FJRR:一种基于阵列型数据划分应用的并行设计模式

FJRR:parallel design pattern for array-based data decomposition applications

并行设计模式能够降低并行编程的难度与复杂度。针对科学与工程计算领域大量耗时的数据密集型应用的计算,提出了一种能够适用于阵列数据划分应用的FJRR并行设计模式;鉴于开发者更习惯串行编程,在FJRR模式中提出并实现了一种类串行化并行程序设计的思想。通过两个示例进行了模式应用说明与性能分析,并将该模式在EasyPAB可扩展并行应用开发平台中实例化并应用。

Parallel design pattern can ease parallel programining. Aiming at the computations of time-consuming data-intensive applications in scientific and engineering fields, the paper proposed a parallel design pattern named FJRR from the aspect of design pattern. In consideration of user＇ s familiarization with serial programming, brought forward a serial-like parallel programming idea from user＇ s aspect and well implemented in the pattern. As illustration, gave two examples and well experimentally analyzed. The pattern has been put into use in the ongoing extensible IDE framework named EasyPAB.