[gmx-users] Procedure for running Gromacs in Parallel??

[Dallas Warren]

Vivek,

>I was going through the previous mails on users list... which confused me 
>even more. Some mails say that "I have to run mdrun 'under' mpirun." I 
>didn't find any command such as mpirun in gromacs. Some mentioned that we 
>have to use "np= node numbers  & then -shuffle -sort in grompp command", 
>and then run the "mdrun under mpirun".  It is still not clear to me what 
>is exactly meant by that. I didnt imagine that running a parallel job can 
>be so confusing!!!

mpirun is a command that comes from the LAM MPI program, which is why you 
don't see it in the GROMACS package.  It is what sets up and manages the 
mdrun program to run across more than one processor.  However, you don't 
necessarily have LAM MPI, there are other programs that can handle the 
parallel stuff too.

>Could anyone explain in detail, what commands exactly I need to use to 
>make the simulation run in parallel, say on 10 nodes?

Assuming running LAM MPI, you need to initialise LAM using the "lamboot 
hostfile" command.  Hostfile contains a list of the IP address of all of 
the CPUs that you want to run it on.  If a dual machine then simply list 
the IP twice.

Then generate the binary topology file using "grompp -f grompp.mdp -c 
conf.gro -t traj.trr -p topol.top -o topol.tpr -np 10".  Can also use the 
-shuffle and -sort switches to get better performance.

Then "mpirun C mdrun -s topol.tpr -o traj.trr", no need to put the -np 10 
switch in here for some reason.

>Also, how to check your node performance??... I want to see how the nodes 
>perform after I start the parallel jobs. Which command should I use for this??

Look at how fast the calculation is performed, either in the GROMACS log 
file listing the total calculation time, speed of the simulation, in 
ps/hour.  Calculate a scaling factor by dividing the time on a single CPU 
by the time on n CPUs multiplied by n.  Closer the value is to 1, the more 
efficient it is.

Hope that helps.

Catch ya,

Dr. Dallas Warren
Research Fellow
Department of Pharmaceutical Biology and Pharmacology
Victorian College of Pharmacy, Monash University
381 Royal Parade, Parkville VIC 3010
dallas.warren at vcp.monash.edu.au
+61 3 9903 9083
--------------------------------------------------------------------------
When the only tool you own is a hammer, every problem begins to resemble a nail.