Fire and Flame Simulation using Particle Systems and Graphical Processing Units

摘要

Simulating fire, flames or other natural phenomena can be difficult because of the inherently complex systems used to model them, while also requiring an adequate amount of realism visually. Simulating such a system in real-time can also be a problem if the system is too large, so a parallel computing techniques can be used to good effect. Particle systems have been shown to simulate flames and fires particularly well at relatively low computational cost. We describe how a simple particle system approach can be used to simulate a fire or flame in real-time in conjunction with using data parallelization, achieving a substantial performance speed up on graphical processing units (GPUs). Using NVidia's Compute Unified Device Architecture (CUDA) and OpenGL interoperability functionality allows for further performance increases when rendering the simulation with GPUs. Additionally, different rendering techniques are used to investigate trade-offs between performance speed and visual realism.