[gmx-users] About gromacs run...

[Mark Abraham]

Mark Abraham wrote:
> Henry Yang wrote:
>> Dear users,
>> Happy new yera to all.
>>
>> I have just a very simple query..
>> I am running a simulation for 100 ns. I have decided to run it by 10 
>> ns step of each. In every run, the gromacs generated confout.gro file, 
>> the traj.trr, traj.xtc, md.log , ener.edr file, state.cpt etc. The 
>> next time when i run the simulation for the next 10 ns , I have edited 
>> the em.mdp 
> 
> It's probably best not to name your simulation .mdp file "em.mdp" which 
> would normally be for an Energy Minimization.
> 
>> file and just change the tinit to the last time that i was run and the 
>> nsteps to 5000000 for 10 ns.. And the command i used for grompp is ...
>>
>> grompp -f em.mdp -c confout.gro -p topol.top -o topol.tpr
>> And then it will generate the topol.tpr file. and the command is then
>>
>> mdrun
>>
>> My question is is that oaky ... or I have to use any more file in the 
>> grompp command like that traj.trr or state.cpt for tracking velocity??
> 
> With the advent of checkpointing in GROMACS 4, there's no reason to want 
> to use this strategy. Consult the GROMACS wiki pages on restarts and 
> continuing runs. Set up your original grompp for your expected maximum 
> run, and choose an appropriately conservative checkpoint interval in the 
> mdrun command line, 

Sorry, that's in the .mdp file. If you have to cope with queue system 
time limits, use mdrun -maxh. Read the wiki! :-)

Mark

> and suitable n*out .mdp options. If later you need 
> to extend the run, use tpbconv -extend and supply the final checkpoint 
> file to the next mdrun. There should be no need in 4.x to re-use grompp 
> unless you need to change something more than the length of the simulation.
> 
> Mark