This thesis aims to understand how to design high-performance compression algorithms suitable for hardware implementation and to provide hardware support for an efficient compression algorithm. Lossless data compression techniques have been developed to exploit the available bandwidth of applications in data communications and computer systems by reducing the amount of data they transmit or store. As the amount of data to handle is ever increasing, traditional methods for compressing data become· insufficient. To overcome this problem, more powerful methods have been developed. Among those are the so-called statistical data compression methods that compress data based on their statistics. However, their high complexity and space requirements have prevented their hardware implementation and the full exploitation of their potential benefits. This thesis looks into the feasibility of the hardware implementation of one of these statistical data compression methods by exploring the potential for reorganising and restructuring the method for hardware implementation and investigating ways of achieving efficient and effective designs to achieve an efficient and cost-effective algorithm.

In general digital control algorithms are multi-input multi-output (MIMO) recursive digital filters, but there are particular numerical requirements in control system processing for which standard processor devices are not well suited, in particular arising in systems with high sample rates. There is therefore a clear need to understand the numerical requirements properly, to identity optimised forms for implementing control laws, and to translate these into efficient processor architectures. By taking a considered view of the numerical and calculation requirements of control algorithms, it is possible to consider special purpose processors that provide well-targeted support of control laws. This thesis describes a compact, high-speed, special-purpose processor which offers a low-cost solution to implementing linear time invariant controllers.