[gmx-users] Re: gmx-users Digest, Vol 19, Issue 26

[George Rydlew]

On Mon, 2005-11-07 at 20:48 +0100, David wrote:
> > Dear GROMACS users,
> >
> >   I am doing POPC membrane simulation using gromacs 3.1.4 with:
> >
> > pbc = xyz
> > constraints =  hbonds
> > constraint_algorithm = shake
> >
> > put in parameter file. Everything goes without any problems as long as
> > I run it in serial. However any attempts to continue simulation in
> > parallel causes either "inconsistent shifts" or "Shake did not
> > converge in 1000 steps" errors.
> > In particular when
> >
> > unconstrained_start = yes
> > gen_vel =  no
> >
> > is set in md.mdp and followed by:
> >
> > grompp -f md.mdp -n popc.ndx -deshuf popc_deshuf.ndx -p popc.top -t
> > popc-serial.trr -c popc-serial.gro -np 2 -shuffle -o popc-parallel.tpr
> > mpirun -np 2 mdrun_mpi -deffnm popc-parallel
> >
> > mdrun crashes with "inconsistent shifts". Another way around:
> The problem may be that your topology gets shuffled, but not your
> coordinates. Try it without shuffle first. If that work make an inverse
> deshuf.ndx file from the original deshuf.ndx and apply that to your
> original trr (last frame only will suffice):
> trjconv -f popc-serial.trr -b XXX -n reshuf.ndx -o reshuffled.trr
> and feed that trr file to grompp.

There seams to be two different issues. The trajectory file indeed
don't get shuffled. After manual reshuffling it was possible to run
the simulation in gmx 3.2.1 Resultant trajectory, after  deshuffling,
seams to be ok. However, in gmx 3.1.4 I got "Shake did not converge"
error - same as when starting from ".gro". Can this be attributed to
athlon processors I am running this simulations on? On gromacs.org
there is a mention of work around for problems with SSE on athlon,
that was introduced in v3.2.1

> >
> > trjconv -f popc-serial.trr -o par-in.gro -s popc-serial.tpr -n
> > popc.ndx -b 25480 -e 25480
> > grompp -f md.mdp -n popc.ndx -deshuf popc_deshuf.ndx -p popc.top -c
> > par-in.gro -np 2 -shuffle -o popc-parallel.tpr
> > mpirun -np 2 mdrun_mpi -deffnm popc-parallel
> >
> > causes "Shake did not converge in 1000 steps" and simulation crash
> > within few subsequent steps (here md.mdp was also altered to perform
> > unconstrained run and preserve velocities). I tried to solve this by
> > setting GMXFULLPBC to 1 but to no avail in both cases.
> >
> > Now, I managed to continue simulation in gromacs 3.2.1 by extracting
> > last frame from the trajectory (second approach from the above). Doing
> > grompp with "-t popc-serial.trr" was causing "inconsistent shifts"
> > also in this version. Anyway, I don't like that solution, as it
> > requires changing of gromacs version in the middle of simulation and,
> > as far as I understood, does not perform accurate restart due to
> > precision of velocities in .gro file.
> >
> > Thanks in advance for any clue on what is the reason for such behavior
> > and possible solutions.
> >
> --
> David.
> ________________________________________________________________________
> David van der Spoel, PhD, Assoc. Prof., Molecular Biophysics group,
> Dept. of Cell and Molecular Biology, Uppsala University.
> Husargatan 3, Box 596,          75124 Uppsala, Sweden
> phone:  46 18 471 4205          fax: 46 18 511 755
> spoel at xray.bmc.uu.se    spoel at gromacs.org   http://xray.bmc.uu.se/~spoel
> ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Thanks
  George