Ray,
Thanks. I'll prepare the files and let you know.
Rami
On Sep 16, 2011, at 11:33 PM, Ray Luo, Ph.D. wrote:
Roman,
That's interesting ... do you mind sharing with us your data file so
we can reproduce your observation?
All the best,
Ray
On Fri, Sep 16, 2011 at 3:18 PM, Osman, Roman <roman.osman.mssm.edu<mailto:roman.osman.mssm.edu>> wrote:
Hello,
I have a peculiar problem with MMPBSA. I get the same results for GB but different results for PB depending on the number of processors I use.
Here is the summary:
The results of gb are identical regardless of the number of processors;
the results of pb depend on the number of processors:
gb pb
5proc -57.7880 -77.1179;
4proc -57.7885 -68.7393;
3proc -57.7866 -71.9419;
2proc -57.7888 -95.5026
1proc -57.7881 -176.9453
These are the final DELTA
The question is whether this is a problem with the implementation on our computers or a more fundamental problem.
The input file is:
Input file for running GB
&general
verbose=2, strip_mdcrd=1, strip_mask=":WAT",
entropy=0, startframe=1000, endframe=10000, interval=1000
/
&decomp
idecomp=2, dec_verbose=3,
/
&gb
igb=5, saltcon=0.150
/
&pb
istrng=0.150
/
and the script (this case for 2 processors) is:
#!/bin/bash
#PBS -N MM_ALL
#PBS -j oe -V
#PBS -q para_std
#PBS -l select=2:ncpus=1
#PBS -m e
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/share/apps/intel/lib/intel64
export AMBERHOME=/share/apps/amber11
export PYTHONPATH=/share/apps/python2.6/lib/python2.6/site-packages\:$PYTHONPATH
MMPBSA="/share/apps/amber11/bin/MMPBSA.MPI"
PYTHON="/share/apps/python27/bin/python2.7"
cd $PBS_O_WORKDIR
SUB="mpirun -np 2 $MMPBSA -O -i decomp_tb.in -o decomp_tb.dat -sp ../tryp_bpti_1000.top -cp ../GENR/tryp_bpti_nw.top -rp ../tryp_nw.top -lp ../bpti_nw.top -y ../tryp_bpti_1000.trj -do decomp_tryp_bpti.dat"
echo $SUB
$SUB
Thanks for your help
Roman Osman
roman.osman.mssm.edu<mailto:roman.osman.mssm.edu>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org<mailto:AMBER.ambermd.org>
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org<mailto:AMBER.ambermd.org>
http://lists.ambermd.org/mailman/listinfo/amber
Roman Osman
roman.osman.mssm.edu<mailto:roman.osman.mssm.edu>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Sep 17 2011 - 08:30:02 PDT
