基于GPGPU的Lattice-Boltzmann数值模拟算法
Numerical Simulation Algorithm of Lattice-Boltzmann on GPGPU
-
摘要: 对Lattice Boltzmann方法(LBM)在GPGPU下的建模和算法进行了一系列研究,使得该方法在GPU下的计算加速比提升,大大缩短计算过程的时间消耗.重新设计了GPU的计算流程,在舍弃pixel buffer离屏渲染的同时,采用最新的帧缓存对象,多重纹理、多通道渲染和乒乓技术来设计一套基于方腔的LBM数值模拟程序,最终使GPU的计算时间缩短到CPU计算时间的六分之一.Abstract: A new algorithm on the numerical simulation of LBM (Lattice-Boltzmann method) based on GPGPU (General Purpose GPU) is studied in the paper.The method performs well for improving the computational efficiency on GPU and substantially reduces the time consumption of the calculation process.The computational procedure of GPU is redesigned based on the previous results.By removing the off-screen rendering on pixel-buffer,the modern technique of Frame Buffer Object,Multiple Textures,Multiple Channel Render and Ping Pang is adopted to code a new sort of numerical simulation procedure based on Lattice-Boltzmann method for cavity flow.The test results indicated that the GPU-based implementation is faster than the CPU-based,with a factor of speedup in about 6.22 times.
下载: