Abstract

A Multi-threaded application is one which is capable of performing several tasks concurrently. This work describes the design and issues relating to the implementation of a user-level thread facility for an interactive, persistent language for statistical data analysis, S. We outline some of the possible uses of this threaded environment, including parallel and distributed computing. Also, we examine the benefits of using this framework. We describe the interface to the user level threads and the design of the internal structures that control the run-time behavior of these threads.

Finally, an algorithm is described which facilitates the develpoment of correctly synchronized applications and prototypes for other languages.

We proceed in chapter II to describe the application programming interface (API) to the thread primitives and provide examples illustrating how higher level structures and functions can be constructed and how applications might utilize these functions. Chapter III describes how much of the complexity associated with developing multi-threaded code can be eliminated by using tools to identify potential synchronization problems. The remainder of the document describes the more technical design and implementation of the threading capabilities in the application code space (rather than the language definition). The ideas presented here have been implemented in the newest version of the language S created by Lucent Technologies Bell Labs Innovations. We conclude the thesis by discussing the generalizability of the approach to other similar languages.


Duncan Temple Lang<duncan@stat.Berkeley.EDU>
Last modified: Thu Jun 12 08:51:41 PDT 1997