Subject-oriented programming supports composition of object-oriented programs or program fragments called subjects. This paper presents an approach to the composition rules used to specify composition details. Rules can be generic, allowing different subrules to be "plugged into" higher-level rules, and they include a means of specifying exceptions to general rules. We give definitions of a number of useful, generic rules, including merge and override, as a first step towards a generally-useful composition rule library. We also outline an object-oriented framework for implementing rules, which we are currently building as part of our support for subject-oriented programming in C++.

An experimental APL compiler has been developed for the IBM SP1 distributed memory parallel machine. It accepts classical APL programs, without additional directives, and generates parallelized C code for execution on the SP1 machine. The compiler exploits data parallelism in APL programs based on parallel high-level primitives. Program variables are either replicated or partitioned. We also present performance data for five moderate-size programs running on the SP1.>