一体化信号处理与先进处理架构展望

Prospect of Integrated Signal Processing and Advanced Processing Architecture

多功能一体化系统是电子信息技术领域的重要发展方向之一,一体化信号处理是其中的关键技术,对实现各种功能之间资源共享与高效协同具有重大意义,同时也从算法到处理架构提出了新的要求。本文首先对一体化信号处理的研究现状进行了归纳总结,给出一种基于一体化信号的多功能系统模型,指出进一步开发空域资源是当前一体化信号设计的主要研究方向之一,并且给出相关信号处理的典型算法;然后,针对上述算法进行了核心算子提炼,指出该类算法以大规模矩阵运算为主要特征,对处理架构提出高算力、高能效的需求;最后,针对上述算法的核心算子设计了基于存内计算和光子计算的处理架构,由于避免了数据搬移并采用高性能模拟计算模式,因此可以大幅提升算力和能效,为一体化信号处理先进架构设计提供了新的技术途径。

The multifunctional integrated system is one of the development directions in the electronic information technology field.Integrated waveform,the crucial technology in the multifunctional integrated system,has great significance for resource sharing and efficient cooperation between various functions.Meanwhile,new requirements are raised from algorithm to processing architecture.In this study,the research status of the integrated waveform processing is summarized on the comb,and an integrated signal processing system model is suggested.The further development of spatial resources is pointed out as a substantial direction in the research of integrated waveform processing.And the typical algorithms of correlation signal processing are given.Then,the large-scale matrix operation is proposed as the key characteristic by abstracting the core operator of this kind of integrated signal processing method.The corresponding system processing architecture is required to have high computing power and high energy efficiency.Finally,the In-Memory computing and photonic computing advanced architectures are designed to solve the above requirement.By reducing the data movement and employing high-performance analog computing mode,these two architectures significantly improve the capability of high computing power and high energy efficiency,providing a novel technical approach for integrated waveform advanced processing architecture research.