ABSTRACT
This chapter explores the alternatives of replication in computer architecture. There are only two techniques for introducing parallelism into computer hardware, replication and pipelining; pipelining can be considered as replication which has been made possible through sequence, as each component of replication in a pipeline follows another in time. Considering just control flow architectures, there is a wide choice in the implementation of control strategy in a replicated system. The control of an array may also have a number of levels; indeed this technique is often used to combat complexity in control unit design. The implementation of a packet-switched network, if efficient, can provide a very powerful basis for many parallel computational models. With a switching network that can connect all memory banks to a given processor, it is very desirable to have a memory or data structure in which rows, columns and other principle substructures from arrays may be accessed without conflict.
