This paper studies some of the problems involved in attaining machine independence for a code generator, similar to the language independence and the token independence attained by automatic parsing and automatic lexical systems. In particular, the paper examines the logic involved in two areas of code generation: computation and data reference.
