[gmx-users] Re: MPI scaling (was RE: MPI tips)

[Lubos Vrbka]

David Mathog wrote:
>>You are running an extremely short number of timesteps, but increasing the
>>system *size* to make the simulations take longer. That most likely means
>>you're *mostly* increasing the overhead involved in setting up the system.
>>So of course scaling is abysmal -- you're still running trivially short
>>calculations, just *really big* trivially short calculations. Try running
>>reasonable sized calculations that are long (MORE STEPS) and then
> I don't really understand your reasoning for increasing the number
> of steps since the nodes have to do some sort of communication at
> each step, and whatever the ratio of that time to computation time
> per step is, it shouldn't vary with the number of steps.
in my opinion, the overhead mentioned above is the problem. imagine the 
following situation:

1) simulation setup
	takes 30 seconds (allocation of resources, initial creation of pair 
lists, fft optimization, lincs initialization, or whatever else might 
happen)
2) simulation itself
	takes 30 seconds on 1 processor
	takes 15 seconds on 2 processors

total 4:3 (30+30)/(30+15) instead of 'real' scaling 2:1 (30:15). if you 
increase the simulation by order of magnitude you get (since setup takes 
the same time) 330:180 that is much closer to the 'real' scaling 2:1. 
making the simulation longer by 2 orders of magnitude gives you even 
better results...

this example is an exxageration, but it should illustrate why you should 
make the test simulation longer...

regards,
-- 
Lubos
_ at _"