[gmx-users] 3.3 parallel runs - Lincs/PME/sorting errors?

[Alan Dodd]

I'm still playing with the use of 3.3 with my lipid
systems.  I've been trying to equilibrate a 257-lipid
patch of bilayer generated with genbox and a smaller
block of lipid.  Obviously, it has big cavities around
the edges, so I wasn't expecting a flawless run. 
However, examining a crashed trajectory made with 4
nodes showed that water molecules were mysteriously
clustering towards the centre of four blocks along the
x-axis.  Is there some kind of maximum size of a
system, beyond which PME no longer works properly over
multiple processors?  Could it be an error in
deshuffling?  Or is it not a problem at all?

Alan Dodd
University of Bristol

details from .mdp file:
cpp                 =  /lib/cpp
constraints         =  all-bonds
integrator          =  md
dt                  =  0.0015   ; ps !
nsteps              =  10000    ; total 7500 ps.
nstcomm             =  1
nstxout             =  1000
nstvout             =  1000
nstfout             =  0
nstlog              =  1000
nstenergy           =  1000
nstlist             =  10
ns_type             =  grid
coulombtype         =  PME
rlist               =  1.5
rcoulomb            =  1.5
rvdw                =  1.5
pme_order           =  8
fourierspacing      =  0.2
optimize_fft        =  yes
; Berendsen temperature coupling is on in two groups
Tcoupl              =  berendsen
tc-grps             =  DOPC     SOL
tau_t               =  0.1      0.1
ref_t               =  300      300
; Energy monitoring
energygrps          =  DOPC     SOL
; Isotropic pressure coupling is now on
Pcoupl              = berendsen
Pcoupltype          = surface-tension
tau_p               =  5
compressibility     =  4.5e-5 4.5e-5
ref_p               =  410 1.0
; Generate velocites is off at 300 K.
gen_vel             =  no
gen_temp            =  300.0
gen_seed            =  173529


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com