2018-2019春季学期

可重构系统基础  Fundamentals of reconfigurable systems

  • 课程类别:任选课
  • 授课时间:周一10-12节
  • 授课学期:2018-2019春季学期
  • 授课地点:
  • 课程学分:3.0
  • 授课教师:王韬

 

课程简介:
        近年来,用以现场可编程门阵列(FPGA)为代表的可重构逻辑进行计算的想法,引发了人们的兴趣。可重构逻辑具有一个独特的性质:当它被制造出来后,可以被重新配置以改变其实现的逻辑功能;而一旦被配置后,它就可以被视作一个能够实现特定逻辑的硬件。
        对于许多应用,与通用处理器相比,可重构逻辑通常可以实现好得多的性能/功耗指标,在一些计算中得到几倍到几十倍的性能提升。与针对特定任务设计的专用集成电路(ASIC)相比,虽然可重构逻辑的性能/功耗指标通常要更弱一些,但是它的可重构特性给它带来了一个重要优势:它可以被用来实现那些用通用处理器无法满足其性能/功耗需求的,而又无法及时(或者因代价太高而不能承受)用ASIC实现的新应用。
       目前国际上逐渐出现一个热点—研究者们将可重构逻辑作为可以及时实现新功能的加速器,构建可重构系统,以提高整体性能/功耗指标,并对创新应用提供更好的特性/市场响应时间。最近,国际最大两个FPGA厂商之一的Xilinx公司,也发布了由两个ARM Cortex-A9处理器核心与FPGA组成的可扩展处理平台(Xilinx Extensible Processing Platform)。
      本课程拟对可重构系统的原理、组成、结构进行介绍,讲授在可重构逻辑上的硬件编程语言(Verilog)、编程方法,讨论在可重构系统中适合的应用与算法,并介绍若干前沿研究专题。在课程中,还将有一定的实践练习,在FPGA开发平台上进行设计。
     通过此课程的学习,同学们将掌握有关可重构系统的基础知识,并可以针对可重构系统进行应用/算法开发,以充分利用高性能/低能耗的可重构系统。
 

        Recently people have shown interest in using reconfigurable logic (RL), which is represented by field programmable gate array (FPGA), to do computing. RL has an unique property: after fabricated, it can be configured many times to implement various logic functions; and after configured, it can be treated as a fixed hardware to implement a specific logic function.
        For many applications, RL can generally achieve much better performance/power, and several times to several tens of times speedup over general processors. Compared with application specific integrated circuit (ASIC), which is specifically designed for a specific function, RL shows some disadvantages in performance/power, but its reconfigurability brings out an important advantage: it can be used to implement those new-coming creative applications whose performance/power requirments cannot be satisfied by general-purpose processors and cannot be implemented in time by ASICs.
        People have shown great interest in using RL as accelerators to implement new-coming creative applications in time and building reconfigurable systems; this will improve the performance/power of the whole system, and provide better time-to-property/market to new-coming creative applications. Recently, Xilinx, one of the two most important FPGA vendors in the world, announced the Xilinx Extensible Processing Platform, which is composed of two ARM Cortext-A9 processor cores and FPGA.
        In this course, the principles, organization, and structure of RL will be introduced. The hardware description language Verilog and its programming model on RL will be taught. The algorithms and applications suitable on RL systems will be discussed. Some advanced topic in RL research will be presented. And there will be also some practices on FPGA developing platform.
        The students enrolled in this course will understand the fundamentals of reconfigurable systems, and have the capability of developping algorithms/applications in reconfigurable systems to fully leverage the performance/power advantages of the reconfigurable systems.