State Abstraction-Based Synchronizatoin

• GSSM-RR-96-02[PDF]: Kuno, Y. and Ohki, A.: p6 -- A State Abstraction-Based Parallel Object-Oriented Language, Reserch Report 96-02, Graduate School of Systems Management, Univ. of Tsukuba, Tokyo, 1996.
  Roughly corresponds to: Kuno, Y. and Ohki, A.: p6 -- A State Abstraction-Based Parallel Object-Oriented Language (in Japanese), Transactions of IPSJ, vol. 38, no. 3, pp. 563-573, 1997.

Concurrent/parallel programming languages require synchronization. We are proposing concept of "abstract state," in which objects' internal states are made visible from outside, in a controlled fashon. State abstraction can be an ideal tool for specifying synchronization in parallel object-oriented programming language because it is simple, clean, and efficiently implementable. To demonsrate our correctness, we have designed and implemented a parallel object-oriented language called "p6." P6 implementation on 4CPU SparcServer 1000 has shown that very efficient implemtation of state abstraction-based synchronization can be constructed.

• OBPCD'97 Paper [PDF]: Kuno, Y.: Solving Inheritance Anomaly Problems by State Abstraciton-Based Synchronization, Proc. 1997 France-Japan Workshop on Object-Based Parallel and Distributed Computing (OBPDC'97), October 1997.
• OBPDC'97 Slides: Presentation slides for above.

At first glance, state abstraction-based synchronization seems to suffer from the famous "inheritance anomaly" problem, but the contray is true. We can solve the problem by introducing state-splitting and cartesian product functionality along with interface inheritance. However, what is more important (as we think) is that state abstraction-based synchronization provides a natural way to caputure objects' synchronization behavior as a part of their interface signagures. Such functionality is the key to solve the problem in "the right manner."