[gmx-users] BUG in GROMACS 4.0.5, related to very log total integration time

[Mark Abraham]

Daniel Adriano Silva M wrote:
> Dear GROMACS users and developers.
> 
> I don known if this issued had been previously addressed, but I found
> that when I try to run a MD with time-step of 2fs and 2500000000 steps
> (yes 500us!!!) the dynamics aborts with the next message (64bit-LINUX
> and icc 10 compiler):
> 
> ####################
> WARNING: This run will generate roughly 20576946697451257856 Mb of data
> 
> starting mdrun 'F1-ATPASE'
> -1794967296 steps, -3589934.8 ps.
> 
> nodetime = 0! Infinite Giga flopses!
>         Parallel run - timing based on wallclock.
> 
> #####################
> 
> If i reduce the steps number by one order of magnitude then all goes
> ok.  My MDP when I obtined this error was:
> 
> 
> ; VARIOUS PREPROCESSING OPTIONS
> title                    = NPT simulacion
> cpp                      = /lib/cpp
> 
> ; RUN CONTROL PARAMETERS
> integrator               = md
> dt                       = 0.002
> nsteps                   = 2500000000
> nstxout                  = 5000
> nstvout                  = 5000
> nstlog                   = 2500
> nstenergy                = 2500
> nstxtcout                = 2500
> energygrps               = Protein Non-Protein
> nstlist         = 10
> rlist           = 1.0
> ns_type         = grid
> pbc             = xyz
> coulombtype              = pme
> rcoulomb                 = 1.0
> vdw-type                 = Cut-off
> rvdw                     = 1.0
> fourierspacing       =  0.12
> pme_order            =  4
> optimize_fft         =  yes
> ewald_rtol           =  1e-5
> tcoupl                   = Berendsen
> tc-grps                  = Protein  Non-Protein
> tau_t                    = 0.1      0.1
> ref_t                    = 300      300
> Pcoupl                   = Parrinello-Rahman
> Pcoupltype               = Isotropic
> tau_p                    = 1.0
> ref_p                    = 1.0
> compressibility          = 4.5e-5
> gen_vel                  = no
> constraints              = all-bonds
> constraint-algorithm     = Lincs
> unconstrained-start      = yes
> lincs-order              = 4
> lincs-iter               = 1
> lincs-warnangle          = 30
> 
> I known this kind of error is not a priority since the total
> integration time is ridiculous big, but anyway I want to comment it to
> you.

Yes, it's known. IIRC there was some discussion on the developers list 
about changing the relevant data type so that it can store bigger 
numbers. It's also a non-problem inasmuch as if you are ever able to run 
a simulation that long, manually resetting the number of steps to zero 
at a suitable point will be workable.

Mark