1. Introduction
Compilers for high-level languages form a significant part of most computer systems, and with an ever increasing number and variety of machine architectures on the market the problems of compiler development, testing, and maintenance consume more and more manpower and computer time. Moreover, as computer technology is improving and changing rapidly it is becoming evident that software costs will increasingly dominate the total cost of a system. Indeed, it may not be long before the lifetime of software regularly exceeds that of the hardware on which it was originally implemented, a state of affairs quite different from that envisaged by Halpern when he concluded that "the importance of the entire question of machine-independence is diminishing .." [Halpern, 1965]. In addition, there is a need to encourage the slowly-developing trend to write the majority of software in high-level languages. Even though the advantages of such an approach are many, a large number of users still have a love of machine-code, usually fostered by thoughts of "machine efficiency". Clearly, techniques must be developed to simplify the production of usable compilers which can "optimise" the match between the executing program and the user's requirements, be they for fast execution, small program size, reasonable execution time but with good run-time diagnostics, or whatever.