| 引用本文: | 唐新春,郭克榕.MPP Fortran 程序中串行循环的优化.[J].国防科技大学学报,1996,18(3):102-108.[点击复制] |
| Tang Xinchun,Guo Kerong.Optimizing Sequential Loops in MPP Fortran Programs[J].Journal of National University of Defense Technology,1996,18(3):102-108[点击复制] |
|
| |
|
|
| 本文已被：浏览 6043次 下载 6653次 |
| MPP Fortran 程序中串行循环的优化 |
| 唐新春, 郭克榕 |
|
(国防科技大学 计算机系 湖南 长沙 410073)
|
| 摘要: |
| 在大规模并行处理系统中，通讯开销是影响程序性能的一个重要因素。本文提出了一种 MPP Fortran 程序中串行循环的优化技术：在串行循环中加入同步控制、将串行循环转换成共享循环。该技术能减少通讯开销、提高程序的性能。 |
| 关键词: 数据分布，通讯开销，依赖距离向量，同步控制 |
| DOI： |
| 投稿日期：1995-12-29 |
| 基金项目:国防科技预研基金资助项目 |
|
| Optimizing Sequential Loops in MPP Fortran Programs |
| Tang Xinchun, Guo Kerong |
|
(Department of Computer Science，NUDT，Changsha，410073)
|
| Abstract: |
| Communication overhead is an important factor which affects program performance in massively parallel processing systems. This paper introduces an optimizing technique for sequential loops in MPP Fortran programs. This technique，through application of synchronization control to sequential loops and changing sequential loops into shared ones，will reduce communication overhead and improve program performance. |
| Keywords: deat distribution,communication overhead,data dependance distance vector,synchronization control |
|
|
