Analyzing and improving the latency and throughput performance of self-timed pipelines and rings
Ted E. Williams · 2003
An analysis of self-timed pipelines and rings is presented. The author defines terminology, qualitatively describes data and bubble flow, and defines the variables locally characterizing each stage in a ring. Values for these variables can be found using the dependency graph analysis method presented. The author then determines the overall performance of self-timed rings in terms of those variables locally describing the stages. The performance is expressed on graphs showing regions defined by the number of stages and tokens. Separate equations give the ring latency and throughput within each region. Cross-sections graphing latency and throughput versus the number of stages and tokens are presented, and the implications of this analysis for self-timed ring design are discussed.>