Why this is not an all-pairs n-body
The textbook n-body simulation computes the force between every pair of particles. For sixty-five thousand stars that is more than four billion force terms per step, and it has to happen sixty times a second. Without compute shaders and shared memory, it is simply not available.
The dominant-mass approximation replaces that with a potential from a central mass plus a few heavy clumps. Every star is pulled by that potential rather than by every other star, which reduces the cost from quadratic to linear in the star count.
Crucially, it keeps the part that matters visually. Real galaxies are dominated by exactly that kind of potential, and the flattened disc, the spiral arms and the shear all come from it rather than from star-to-star interaction. What is lost is local clumping between neighbouring stars, which is invisible at this scale anyway.

