Context parallelism in an indexical programming language

Indexical programming languages enrich conventional programming languages by allowing the meanings of program constructs to depend on an implicit context. In an indexical program, values at different contexts can potentially be evaluated an parallel, resulting in context parallelism. This paper formally defines context parallelism in the indexical language mLucid, by a variable-value dependency graph (VDG) derived from the denotational semantics of the language. Based on different abstractions of the VDG, three levels of context parallelism: fine, medium, and coarse parallelism are defined, which reflect different levels of parallel computations specified by the indexical program.<
>