time, a value in
> request was less than 3500000.

The system exits the first time it has a request for stack sizes that
exceed the "lastist" value. Unfortunately, this means that you have to
use some trial and error to figure out the best value, since a later stack
request might want even more space than the earlier one.

..hope this helps....dac