ABSTRACT
This chapter introduces the implementation of a fully-connected interconnection network with a cross-bar switch. Pipelining and parallel processing can be used to improve system throughputs, often significantly. Parallel architectures and algorithms for general-purpose parallel processing have been extensively studied for several decades. In a parallel processing system, processing elements cooperate to produce the desired result. An informal definition for the term “parallel processing” is that multiple processing elements (PE) are utilized in a coordinated manner to support the solving of a problem. The PEs in these general-purpose parallel architectures is often limited to a few complete central processing units (CPUs). General-purpose parallel architectures are especially efficient for vector and matrix operations. The shuffle-exchange interconnection network can be enhanced by changing the shuffle connections into bidirectional links. Pipeline architecture works in the same principle as an assembly line to speed up a strictly sequential task. A pipeline’s performance depends on the availability of data.
