Hi all,
I have just experienced a strange coordinate writing issue with a pmemd
simulation. I ran a 1 ns standard NPT MD of a small solvated protein
writing to disk very often (10 gig total traj size), when at a random
point (after 418ps) the structure when viewed in vmd appears to become
completely scrambled.
A relevant cutout of my crd file for the first problematic snapshot looks
like this:
[...previous, normal looking snapshot coords, box dims]
66.000 31.000 90.000 39.000 12.000 12.000 39.000 14.000 85.000
4.000
15.000 78.000 40.000 47.000 55.000 20.000 99.000 74.000 45.000
17.000
0.000 89.000 83.000 67.000 75.000 13.000 52.000 4.000 30.000
38.000
[...]
so all digits after the comma are zero, with more or less random numbers
before the comma. This continues until at about 840ps the coordinate
format changes to
5.546 7.463 6.360 4.644 7.784 6.368 5.643 7.044 5.504
5.460
7.453 6.269 3.176 9.175 8.587 2.930 8.372 8.128 3.862
8.902
so non-zero digits after the comma, but all numbers are positive and <10.
Both result in very weird structures, of course.
The system energy, temperature, stability, density etc. are not affected
by this, neither is the final rst file which contains a normal structure,
so I think it is a coordinate writing problem. Has anyone encoutered
something like this before? Is it a possible MPI problem? Any ideas on
where to look further?
input file:
&cntrl
ntx=5, irest=1,
ntc=2, ntf=2,
nstlim=500000,
ntpr=50, ntwx=50,
ntwr=2500000,
dt=0.002, cut=9.,
ntt=1, tautp=10.0,
temp0=298.0,
ntb=2, ntp=1, taup=10.0,
&end
Kind Regards,
Thomas
Dr. Thomas Steinbrecher
formerly at the
BioMaps Institute
Rutgers University
610 Taylor Rd.
Piscataway, NJ 08854
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Apr 05 2012 - 06:00:04 PDT
