A distributed framework to improve high-performance IP
Meng Yeem Kah, Yann Fong Day, Jun Koh Wei · 2010
New and emerging applications in the area of high-speed processing are increasingly stretching intellectual property (IP) to provide an ever higher throughput. It becomes more and more difficult for an IP to deliver a higher throughput on a single core, as the design approaches its practical limit on frequency scaling as well as the degree of parallelism. This drives designers towards a multiprocessing approach to find an optimal balance between performance, power, cost, and design complexity. This paper presents a novel method to deliver a higher performance IP via multiprocessing. This approach uses a highly scalable architecture to create a distributed framework that enables multiple cores to work together in a pipelined manner to achieve the desired throughput. Using this approach helps the IP developer to get to market faster, and by doing so, stay ahead of the competition.