yup. molecular dynamics simulations are hard-core. and the parallelization is highly non-trivial so you need fast fabrics and complex message passing to scale. one of the raisons d'être for supercomputers

Did your simulation use the O(n) gravitational potential algorithm? You don't need to spend O(n^2) time per time step.

IIRC I was calculating forces for every pair, which is ok for small N (e.g. a solar system, rather than a galaxy).

Integrating physics simulations like that is very inefficient and inaccurate, as it does not conserve momentum.

I get it, it was a toy, but if you're really interested look up adaptive step symplectic integrators.

I was just messing doing a bit of recreational programming, with my son. It was good enough for my purposes!

Totally get it ! Just thought of suggesting it in case you wanted to dig in deeper :)