New Challenges for Cellular Automata Simulation on the GPU
John Tran, Don Jordan, and David Luebke
University of Virginia
Poster presented at GP2 and SIGGRAPH 2004
Cellular automata (CA) simulations consist of a collection of independently acting cells and states associated with those cells; cells change state in lockstep, with each cell's next state determined by its current state and the states of cells in some local neighborhood. The discrete nature of CA simulations allows them to be computed efficiently and a vast array of computations have been implemented as CA simulations, including simulations of physical processes more typically modeled with partial differential equations. The best-known CA simulation is Conway's "Game of Life,"' a very simple rule set that exhibits a wide range of complex behavior. An early NVIDIA demo showcased dependent texture mapping by implementing the game of life on a GPU. We have been studying the suitability and implementation of more elaborate CA simulations--in particular CA simulations using more sophisticated local neighborhood computations--on graphics hardware.