摘要
对于分布内存体系结构的并行计算机而言,如何对计算和数据进行合理划分以增加数据本地化减少处理器间的通信是提高其并行性能的关键。本文主要讨论了一种自动实现无数据重组的静态计算和数据划分算法。
Minimizing communication by increasing the locality of data references is an important optimization for achieving high performance on large-scale parallel machines, This paper describes a linear decomposition algorithm that automaticaUy finding computation and data decomposition, including finding data and computations decomposition that has not data reorganization communication.
出处
《微计算机信息》
北大核心
2005年第11X期195-197,共3页
Control & Automation
基金
河南省杰出人才创新基金0521000200
关键词
并行编译
数据分布
计算划分
paralleling compiler, data decomposition, computation decomposition
