Class partitions-a new approach to sequential object oriented programs

The concurrent object model assumes objects communicating via explicit messages. This approach is expressive but puts an additional burden on users of considering parallel issues. We advocate an implicit static analysis of programs written in a sequential object oriented language. We have developed a prototype, OFFERS, which performs implicit class analysis and suggests minimal class partitions. OFFERS views the input program as made up of interacting objects. Objects belonging to the same class are grouped together on the same processor. Such a grouping reduces the code copying (related to sharing of member functions) and avoids data coherence (an effect of static data members). The minimal class partitions suggested by OFFERS can be used by the users to fine-tune their programs and also by the run-time system to make intelligent scheduling decisions, thereby achieving high execution performance