Gravity solvers

Basic

REB_GRAVITY_BASIC

The basic gravity routine works is the default. It works in most cases. It uses direct summation to calculate gravitational forces between all particle pairs. OpenMP parallelization is implemented. If OpenMP is turned on, the scaling is \(O(N^2)\), otherwise, it is \(O(\frac12 N^2)\), where \(N\) is the number of particles.

Compensated

REB_GRAVITY_COMPENSATED

This routine also uses direct summation but in addition makes use of compensated summation to minimize round-off errors. There are only a few special cases where the round-off error in force calculations has a dominant effect. In most cases, the basic gravity routine is faster and equally accurate.

Tree

REB_GRAVITY_TREE

This method uses an oct tree (Barnes and Hut 1986) to approximate self-gravity. It scales as \(O(N \log(N))\).

Tree

REB_GRAVITY_JACOBI

Direct summation, scales as \(O(N^2)\), includes special terms needed for some symplectic integrators.

Custom

REB_GRAVITY_CUSTOM

Some integrators and some special setups require a custom gravity routine. By setting the gravity variable in reb_simulation to REB_GRAVITY_CUSTOM, the function pointer gravity_custom is called whenever REBOUND required an update to the accelerations.

None

REB_GRAVITY_NONE

By using this gravity routine, no self-gravity calculated. It is still possible to include additional forces.