[gmx-users] Possible bug: energy changes with the number of nodes for energy minimization

[Stephen Cox]

Hi,

I'm running a number of energy minimizations on a clathrate supercell and I
get quite significantly different values for the total energy depending on
the number of mpi processes / number of threads I use. More specifically,
some numbers I get are:

#cores      energy
1        -2.41936409202696e+04
2        -2.43726425776809e+04
3        -2.45516442350804e+04
4        -2.47003944216983e+04

#threads    energy
1        -2.41936409202696e+04
2        -2.43726425776792e+04
3        -2.45516442350804e+04
4        -2.47306458924815e+04


I'd expect some numerical noise, but these differences seem to0 large for
that. Before submitting a bug report, I'd like to check:
a) if someone has seen something similar;
b) should I just trust the serial version?
c) have I simply done something stupid (grompp.mdp appended below);

Any help greatly appreciated.

Steve

NB I am using a flexible methane and settles for TIP4P/2005. I have
reproduced the above trend with ubuntu repository versions of
gromacs/openmpi and my own compiled versions of gromacs/mpich2.

; run control
integrator = steep
dt         = 0.001
nsteps     = -1
comm_mode  = linear
nstcomm    = 1

; energy minimization
emtol  = 0.01
emstep = 0.01

; output control
nstxout       = 1000
nstvout       = 1000
nstfout       = 1000
nstlog        = 1000
nstcalcenergy = 1
nstenergy     = 1000

; neighbour searching
nstlist            = 1
ns_type            = grid
pbc                = xyz
periodic_molecules = no
rlist              = 0.9

; electrostatics
coulombtype = pme
rcoulomb    = 0.9

; vdw
vdwtype  = cut-off
rvdw     = 0.9
dispcorr = ener

; ewald
fourierspacing = 0.1
pme_order      = 4
ewald_geometry = 3d
optimize_fft   = yes

; temperature coupling
tcoupl          = nose-hoover
nh-chain-length = 10
tau_t           = 0.5
ref_t           = 300.0
tc_grps         = system

constraints
constraint_algorithm = lincs
shake_tol            = 0.0001
lincs_order          = 8
lincs_iter           = 2
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20120529/182f6372/attachment.html>