并行计算性能可视化方法研究

Research on Parallel Computing Performance Visualization

分布式并行系统中，并行程序通常表现出比传统串行程序复杂得多的行为。影响并行程序性能的各种因素的交互作用，使得获得好的并行计算性能成为系统设计者和用户所面临的极大挑战。近年来的研究表明，并行程序的性能可视化方法和工具在帮助用户理解并行程序行为和优化性能等方面起到了很大的作用，成为各种并行性能分析优化工具不可缺少的一部分。本文较为系统全面地综述了并行计算性能可视化方法研究的发展状况，讨论了其面临的主要困难及可能的解决方法，并指出了未来的若干研究方向。

Parallel and distributed computing often exhibits more complicated behavior than conventional serial counterpart. The correlated impacts of all facets of parallel computing make achieving optimal performance a formidable challenge for system designers and programmers. Parallel performance visualization research and tools have shown great help in gaining insight into parallel program behavior and performance optimization. This paper systematically reviews the development of parallel performance visualization. We discuss the key technical issues in this area and point out some future directions.