17.2.1 Introduction to Fair Threads

Fair threads are cooperative threads run by a fair scheduler which gives them equal access to the processor. Fair threads can communicate using broadcast events. Their semantics does not depends on the executing platform. Fine control over fair threads execution is possible allowing the programming of specific user-defined scheduling strategies.

Contrary to standard sequential programming where the processor executes a single program, in concurrent programming the processor is a shared resource which is dispatched to several programs. The term concurrent is appropriate because programs can be seen as concurrently competing to gain access to the processor, in order to execute.

Threads are a basic mean for concurrent programming, and are widely used in operating systems. At language level, threads offer a way to structure programs by decomposing systems in several concurrent components; in this respect, threads are useful for modularity.

However, threads are generally considered as low-level primitives leading to over-complex programming. Moreover, threads generally have loose semantics, in particular depending on the underlying executing platform; to give them a precise semantics is a difficult task, and this is a clearly identified problem to get portable code.

Bigloo proposes a new framework with clear and simple semantics, and with an efficient implementation. In it, threads are called fair; basically a fair thread is a cooperative thread executed in a context in which all threads always have equal access to the processor. Fair threads have a deterministic semantics, relying on previous work belonging to the so-called reactive approach.

This document was generated on October 23, 2011 using texi2html 5.0.