Re: peek_ewald_inpcrd

From: darden <darden_at_gamera.niehs.nih.gov>
Date: Mon 7 May 2001 16:01:08 -0500 (CDT)

Actually sander in amber6 will run ok on really big systems. I've run
200000+ atom md simulations using pme. The problems
I've seen are all to do with fortran i/o formatting. The first difficulty
happens when you pass 100000. Unfortunately several reading and writing
subroutines use I5 for the statement where number of atoms is read or
written. The problem you saw is due to peek_ewald_inpcrd using the same
format. So rdparm, which reads the prmtop file uses I6 and gets the number
of atoms right, but many other places including peek_ewald_inpcrd misread
it. The symptom you saw was due to this. To fix it you need to change this
format several places (getcrd, peek_ewald_inpcrd, minrit, mdwrit, rdrest).
For example in peek_ewald_inpcrd, where it says
      read(9,9018) natom,tt
...
...
9018 format(i5,5e15.7)

change the i5 to i6. similar changes in the other subroutines I mentioned.
I think this is all you need to do.
Note that the changed subroutines are not compatible with coordinate files
produced by the standard sander release, and you will get errors in
regular sander if you try to read coordinate files produced by your new
big sander. So, best to keep copies of both types of sander. Hopefully
this will all be fixed in the next sander release
when we go to a more general i/o scheme.
By the way the next problem I'm aware of is when the number of solute
atoms gets to around 30000. Then the dihedral list gets too big. A harder
problem to overcome, but it can be done.
Hope this helps
Tom Darden

On Mon, 7 May 2001, Andrew Aird wrote:

> Dear amber users
>
> I'm trying to minimize a very large system: # atoms about 100000!
>
> When I start sander, it stops after a short while. I get the following error
> message from the output file:
>
>
> ----------------------------------------------
> &cntrl
>
>
>
> 1. RESOURCE USE:
>
> getting box info from bottom of parm
> getting new box info from bottom of inpcrd
> peek_ewald_inpcrd: SHOULD NOT BE HERE
>
>
> I've solvated the system with xleap using the solvatebox command (WATBOX216).
> The system looks OK. After producing the *.prmtop *.prmcrd file I've converted
> them back to pdb using ptraj to look at the structure with vmd and again, it
> looks OK.
>
>
> Is there a maximum number of atoms with Amber6? And what does peek_ewald_inpcrd
> mean? Did someone encounter a similar problem?
>
> Thanks for any comments.
>
>
> Andrew
>
>
>
>
> ---------------------------------------------------
> Andrew Aird
> 3.Physikalisches Institut
> Universitšt Stuttgart
> Pfaffenwaldring 57
> D 70569 Stuttgart
> Tel: +49 711 685-5232
> e-mail: a.aird_at_physik.uni-stuttgart.de
> ---------------------------------------------------
>
Received on Mon May 07 2001 - 14:01:08 PDT
Custom Search