[gmx-users] Parallelization in large system

[Rachel Baarda]

Dear Gromacs users,

I am trying to simulate a large system (roughly 3M atoms including solvent)
on Gromacs version 4.5.5. I am having difficulty having mdrun execute
parallel calculations. I have run several variations (described below) of
attempting steepest descent minimization and the simulation finishes
successfully only if I set the number of threads to just 1.

variation 1: mdrun -v -deffnm steep_to_1000
Result:
Making 1D domain decomposition grid 8 x 1 x 1, home cell index 0 0 0
Program mdrun, VERSION 4.5.5
Source code file: /share/apps/pkg/gromacs-4.5.5/src/mdlib/domdec_top.c,
line: 356
Fatal error:
8346 of the 4590648 bonded interactions could not be calculated because
some atoms involved moved further apart than the multi-body cut-off
distance (1 nm) or the two-body cut-off distance (1 nm), see option -rdd,
for pairs and tabulated bonds also see option -ddcheck

variation 2: mdrun -v -deffnm steep_to_1000 -noddcheck
Result: identical to variation 1 (6528 of the 4422192 bonded interactions
...)

variation 3: mdrun -v -deffnm steep_to_1000 -nt 1
Result:
Successfully finishes run without incident (but very slowly)

I also tested a simulation of a smaller system (~15k atoms), which ran
without incident on a 1D domain decomposition grid 8 x 1 x 1.

What parameters do you recommend I change to allow for multithreading?

Thank you in advance,

Rachel Baarda