面向DSP应用的可重构计算(英文)

Reconfigurable Computing Exploration for DSP Applications

DSP应用的特点是计算密集并适合并行处理,传统的可编程处理器与ASIC在性能和灵活性上各有优劣。因此出现了一种新的计算模式—可重构计算。由于它能将效率和灵活性很好地结合在一起,故正得到广泛的关注和研究。本文在介绍可重构计算的概念和分类的基础上,着重讨论了一些主流的可重构计算系统,分析了各类系统应用于DSP的特点,对可重构计算在计算模型,编译器,映射技术以及开发环境等方面的现状和趋势进行了探讨,并给出了自己的思考。

Within the DSP applications, which are computationally intensive and desirable for parallel processing, programmable processors and ASICs have their own advantages and defects respectively. Consequently a new computing paradigm-reconfigurable computing has received wide attention and under extensive researches due to its ability to combine the efficiency with flexibility. The paper introduces the basic concept and taxonomy of reconfigurable computing, describes some main-stream reconfigurable systems and analyzes their characteristics. Then the paper discusses the challenges and directions in developing computation models, compiler, mapping technique, as well as entire development environment for the reconfigurable architectures, and draws the conclusions.