中文摘要：

基于现代计算机的多级存储结构，采用消息传递并行编程模型对格子Boltzmann并行程序进行了Cache优化．实验结果表明，优化后的程序能够减少80％的Cache缺失，性能提高20％，而且经过预处理的并行程序性能也有很大提高．

英文摘要：

Based on multi-level storage architecture of popular computer and message passing parallel programming mode, this paper discussed that the parallel program of lattice Boltzmann is optimized in cache. The results show that the Cache missing rate can be reduced by 80 %, performance can be increased by 20 % and preprocessing method can also improve parallel performance.