Towards a Field-Programmable Physics Processor (FP?)

Abstract

In this paper, we outline the design and implementation of an FPGA-based numerical integrator that will ultimately form the basis of our FPGA-based physics engine. Physics engines are considered one of the most important of a multitude of components requesting CPU time in a modern computer game, and so we propose offloading aspects of this physics computation to an FPGA-based physics engine. We aim to ameliorate the speed of the physics computation in concert with the other game components. Currently, our physics processor uses the fourth-order Runge-Kutta numerical integration algorithm to solve the ordinary differential equations used in game physics. Our analyses indicate that the performance of our physics processor should surpass that of the equivalent software executing on a CPU when several objects are simulated.