Classification of Parallel Architectures

Parallelism appears in various forms, including look ahead, pipelining, vectorization, multiplicity, replication, time sharing, multitasking, multi programming, multi threading, and distributed computing at different processing levels. Let’s know more about classification of parallel architectures.

The most commonly used classification of parallel architectures is that given by Flynn in 1966. Flynn considered the presence or absence of potential multiplicity in the instruction and data streams of computer systems.

The instruction stream is the programmed instructions of operations. Thus, an architecture is categorized by the multiplicity of the hardware used to manipulate the instruction and data streams.

In this manner, it is possible to obtain four classes of computers as described below.

Classification of parallel architectures

Classification of parallel architectures

Classification of parallel architectures

Single instruction single data StreamConventional computers belong to the single instruction single data stream (SISD) category where there is one central processing unit in the computer which can perform  one instruction at a time.

This represents the serial architecture illustrated in figure 6.8. this, until recently, has formed the basis for nearly all computers.

Single Instruction Multiple Data StreamThe single instruction multiple data stream (SIMD) systems are those in which the same instruction is executed, at the same time, by several processors, as illustrated in Figure 6.9.

Clearly such a system comprising of N processors can have a throughput of N times more than an SISD machine. Examples include pipelined vector processors and array processors.

Multiple Instruction Single data StreamIn a multiple instruction single data stream (MISD) architecture, the processors can share common memory and each processor operates according to the instructions it receives via it own instruction stream (Fig. 6.10).

This class includes all forms of multiprocessor configurations form linked main-frame computers to large arrays of microprocessors. Multiple processors capable of independent operations is the main characteristic of MIMD computers.

You may like also:

Leave a Reply

Your email address will not be published. Required fields are marked *