Code generation for NEW on the Rice Research Computer

A code generation scheme is given for the implementation language NEW to be used on the Rice Research Computer. A discussion of the run-time organization of NEW programs is given which provides for independent compilation of routines and generalized jumps using label variables. The code generator is based on an algorithm given by Sethi and Ullman, which has been extended to aid in efficient compilation of NEW constructs. Modes of elaboration are introduced to aid in the handling of assignations and jumps. A heuristic scheme is used for register allocation based on context. Finally, some problems with the language and the computer are discussed and areas for future work are presented