Section 13.4
Computer Architecture
13.4 Processing Architectures
13.4.1 The Von Neumann Machine
Instructions are fetched and executed one at a time.
These instructions are fetched and executed one at a time. These instructions
are stored in main memory just like any other data. This is known as the stored
program concept.
13.4.2 Pipelining
This is a technique to speed up processing by introducing a degree of processing
by intrucing a degree of parallelism; whilst once instruction is been executed the
next instruction can be fetched.
Both the Von Neumann architecture and its pipelined version are classified as
single instruction stream, single data stream (SISD).
13.4.3 Vector Array Processing
In the CPU architecture a single instruction can operate on an array of data.
This is possible because the CPU has:
- one control unit
- several processing elements (ALUs and registers)
This architecture is known as single instruction stream, multiple data stream
(SIMD).
13.4.4 Multiple Processor Architecture
In this system, multiple instruction streams operate on multiple data streams
(MIMD). The CPU consists of several control units and several ALUs. These processing
elements have to be co-ordinated.
