2 Intermediate codes
This section gives a brief account of the more important intermediate codes which have been discussed and have had an influence on the design of I-code.
2.1 Uncol
UNCOL, UNiversal Computer Orientated Language, [Mock, 1958], was an early attempt to specify a means for solving the M*N problem of producing compilers for M languages to run on N machines. It was proposed that an intermediate language, UNCOL, be defined which would be able to express the constructs from any language, and which could itself be translated into code for any machine, resulting in the need for only M+N compilers. Indeed it was even suggested that programs would be written directly in UNCOL rather than in machine code.
These ideas were very ambitious, but were presented without any concrete examples of what UNCOL might look like. Proposals were made for an UNCOL in [Steel, 1961] but the work was abandoned before anything like a complete specification had been produced.
An UNCOL-like technique which has been used extensively is to compile for a known type of machine, such as the IBM 360, and then emulate that machine on the target machine. Unfortunately, to give this any chance of being efficient, microcode support will be necessary, and this is rarely available to compiler writers.