[gmx-users] crash nsgrid.c

[Fabio Affinito]

Dear all,
During my simulations of a GFP protein (force field made with amber,  
then converted to gromacs), I usually get this error:

Fatal error: ci = 8108 should be in 0 .. 7937 [FILE nsgrid.c, LINE 218]

What this that means?

If I restart the simulation from the frame just before the crash, the  
run is ok for a (random) time and then it crashes again.
I checked this on different kind of machines and I get always the same  
behavior.

To make things clearier, I include the input.mdp file:

title                    =
cpp                      = /lib/cpp
include                  =
define                   =
integrator               = md
dt                       = 0.002
nsteps                   = 500000
nstxout                  = 10000
nstvout                  = 0
nstlog                   = 500
nstenergy                = 500
nstxtcout                = 500
energygrps               = System
pbc                      = xyz
nstlist                  = 15
epsilon_r                = 1.
ns_type                  = grid
coulombtype              = pme
vdwtype                  = Cut-off
rlist                    = 0.8
rcoulomb                 = 0.8
rvdw                     = 0.8
tcoupl                   = Nose-Hoover
tc-grps                  = System
tau_t                    = 2.0
ref_t                    = 300.00
pcoupl                   = Parrinello-Rahman
pcoupltype               = isotropic
tau_p                    = 4.0
compressibility          = 4.5e-5
ref_p                    = 1.0
gen_vel                  = no
gen_seed                 = 173529
constraints              = All-bonds
constraint_algorithm     = lincs
shake_tol                = 0.0001

Any suggestion is welcome.

Thanks

F.A.