摘要
科学计算应用问题在计算模型、处理过程以及对处理器、存储器和通信的要求方面存在巨大的差异。同一种结构难以适应差异巨大的应用,造成效率低下。解决途径在于可变的体系结构,让体系结构去适应不同的应用,而不是让应用去适应单一的体系结构。不同的应用决定了不同的应用程序结构,对于不同的应用程序结构,采用最适合的体系结构去匹配,实现应用决定体系结构的模型。结果表明,基于应用驱动的异构体系结构异构模型能够达到使运算高效率、低能耗的目的。最后给出了从不同应用得到最适应体系结构的方法。
Scientific computing application problems greatly differ in computing models, processing and the requirements of processors, memories and communications. The same structure is difficult to adapt to vastly different application problems, which results in low efficiency. The solution depends on the variable architecture, while making the architecture adapt to different applications, rather than making the different applications adapt to the same architecture. Differenl application problems determine different application structures. The most suitable system architecture was used to match a specific application structure to implement the model in which the application determines the architecture. The results show that the heterogeneous blended architecture model based on application drive has high computing efficiency and low power consumption. The method is given to obtain most suitable architectures from different applications in this paper.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2013年第2期234-241,共8页
Journal of Tsinghua University(Science and Technology)
基金
国家"八六三"高技术项目(2007AA01Z425)
国家自然科学基金资助项目(90718015
60673157)
安徽省优秀青年人才基金项目(2012SQRL183)
关键词
异构体系结构
匹配
应用驱动
效率
heterogeneous architecture
match
application drive
efficiency
