Closed Francyrad closed 1 year ago
The tree code scaling is not $O(N)$ but $O(N log(N))$. Regardless, the issue with your setup is that you initialize your particles in 2D but scale their radii as if the particles were 3D. As a result you will have significantly more collisions between particles as you increase the number of particles.
Dear @hannorein
I did some experiments for my moon formation scenario and i found this pattern:
Basically when i run my scenario using 10k particles, it takes usually 0.2 seconds with my cpu power of 20000. If i run 100k particles for the same problem, it's gonna cost up to 1 minute to do 10 s of simulation.
My question is: is this normal with hardsphere? I would expect just a 10x of execution time (2s), my particles are not even touching in the beginning of my simulation, or my code as somekind of a bottleneck using mpi?
Still thank you for your help and best regards