Qinghua Liao wrote:
Dear gmx users,
I tried to do simulations of a small peptide in vacuum, I found that it
failed to be run in parallel, even when I use only 8 cores.
My system only have hundreds of atoms. the problem may be resulted from
domain decomposition. When I choose particle
decomposition method, for small system, I can use 4 threads but not 8,
and for a little bigger systems, I can only use 8 threads.
For this situation, is it normal? Is there some solution to this
problem? Thanks very much!
You can't necessarily parallelize any system over an arbitrary number of
processors. Even if you could, you'd likely lose performance due to the latency
of node-node communication. For a system of a few hundred atoms, running in
parallel may not even yield significantly better performance over a serial run.
-Justin
The following lines are my mdp file for the vacuum simulation:
title = PDXN of Abeta in H2O
;cpp = /lib/cpp ; prepocessor of the current machine
define = ;-DPOSRES
integrator = md ; molecular dynamics algorithm
tinit = 0.0 ; start time and timestep in ps
dt = 0.002 ; time step in ps
nsteps = 500000000 ; number of steps for 1000ns run
emtol = 100 ; convergence criterion
emstep = 0.05 ; intial step size
nstlist = 10 ; step frequency for updating neighbour list
ns_type = grid ;simple ; method for neighbour searching (?)
nstxout = 5000 ; frequency for writing coords to output
.trr file
nstvout = 0 ; frequency for writing velocities to
output...should be same as nstxout
nstfout = 0 ; frequency for writing forces to output
nstlog = 5000 ; frequency for writing energies to log
file
nstenergy = 5000 ; frequency for writing energies to
energy file
nstxtcout = 5000 ; frequency for writing coords to xtc traj
xtc_grps = system ; group(s) whose coords are to be
written in xtc traj
energygrps = system ; group(s) whose energy is to be written
in energy file
pbc = no ; use pbc
rlist = 0 ; cutoff lengths (nm)
epsilon_r = 1.0 ; Dielectric constant (DC) for
twin-range or DC of reaction field
niter = 100 ; Some thingies for future use
fourierspacing = 0.16
fourier_nx = 30
fourier_ny = 30
fourier_nz = 30
coulombtype = Cut-off ; truncation for minimisation, with
large cutoff
rcoulomb = 0
rcoulomb-switch = 0
vdw-type = Cut-off ; truncation for minimisation, with
large cutoff
rvdw-switch = 0
rvdw = 0 ; cut-off lengths
;pme_order = 6 ; EWALD/PME/PPPM parameters
;ewald_rtol = 1e-05
;ewald_geometry = 3d
epsilon_surface = 0
optimize_fft = yes
Free energy control stuff
free_energy = yes
init_lambda = 0.0
delta_lambda = 0
sc_alpha =0.5
sc-power =1.0
sc-sigma = 0.3
comm_mode = angular
nstcomm = 10 ; number of steps for centre of mass
motion removal (in vacuo only!)
Tcoupl = V-rescale
tc_grps = system ; MVN_Protein ;SOL_Ion ; Non-Protein
tau_t = 0.01
ref_t = 300
Pcoupl = no ; Parrinello-Rahman ; Pressure coupling
;Pcoupltype = Isotropic
;tau_p = 1.0 1.0 1.0
;ref_p = 1.0 1.0 1.0
;compressibility = 4.5e-5 ; compressibility
;
annealing = no ; SIMULATED ANNEALING CONTROL
;zero_temp_time = 0 ; Time at which temperature should be
zero (ps)
gen_vel = yes
gen_temp = 300
gen_seed = -1
constraints = all-bonds ; OPTIONS FOR BOND CONSTRAINTS
constraint-algorithm = Lincs ; Type of constraint algorithm
lincs_order = 4 ; Highest order in the expansion of the
constraint coupling matrix
lincs_iter = 1
lincs_warnangle = 30 ; Lincs will write a warning to the
stderr if in one step a bond rotates
; over more degrees than
unconstrained-start = no ; Do not constrain the start configuration
;Shake-SOR = no ; Use successive overrelaxation to
reduce the number of shake iterations
;shake-tol = 1e-04 ; Relative tolerance of shake
morse = no ; Convert harmonic bonds to morse potentials
--
Best Regards,
Qinghua
--
========================================
Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin
========================================
--
gmx-users mailing list gmx-users@gromacs.org
http://lists.gromacs.org/mailman/listinfo/gmx-users
Please search the archive at
http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
Please don't post (un)subscribe requests to the list. Use the
www interface or send it to gmx-users-requ...@gromacs.org.
Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
