AMBER Archive (2002)

Subject: Re: help! Non bond list overflow!

From: Jian Zhang (jzhang_001_at_yahoo.com.cn)
Date: Wed Nov 13 2002 - 20:32:01 CST

Dear David, A. Case,

The problem only occurs in parallel run, min and MD on
a single processsor are ok.

I changed the dominator to a smaller value (2.5D0),
then everything is ok.
I think you are right, the nonbonds are not being
divided roughly equally among processors. cz in my
simulations, the protein (in TIP3P water) becomes more
and more extended (denaturation simulation).

Thank you for your help!

Best Regards,
Jian Zhang

--- "David A. Case" <case_at_scripps.edu> wrote:
> On Wed, Nov 13, 2002, Zhang Jian wrote:
> >
> > I run sander in parallel to calculate a protein
> with
> > 36 residues and about 3000 waters, totally 8864
> atoms.
> >
> > I gave the following command in a SGI IRIX
> machine,
> > ==============================================
> > mpirun -np 8 sander -O -i ...
> > ==============================================
> > then got
> > ==============================================
> > * NB pairs 206 317419 exceeds
> capacity
> > ( 317571) 4
> > SIZE OF NONBOND LIST = 317571
> > EWALD BOMB in subroutine ewald_list
> > Non bond list overflow!
> > check MAXPR in locmem.f
> > ==============================================
> > I examined the file locmem.f, ew_direct.f and
> > ew_setup.f, then know that it is because
> > num+numpack>maxnblst (206+317419>317571)
> > in ew_direct.f.
> > Since maxnblst=MAXPR, I changed the line
> > MAXPR=MAXPR/numtasks
> > to
> > MAXPR=MAXPR/numtasks+1000
>
> You probably need to change MAXPR by much more than
> 1000. Look for line
> 479 in locmem.f:
>
> maxpr_float = natom * (cutoffnb +
> skinnb)**3 / 3.5d0
>
> and change the denominator to a smaller value (say
> 3.0 or 2.5).
>
> However, the fact that you have many more non-bonded
> interactions than the
> program expects suggests that something is wrong
> with your coordinates (maybe
> lots of things are on top of each other?? or the
> system is very
> inhomogeneous??) Check (visually or otherwise) that
> the coordinates are OK.
> "Ordinary" sets of coordinates should not have this
> sort of error.
>
> > To increase maxnblst in ew_setup.f will result in
> same problem.
>
> No; as the message indicates, you need to work with
> MAXPR in locmem.f.
>
> >
> > p.s. this error seems only occur when running
> sander
> > in parallel.
>
> Try a short minimization or equilibration on a
> single processor, then see
> if those coords work OK in parallel. I suspect that
> somehow, you have a
> fairly inhomogeneous system, so that the nonbonds
> are not being divided
> roughly equally among processors.
>
> ..good luck...dac
>
> --
>
>
==================================================================
> David A. Case | e-mail:
> case_at_scripps.edu
> Dept. of Molecular Biology, TPC15 | fax:
> +1-858-784-8896
> The Scripps Research Institute | phone:
> +1-858-784-9768
> 10550 N. Torrey Pines Rd. | home page:
>
> La Jolla CA 92037 USA |
> http://www.scripps.edu/case
>
==================================================================
>

=====
Jian Zhang, Dr
Institute for Biophysics
Nanjing University
22 Hankou Road, Nanjing 210093
P. R. China
Tel: 86-25-3594476
Fax: 86-25-3595535

__________________________________________________
Do you Yahoo!?
Yahoo! Web Hosting - Let the expert host your site
http://webhosting.yahoo.com