[gmx-users] parallel run with integrator=nm ?

[J. Rui Rodrigues]

Dear all,

In an attempt to speed up things, I tried to calculate an hessian matrix in 
parallel, using two processors and it finished in about half the time when 
compared to a single-node run. However, the parallel run produces an .mtx 
file that has about half the size of the one obtained in the single-node run.
Running g_nmeig_d on the .mtx from the parallel run generates negative 
eigenvalues, while eigenvalues from the single-node run are OK.

Is this a known issue? I could not found anything in the archives. 
mdrun_d_mpi runs just fine in parallel with integrator=steep or md, and 
produces the expected error with l-bfgs. This is with gromacs 4.0.5 and 
4.0.7, compiled in 64 bit, double precision, with or without mpi, in a Mac
(OSX 10.5.8).

Commands used:
grompp_d -f nm.mdp -c lbfgs.gro -t lbfgs.trr -p protMg.top -o nm.tpr
mpirun -np $NSLOTS /usr/local/gromacs/bin/mdrun_d_mpi -s nm.tpr -v -deffnm nm -mtx hessian.mtx
g_nmeig_d -f hessian.mtx -s nm.tpr

Thanks for your time,
Rui Rodrigues