Discussion

We have reviewed the compilation strategies the current HPJava adopts in this chapter. For multiarray types and HPspmd classes, we have explained in detail the philosophy of design for multiarray types as well as the programming restrictions on HPspmd classes. We have seen the whole picture of the current HPJava type-checking system with some explanations for the functionality of each phase such as type analysis, reachability nalysis, definite (un)assignment analysis. The current HPJava type-checker is totally based on the specification from ``The Java Language Specification'' [26]. Moreover, we reviewed the rules and restriction of the pre-translation phase. Through the isolation of the pre-translator from the translator itself, we have seen that the translator phase becomes simpler and more efficient. The main issue of this chapter is to create logically well-defined and well-designed ``basic'' translation schemes for the HPJava system. We have reviewed the basic translation schemes we adopted to the current system. Because of the simplicity of the current HPJava translator, we need to keep researching to create better translation schemes for the system. Moreover, the naivety leads our research to a new direction; optimization. With simple and efficient translation schemes and effective optimization we will see in the next chapter, we can say the future of the HPJava system will be promising.