基于Matlab的遗留系统并行化重构方法

摘要

While the multicore platform became popular, the applications got complicated and data set expanded, the full power of system can' t be exploited by the serial code in the Matlab-based legacy program, the requirement of large scale data processing can't be fulfilled. The parallel computing model of Matlab provides the support for parallelization of the data-intensive application. The key-points and methods of refactoring for parallelization of legacy system are analyzed from the view of system architecture expanding and business logic parallelization, the performance promotion of large scale data processing is verified by the application test results in the refactored system.%随着CPU多核架构的普及,应用的复杂和数据集的膨胀,基于Matlab的遗留系统中的串行程序代码无法充分发挥系统潜在的性能优势,无力应对当前大型数据集的处理应用需求.Matlab的并行计算模型为数据密集型的处理任务提供了并行支持.本文首先从系统架构扩展和业务代码并行化入手,分析遗留系统并行化重构过程要点和方法,应用案例的并行化重构实验数据表明了系统重构处理大型数据集的性能提升.