Paper by Erik D. Demaine

Reference:

Erik D. Demaine, “Higher-Order Concurrency in PVM”, in Proceedings of the Cluster Computing Conference (CCC'97), Atlanta, Georgia, March 9–12, 1997.

Abstract:

Message-passing systems are typically used to achieve high-performance parallel computing. In this paper we examine how advanced concurrent programming can be achieved using existing message-passing systems. In particular, we look at Reppy's extension to Hoare's basic CSP model, called higher-order concurrency, where communication events (such as send, receive and non-deterministic choice) are first-class just like normal program variables (e.g., integers), that is, they can be created at run-time, assigned to variables, and passed to and returned from functions. This allows the construction of high-level concurrent features without explicit support for them in the model itself. We see how first-class events can be implemented in PVM, showing that it forms the base of a powerful concurrent-programming language. We also provide garbage-collection of processes, which simplifies selective communication and dealing with end-cases. The implementation promotes more widespread use of the flexibility present in the higher-order-concurrency model, and allows multiple processors to be exploited in such programs.

Length:

The paper is 17 pages and the talk is 25 minutes.

Availability:

The paper is available in PostScript (202k).

The talk is also available in PostScript (130k).

See information on file formats.

[Google Scholar search]

Related papers:

WoTUG20 (Higher-Order Concurrency in Java)

IPPS98 (Protocols for Non-Deterministic Communication over Synchronous Channels)

ProtocolsTR (Adaptive Protocols for Negotiating Non-Deterministic Choice over Synchronous Channels)