[AMBER] Error with multiple GPU run: kCalculateMolecularVirial

From: Andre Serobian <serobian.a.student.unsw.edu.au>
Date: Sun, 19 Dec 2010 22:05:45 +1100

Hi All,

On a 2X C2050 system a kCalculateMolecularVirial error terminates my md run when I execute the below run file looping it by 100 using the following mdcont.in file:

mdcont.in file:

# constant pressure dynamics 10ps
&cntrl
 imin = 0,
 ntx = 7, irest = 1,
 ntt = 1, temp0 = 288.0, tautp = 2.0,
 ntwr = 500, ntwx = 500, ntwv = 500,
 ntc = 2, tol = 0.00001,
 cut = 10.0,
 ntpr = 500,
 ntb = 2,
 ntp = 1,
 nstlim = 100000,
&end


run file:

#!/bin/bash

COUNT=0

until [ $COUNT -gt $1 ]; do
  let NJOB=COUNT+1
  let LJOB=COUNT
  mpirun -np 2 pmemd.cuda.MPI -O -i mdcont.in -o md$NJOB.out -c md$LJOB.restrt -p input.prmtop -r md$NJOB.restrt -x md$NJOB.mdcrd -v md$NJOB.mdvel -ref md$LJOB.restrt
        let COUNT=COUNT+1
done


Any ideas why this may be happening? Someone suggested a bugfix.9 issue.


Kind regards,

Andre
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Dec 19 2010 - 04:00:03 PST
Custom Search