[gmx-users] memory usage grows without bounds
macarignano at gmail.com
Wed Oct 24 17:48:18 CEST 2012
the same happens to me on several computers, but it depends on the system.
I'm sorry I cannot dig the examples right now to add more precise information.
Could be shake? The system that fails is full of constraints.
mdrun will not leak memory that badly. I would guess your output buffers
are not being flushed, so grow without bound.
On Oct 25, 2012 12:36 AM, "Pablo Serra" <serra at famaf.unc.edu.ar> wrote:
> Dear users,
> I'm running a NVT simulation with gromacs 4.5.5,using
> the command:
> mdrun -nt $NSLOTS
> and the memory usage grows without bounds until a crash with the message
> /opt/gridengine/default/spool/compute-0-8/job_scripts/1711: line 29: 25200
> Killed /usr/local/bin/mdrun -nt $NSLOTS
> I've got the same result in several machines with 4, 6 and 8 cores, and
> ram between 4 to 23 G. Attached is the memory map of one of a typical run.
> I'm running a system of 512 CBrCL_3 molecules (2560 atoms) using the shake
> algorithm for the constrains.
> The options in the mdp file are:
> constraints = all-bonds
> integrator = md
> dt = 0.005
> nsteps = 10000000000
> ns_type = grid
> nstxtcout = 1000
> xtc-precision = 1000
> nstxout = 0
> nstvout = 0
> nstfout = 0
> rlist = 1.6
> coulombtype = Cut-off
> rcoulomb = 1.6
> vdw-type = Cut-off
> rvdw = 1.6
> DispCorr = EnerPres
> Tcoupl = Nose-Hoover
> tc-grps = System
> tau_t = 2.0
> ref_t = 180
> Pcoupl = no
> gen_vel = no ;yes
> gen_temp = 180
> gen_seed = 1993
> ; Type of constraint algorithm
> constraint-algorithm = shake
> shake-tol = 0.0001
More information about the gromacs.org_gmx-users