The allocation problem in distributed signal processing

Functional allocation in a distributed signal processor is examined from the point of view of efficient use of system processing resources. An efficient allocation is one that either 1) achieves a smaller overall cycle time for a given processing sequence for the same, or fewer, resources than that of any other, or 2) uses a smaller number of resources than any other to satisfy a maximum allowable cycle time. It is shown that certain kinds of allocation will usually result in more efficient system utilization when intermediate data transfers among resources are a requisite part of the processing sequence. Thus, for example, cornerturning processes are usually more efficiently carried out when separate groups of processing elements are allocated to the pre-and post-corneturning functions, rather than uniformly sharing the implementation of all the functions among all the processing elements.