Multirate process scheduling and synchronization in distributed signal processors

Multirate signal processing requires different segments of a signal processing graph to be executed at different rates. Graph segments executed in each distributed processing element of a distributed signal processor must be scheduled to avoid excessive peaks and valleys in computation loading. Also, graph segment scheduling in pipelined distributed processing elements, performing different parts of the same problem, must be synchronized. When application control code and parameter tables for individual distributed processing elements can be developed relatively autonomously, and when control interaction between distributed processing elements is minimized, programming complexity is reduced. This decomposition of large system processing requirements makes the use of high order, graph based languages more feasible. Scheduling and synchronizing signal processing tasks using high order language graph specification has been studied extensively in recent years 2,3,4,5. This paper describes a method and algorithms for scheduling graph segments within distributed processing elements, load smoothing within distributed processing elements, and distributed processing element synchronization for multirate signal processing problems. The criterion employed for all scheduling, load smoothing and synchronization is a modulo processing interval count. The information required to schedule each multirate graph segment is represented in a single high order language statement containing a vector of nine parameters which are generated at compile time.