首页 | 本学科首页   官方微博 | 高级检索  
     

面向共享虚拟存储系统的编译支持的通信优化技术
引用本文:洪锦伟,陈国良,张兆庆,章峰.面向共享虚拟存储系统的编译支持的通信优化技术[J].计算机学报,2000,23(7):738-743.
作者姓名:洪锦伟  陈国良  张兆庆  章峰
作者单位:1. 中国科学技术大学计算机科学与技术系,合肥,230027
2. 中国科学院计算技术研究所高性能中心,北京,100080
基金项目:国家“八六三”高技术研究发展计划!( 863 -3 0 6-ZD0 1-0 2 -3 )
摘    要:当前的面向消息传递的自动并行编译器由于各种原因,只能支持极有限的一类数据并行应用程序,共享虚拟存储系统SVM和面向共享存储的编译技术的相结合有效地解决了这一问题,使得编写并行应用程序变得十分容易,而且易于移植,然而,研究表明通信是影响基于SVM的应用程序性能的主要因素之一。在文中,针对底层SVM的特性,提出了页面对齐、软流水等编译支持的通信优化技术,在SVM系统JIAJIA的平台上实现了制导辅助折

关 键 词:共享虚拟存储系统  自动编译  通信优化
修稿时间:1999-09-28

Compiling-Support Communication Optimizations for SVMs
HONG Jin-Wei,CHEN Guo-Liang,ZHANG Zhao-Qing,ZHANG Feng.Compiling-Support Communication Optimizations for SVMs[J].Chinese Journal of Computers,2000,23(7):738-743.
Authors:HONG Jin-Wei  CHEN Guo-Liang  ZHANG Zhao-Qing  ZHANG Feng
Abstract:Currently parallelizing compilers for message passing machines can only support a limited class of data parallel applications. One method for eliminating this restriction is to combine powerful shared memory parallelizing compilers with software Shared Virtual Memory systems. However, studies show that communication is one of the main factors of system's overhead. This paper presents two compiling support communication optimizing technologies such as page aligning computation partition and software pipelining. It implemented a directives assisted automatic compiler Autopar4jiajia on the platform of a SVM system called JIAJIA. By testing with the suite of NAS, it is showed that these compiling support communication optimizing technologies are very useful to improve the performance of applications running on SVMs.
Keywords:Dawn1000A  Shared Virtual Memory system (SVM)  directives-assisted automatic compiling  computation partition  software pipelining
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号