一种异构多核系统的编译方法及实现  

A Compilation Method and Realization for Heterogeneous Multi-core Systems

作　　者：刘丹丹[1] 杨灿美[1] 倪素萍[2] 杜学亮[2] 

机构地区：[1]中国科学技术大学电子科学与技术系微纳电子系统集成研究中心,安徽合肥230027 [2]中国科学院自动化研究所国家专用集成电路设计工程技术研究中心,北京100190

出　　处：《微电子学与计算机》2015年第11期1-5,共5页Microelectronics & Computer

基　　金：中国科学院战略性先导科技专项(XDA06010402)

摘　　要：面向专用领域计算加速的异构多核处理器近年来得到长足发展,异构多核处理器中集成了多个不同架构的处理器核.由于该类处理器的异构性,其编程方法较传统的同构多核处理器有很大不同,编程者需要就不同架构的处理器核分别编写程序代码并分别编译,增加了软件开发难度.在分析异构多核处理器体系结构、程序执行模型的基础上,提出了一种异构多核系统的编译方法,并给出系统实现,解决了分别编写程序代码和编译的困难,支持异构多核代码的统一编程,屏蔽底层硬件的异构性,为上层用户开发提供方便.Dedicated areas for heterogeneous multi-core processors to accelerate computing has made rapid progress in recent years.Heterogeneous multi-core processor integrates a number of different processor cores.Due to the heterogeneity of the class processor,its programming are very different compared to the traditional homogeneous multicore processors.The programmer needs to write program code separately and compiled respectively for different architecture processor core,which increasing the difficulty of software development.On the basis of on the analysis of the heterogeneous multi-core processor architecture,program execution model,we propose a compilation method for compiling heterogeneous multicore systems,solving the difficulties of programming code and compiled separately,and supports heterogeneous multicore unified programming code.It shields the underlying hardware heterogeneity,providing convenience for the development of upper user.

关 键 词：异构多核 异构编程 编译方法 

分 类 号：TP302.1[自动化与计算机技术—计算机系统结构]