论文部分内容阅读
在当前共享存储对称多处理 ( SMP)并行机上 ,基于指导语句的并行程序设计模式 ,讨论了多功能油藏数值模拟软件中求解超过百万节点规模的解法器 ( MFS)的并行和优化技术。首先 ,结合当前微处理器的高性能特征 ,为了提高 Cache命中率 ,改进了数据和循环结构 ,并组织了 MFS的性能优化 ,在 R50 0 0上获得了 2 0 %的性能提高 ,并消除了并行化将可能引入的 Cache一致性冲突 ;然后 ,基于循环合并、区域分解和大粒度流水线并行技术 ,实现了 MFS的并行化 ;最后 ,在 POW-ER CHALL ENGE R80 0 0的 6台处理机和 R1 0 0 0 0的 8台处理机上 ,对三维三相 50万和 1 0 0万节点规模问题 ,分别组织了数值实验 ,并取得了超过 60 %的并行效率。
Based on the parallel programming paradigm of instructional statements in current shared memory symmetric multiprocessing (SMP) parallel machines, the parallelism and optimization techniques for solving MFS with multi-million-node scale in multi-function reservoir numerical simulation software are discussed. First of all, combined with the high-performance features of current microprocessors, to improve the cache hit ratio, the data and the loop structure are improved and the performance optimization of the MFS is organized, a 20% performance improvement is obtained on the R50 0 0 and the Parallelization will probably introduce the Cache consistency conflict; then, based on the loop merge, the regional decomposition and large-grain parallelism of the pipeline technology to achieve parallel MFS; Finally, POW-ER CHALL ENGE R80 0 6 processors and On the eight processors of R1000, the numerical experiments were organized respectively for the three-dimensional three-phase 500,000 and 10 million node size problems and the parallel efficiency of over 60% was achieved.