An efficient N-body algorithm for a fine-grain parallel computer

Numerical simulations of complicated N-body systems require codes with significantly greater capacity than provided by existing schemes. For example, to study the collision of two galaxies with realistic self-gravitating disks, we will probably need to integrate the equations of motion of ~ 2.5 x 10 s particles. Problems of this kind require algorithms which can handle a wide range of physical scales without any special assumptions about the symmetry of the system. Conventional algorithms have difficulty meeting these requirements: direct-summation codes are far too slow, FFT codes can handle only a limited range of scales, and multipole-expansion codes make restrictive assumptions about the geometry. This report describes an N-body algorithm for a parallel computer combining the generality of direct-summation codes with the favorable scaling properties of FFT and multipole-expansion codes.