AMBER Archive (2006)

Subject: Re: AMBER: nmode memory allocation error

From: Cenk Andac (cenk_andac_at_yahoo.com)
Date: Mon Sep 11 2006 - 04:48:30 CDT

This sounds to me more like a problem germain to your parallel computer facility.
 Although 8 nodes were allocated to share cpus, random access memories are somewhat not shared or not allowed to share high amount of memory. You may want to ask the computer person who is in charge of the parallel computing facility in your local area to see if he/she can manage to get the physical memories shared between different nodes. You guys may also want to increase the size of shared memories in your parallel computer system. Again you may want to ask your computer person about this.
 
 good luck.
 
 jenk.

AYTUG TUNCEL <ATUNCEL_at_ku.edu.tr> wrote: Dear Amber users,

I am trying to calculate binding free energy of a protein complex composed of 6874 and 6902 atoms (with hydrogens) for corresponding ligand and receptor chains using MMPBSA method with Amber8. When I try to make nmode calculation the program gives the following error message after sander minimization,

/home/atuncel/amber8/exe/nmode -O -i nmode_lig.in -o nmode_lig.1.out -c\ sanmin_lig.1.restrt -p /home/atuncel/energy/SS_A+LS_B/SS_A.prmtop not running \properly

and in the nmode_lig.1.out file it says,

 File generated by mm_pbsa.pl \
       ntrun maxcyc ibelly drms
          1 100 0 0.10E+00
       rcut scnb scee dielc idiel
     99.00000 2.00000 1.20000 4.00000 0
      nsave dfpred bdwnhl smx emx alpha ndiag
         20 0.01000 0.10000 0.08000 0.30000 0.80000 10
 ipol = 0
 i3bod = 0
 nvect = 0
     Binary format used for input coords.
| New format PARM file being parsed.
| Version = 1.000 Date = 08/24/06 Time = 11:20:47

    PARM file has the title:

Total memory required : 638255263 real words

Total memory required : 24068644 integer words

Maximum nonbond pairs 23622500
ASSERTion 'ier == 0' failed in nmode.f at line 105.

I am aware that this is a memory allocation problem and I need ~5,1 GB memory. The thing is I used 8 parallel Pentium 4, 2.4 Ghz, 1Gb memory machines.

Here are my questions,

1) Is this source of memory not enough or is it possible to make the calculations if I increase the memory?

2) Before the calculations I made the following changes during installation,
i)changed the MAXAT in molsurf.h to 20000 atoms
ii)changed to single precision using pb_def.h file
iii)changed the parameters in sizes2.h file as follows;

parameter (memdrv=10000000)
parameter (maxatom=15000)
parameter (maxint=500)
parameter (maxvec=500)
parameter (maxdih = 5500000000)
parameter (maxdia = 3500000000)
parameter (maxinb = 9000000000)
parameter (maxbon = 2600000000)
parameter (maxbnh = 2000000000)
parameter (maxang = 3000000000)
parameter (maxanh = 3000000000)

Can the error be somehow related to these modifications ?

3) I know that nmode calculations are very difficult for such big systems,can it be appropirate if I use the ismem option in NMODE, otherwise I will have to decrease the number of atoms by just taking the binding domains into account for the calculations.

Thanks in advance.

Aytug

ps: attached is my sanmin_lig.1.out

                 
---------------------------------
Yahoo! Messenger with Voice. Make PC-to-Phone Calls to the US (and 30+ countries) for 2¢/min or less.
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu