[TinusNl]

I was wondering what i did wrong. Somewhere in the forum i read that number of subdomains must be equal to number of cores to be used. Even that i have a 4 core processor i would like to reserve 1 core for fast testing before starting a real simulation.
However the script below still gives me 4 processors.



numberOfSubdomains 2;

method simple;

coeffs
{
n (1 1 1);
}


Any suggestions? Also changing the n to (1 0 0 ) not changes anything.

[Yann]

Hi TinusNI,

It seems you are a bit confused about decomposeParDict, I suggest you to read the proper section of the user guide: https://www.openfoam.com/documentati...ns-in-parallel

In your decomposeParDict you are requesting 2 subdomains (ie 2 cores) but the coefficients you defined correspond to only 1 subdomain: (1 1 1) means your domain is cut by 1 in each direction => 1x1x1=1 subdomain. Another example to make it clearer: (3 2 1) will gives you 3x2x1=6 subdomains.

If you intend to run on 3 cores, your decomposeParDict should be something like this:

numberOfSubdomains 3;

method simple;

coeffs
{
	n (3 1 1);
}

Of course you also need to start the solver with the proper command, usually:

mpirun -np 3 yourSolver -parallel

Cheers,
Yann

[TinusNl]

Thanks.


That was what i was looking for a small example.


I see that i also need to have a closer look on the other options to see if there is a way to speed up my calculations efficient as i not have that many cores available.