J. Makino et al., GRAPE-4 - A MASSIVELY-PARALLEL SPECIAL-PURPOSE COMPUTER FOR COLLISIONAL N-BODY SIMULATIONS, The Astrophysical journal, 480(1), 1997, pp. 432-446
In this paper, we describe the architecture and performance of the GRA
PE-4 system, a massively parallel special-purpose computer for N-body
simulation of gravitational collisional systems. The calculation cost
of N-body simulation of collisional self-gravitating system is O(N-3).
Thus, even with present-day supercomputers, the number of particles o
ne can handle is still around 10,000. In N-body simulations, almost al
l computing time is spent calculating the force between particles, sin
ce the number of interactions is proportional to the square of the num
ber of particles. Computational cost of the rest of the simulation, su
ch as the time integration and the reduction of the result, is general
ly proportional to the number of particles. The calculation of the for
ce between particles call be greatly accelerated by means of a dedicat
ed special-purpose hardware. We have developed a series of hardware sy
stems, tile GRAPE (GRAvity PipE) systems, which perform the force calc
ulation. They are used with a general-purpose host computer which perf
orms the rest of the calculation. The GRAPE-4 system is our newest har
dware, completed in 1995 summer. Its peak speed is 1.08 TFLOPS. This s
peed is achieved by running 1692 pipeline large-scale integrated circu
its (LSIs), each providing 640 MFLOPS, in parallel.