Re: [AMBER] about MD

From: Jason Swails <>
Date: Fri, 26 Sep 2014 23:51:03 -0400

On Fri, Sep 26, 2014 at 10:47 PM, m13654940810 <> wrote:

> Hi,
> I run 'mpirun -np 16 sander.MPI -O -i -o B2.out -c B.rst -p
> B.prmtop -r B2.rst' using amber10.
> Input file
> Title ...
> &cntrl
> irest=0,
> nstlim=100000000,
> ntpr=100, ntwr=100,ntave=100,
> nscm=500,ntwx=500,
> dt=0.001, vlimit=10.0,
> cut=10.,imin=0,
> ntt=3,temp0=298,gamma_ln=5,
> ntb=2, ntp=1, pres0=1.0, taup=1.0,iamoeba=1
> /
> &ewald
> nfft1=50,nfft2=50,nfft3=50,
> skinnb=0.8,nbtell=0,order=5,ew_coeff=0.36,

skinnb is too small. I would suggest keeping the default of 2. This is
not causing your problem, though.

> &amoeba
> do_induced=1,do_recip=1,do_direct=1,do_adjust=1,
> do_amoeba_nonbond=1,do_amoeba_valence=1,beeman_integrator=1,
> do_bond=1,do_ureyb=1,do_reg_angle=1,do_trig_angle=1,
> do_opbend=1,do_torsion=1,do_str_torsion=1,do_pi_torsion=1,do_strbend=1,
> do_torsion_torsion=1,
> do_vdw=1,do_vdw_longrange=1,do_vdw_taper=1,
> do_self=1,dipole_scf_tol = 0.01,dipole_scf_iter_max=30,
> sor_coefficient=0.7,ee_damped_cut=4.5,ee_dsum_cut=6.8,
> /
> Then there is a error :Unit 16 Error on OPEN: B2.rst.

​This likely means that you do not have write permissions in the directory
where your calculation is running. Try creating a file in that directory
(e.g., by running "touch B2.rst" or something equivalent). I doubt this
particular problem is an Amber problem.

​That said, I would not expect your calculation to succeed even if you
solved this problem. Amber 10 is quite old and no longer supported. The
last time I tried to run a simulation using the AMOEBA force field with
sander in parallel, it did not work (the program hung indefinitely because
the parallel functionality was broken). So only serial execution is
supported for the AMOEBA force field in sander in AmberTools 14.

I have _no_ idea when this break occurred, but it easily could have been
before Amber 10. The pmemd.amoeba.MPI program from AmberTools 14 is the
recommended way to run the AMOEBA force field with Amber with high
performance and reasonable parallel scalability.


P.S. To my knowledge, nobody is working on fixing Amoeba parallelization in
sander. Given that even a working sander would be substantially worse than
pmemd.amoeba in terms of performance and parallel scalability, I doubt
anybody will consider it worthwhile to try.

Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Fri Sep 26 2014 - 21:00:02 PDT
Custom Search