Technical Report Number
Chip Multi-Processors (CMPs) are now available in a variety of systems and provide the opportunity for achieving high computational performance by exploiting application-level parallelism. In the communications environment, network processors (NPs), designed around CMP architectures, are generally usable in a pipelined manner. This leads to the need for static scheduling of tasks on processor pipelines. This thesis considers problems associated with determining optimal schedules for such pipelines. A collection of algorithms is presented with their utility determined by the size and other characteristics of the system. The algorithms employ heuristics, dynamic programming and statistical methods to schedule tasks derived from multiple application ﬂows on pipelines with an arbitrary number of stages. Experimental results indicate that while the dynamic programming algorithm obtains the optimal schedules, heuristics and statistical methods obtain schedules within 10% of the optimal, 95% of the time. Examples are given to show the use of these algorithms for general pipeline/algorithm design and for use in the Network Processor environment with typical networking applications.
Datar, Seema, "Pipeline Task Scheduling with Appication to Network Processors" Report Number: WUCSE-2004-48 (2004). All Computer Science and Engineering Research.
Permanent URL: http://dx.doi.org/10.7936/K7JS9NSQ