CPU与GPU并行计算的火焰模拟

CPU-GPU parallel computed fire simulation

采用基于粒子插值的SPH方法对火焰流体进行模拟,用GPU加速粒子状态地计算,同时用CPU并行地计算粒子邻接关系并控制粒子产生速率。在SPH模型中,较为高效地加入了漩涡场的计算,增加了粒子运动的细节。在粒子渲染过程中,采用了色度场、有向点扩散和颜色锐化技术,由离散的粒子空间分布得到了较为理想的连续火焰图像。由于该方法属于流体模拟的拉格朗日法,所以火焰具有物理真实性,又由于采用GPU为主CPU为辅的计算架构,使得模拟达到了实时。

The Smoothed Particle Hydrodynamics （SPH） method based on the particle interpolation was used to simulate the fire fluid. The Graphic Processing Unit （GPU） was utilized to accelerate the simulation, and the CPU was utilized to parallel search the neighbourhoods of each particle and control the particles＇ birth rate. Moreover, the vorticity field was added to the SPH model efficiently, which improved the details of particles＇ movement. In the final rendering period, the color field, directional point splatting and color refining technologies were applied to obtain a smoothed image of fire from discrete locations of particles. As the method belongs to Lagrangian method for simulating fluid, the fire is almost real in physical. And as the architecture is the GPU computation with auxiliary CPU, the simulation can be implemented in real-time.