The Research Of Reconfgurable Computing Architectures And Applications

Reconfgurable Computing(RC), refers to systems incorporating some form of hardware pro-grammability–customizing how the hardware is used using a number of physical control points.These control points can then be changed periodically in order to execute diferent applicationsusing the same hardware. The emergence of RC is supposed to fll the gap between hardware solu-tion and software problem solution, and to make a compromise: achieving potentially much higherperformance than pure software solution, while maintaining a higher level of fexibility than purehardware solution. Reconfgurable computing now is one of the most hotspot in the high perfor-mance computing feld. The research of RC can be divided into two levels: RC application and RCarchitecture level.In application level, the research major contains two aspects: specifc application and generalapplication research. In the frst aspect, it focuses on the selected application and study for how tooptimizethecomputingmodel,dataorganization,communicationmodel,andalgorithmimplemen-tation to get the best performance. On the other hand, the general application research try to builda template for a collection of similar problems. When facing a specifc problem, the researchersonly need to modify the interface of the template which in accordance with the character of thereconfgurable hardware to get the highly optimized solution.In hardware level, the problem is divided into fne grain reconfgurable hardware and coarsegrain reconfgurable hardware two aspects. The dominant fne grain reconfgurable hardware isFPGA. So the research emphasis about the fne grain reconfgurable hardware is the research aboutFPGA chip. Yet the FPGA enjoys many difculties: too long confguration time, high power con-sumption, long mapping&routing time, low resource utilization and large data transfer latency,the research focal point is transferring to the coarse grain reconfgurable platform.This dissertation includes the whole two levels: beginning with the application research andthen diving into hardware architecture research. At the same time, our works can be divided intothree parts: specifc application research, general application research, and novel heterogeneousreconfgurable architecture research.Inspecifcapplicationresearch, weselectthefacedetectionsystemasexample, whichiswide-ly used in video surveillance, safe access control and may other felds. By improving the image reduction threshold, image reduction algorithm and integral image calculation pipeline, we let theface system ft the reconfgurable computing hardware FPGA. Experiments also demonstrate theperformance of the reconfgurable hardware based system is far better than traditional softwarebased solution.In general application research, we focus on the recursive problem and take it as a example.The recursive problem is one of the basic problem in symbolic logic, artifcial intelligence feld.We build the common, parallel reconfgurable hardware based solution template for the recursiveproblem by dividing recursive search tree, balancing recursive problem partition method.In novel reconfgurable computing architecture research, we design an novel reconfgurablecomputing architecture to settle the difculties existed in present reconfgurable hardware. Thenovel architecture based on wireless communication mechanism, heterogeneous function unit andefcienttaskschedulingmechanism. Ourarchitectureenjoysadvantagebothinpowerconsumption,performance and confguration time issues.