There should be no PDB files among the output files unless specifically requested, which your command should not produce. If they are step*.pdb, that means your run is crashing and these are intermediates that mdrun produces in an effort to help you see where things are breaking apart. In my experience, they are not particularly helpful.

I also STRONGLY recommend the use of the -deffnm option with mdrun, so you get clean, explicitly named output, rather than potentially relying on a host of default files. For example:

# Run 1 ns

gmx mdrun -deffnm md_0_1

# Extend from 1-2 ns

gmx convert-tpr -s md_0_1.tpr -extend 1000 -o md_1_2.tpr

gmx mdrun -deffnm md_1_2 -cpi md_0_1.cpt -noappend

From that, it is obvious which simulation intervals are being done and you have no file name conflicts.

http://www.gromacs.org/Documentation/How-tos/Extending_Simulations

The actual time values are written to the .log file.

tpbconv -s previous.tpr -extend 1000000 -o next.tpr

would that be correct? if I want  to extend my MD time to another 20ns?

or should I use 2000 000.

I am new to Gromacs sorry for silly questions.

The value supplied to the -extend option is time in ps, not steps (see tpbconv -h).  So -extend 20000 is what you want.

thanks!

More easy!!!  Use grompp or use tpbconv. For the grompp case, you will

need to change the gen_vel value in the mdp file to “no”; set unconstrained_start to

“yes” and read in the previous trajectory from previous trr or xtc file and the previous

velocities from the edr file.

Extending runs using tpbconv and mdrun checkpoint files in version 4.0 – Extending

runs is even easier in version 4.0 and is binary identical because you are using the

checkpoint file.

Using grompp to do restarts is not as robust as simply extending with tpbconv and using mdrun -cpi.  One should only use grompp if settings (i.e. output or ensemble) need to be changed, and if you do the restart/extension this way, you should not use .trr/.edr, but use the .cpt instead, because it has full state information.  grompp -t .trr -e .edr will get you close, but it is (strictly speaking) not an exact restart.  Also note that an .xtc is of insufficient precision for continuations.

to check the length at the end use gmxcheck -f traj.xtc

Thank you all ...!! I had same doubt in this regard. Everything has cleared now. The above answers are really very useful. 

What i have to do after new .tpr creation?

After getting new .tpr, just run simply:

mdrun -v -deffnm new.tpr without file extention(tpr).

and once your new run finished just concatenate both previous and new xtc files using trjcat command.

@ Sudhir Kumar

what is the exact command used to do trjcat?

For version 4 or newer you can do:

tpbconv -s previous.tpr -extend timetoextendby -o next.tpr

mdrun -s next.tpr -cpi previous.cpt

Concatenating is simple.

trjcat -f *.xtc -o full.xtc

The program detects the time values in each trajectory and puts them in order automatically.

Thanks Elias and Justin for answers. Ghada sorry for too late reply but above answers are absolutely correct.

I have a similar question, I ran a simulation for 50ns and want to extend to another 40 ns. I do not want to generate the velocity in my second run but want to information to be from previous simulation.

Is it safe to use   grompp -f new.mdp -c old.tpr -o new.tpr -t old.cpt ??

Hi everyone

I also had a same problem. All this conversation really helped me..

Extremely thanks to all of you..

gmx grompp -f md.mdp -c previous.gro -t previous.cpt -p topol.top -o new.tpr

There should be no PDB files among the output files unless specifically requested, which your command should not produce. If they are step*.pdb, that means your run is crashing and these are intermediates that mdrun produces in an effort to help you see where things are breaking apart. In my experience, they are not particularly helpful.

I also STRONGLY recommend the use of the -deffnm option with mdrun, so you get clean, explicitly named output, rather than potentially relying on a host of default files. For example:

# Run 1 ns

gmx mdrun -deffnm md_0_1

# Extend from 1-2 ns

gmx convert-tpr -s md_0_1.tpr -extend 1000 -o md_1_2.tpr

gmx mdrun -deffnm md_1_2 -cpi md_0_1.cpt -noappend

From that, it is obvious which simulation intervals are being done and you have no file name conflicts.

The .log file is rarely useful, but if you want energy and trajectory information, of course you need to keep the output.

In newer version 5, they have modified it. I got confused with that.

Hi Urszula Sylwia Uciechowska-Kaczmarzyk ,

You can extend your GROMACS mdrun using the following commands consecutively:

a) gmx tpbconv -s previous.tpr -extend timetoextendby -o next.tpr b) gmx mdrun -s next -cpi previous.cpt

You can refer to the log file it outputs in the mdrun to check for the number of steps it has completed and hence the total time of the MD run

look at this page:

http://www.gromacs.org/Documentation/How-tos/Extending_Simulations

how much time of extension is this (1000)?

gmx convert-tpr -s md_0_1.tpr -extend 1000 -o md_1_2.tpr

Dear all,

I used the instructions for gromacs 4.5 (http://www.gromacs.org/Documentation_of_outdated_version/How-tos/Extending_Simulations)

tpbconv -s previous.tpr -extend timetoextendby -o next.tpr mdrun -s next.tpr -cpi previous.cpt

I used parallelization (mpi) for mdrun

the log file of parallelization said:

"starting mdrun 'Protein in water'

5000000 steps, 10000.0 ps (continuing from step 2500000, 5000.0 ps).

Writing final coordinates."

"Parallel run - timing based on wallclock.

NODE (s) Real (s) (%)

Time: 92086.629 92086.629 100.0

1d01h34:46

(Mnbf/s) (GFlops) (ns/day) (hour/ns)

Performance: 330.648 19.027 4.691. 5.116"

So, I know that the run was done properly and that the final time was 10 ns

Nevertheless previous .log, .xtc, .cpt, .edr files were not updated. Am I doing something wrong?