Simulation of surface fire fronts using fireLib and GPUs

This paper reports the porting of raster type fire propagation models to graphical processing unit (GPU) architectures. Two known fire growth model algorithms were ported and a new one developed. All models are based in cellular automata, programmed in the CUDA framework developed by NVIDIA and their results compared against serial execution. The novel algorithm targeted for GPU architectures achieved a speedup (SP) of over 200× against serial CPU execution. These results allow ultra faster-than-real-time capabilities that may prove useful toward fire fighting methodologies.