Design of a strongly typed parallel programming language based on CSP

Hoare's CSP espouses a very simple programming model, that of combining sequential processes in parallel so that they cooperate in the execution of a task, communicating values whenever required. Since its introduction in 1978, CSP has undergone major modifications, and has also been given an elegant and reasonably sound mathematical semantics based on the failures-divergence model. However, whereas the original version of CSP was offered as a partial language proposal, the current version is more of a calculus for specifying and reasoning about the observable (possibly nondeterministic) behaviour of processes. This is due in large part to three factors: (1) some of CSP's combinators for process composition are too abstract and powerful, and may have to be either modified or left out altogether; (2) there is no well-defined type system; (3) the syntax is not conducive to large-scale programming. This thesis presents the design of Impala, a strongly-typed imperative parallel programming language based on CSP. Impala, an acronym for 'Imperative Parallel Language', demonstrates that with care it is possible to produce a high-level process-oriented language that is intuitive and powerful, yet simple and practical. It is, in the first instance, suitable for general-purpose use, and achieves maximum functionality with minimum complexity and confusion. Also included are a summary. of CSP's failures-divergence semantics, and a preliminary report on Impala.