基于OpenCL的并行方腔流加速性能分析

摘要

This paper proposed a new approach to accelerate cavity flow problem using OpenCL technology. This paper transformed the cavity flow problem into N-S equation and solved it by finite difference method. During this procedure, used local memory and other techniques to accelerate the code. This paper ran the program on both NVIDIA and ATI platforms. The experiment shows a 30x times speed up.%提出了一种使用OpenCL技术对方腔流问题进行加速计算的方法.在计算方腔流问题时,将其转换为N-S方程通过空间有限差分和龙格库塔时间差分求解,并使用局部缓存等技术进行GPU优化.实验在NVIDIA和ATI平台对所给算法进行评测.结果显示,OpenCL相对其串行版本加速约30倍左右.