Parallelism capabilities are becoming ubiquitous thanks to the widespread use of multi-core processors. This has renewed the interest in language-related designs and tools which can simplify the task of producing parallel programs. The use of declarative languages is considered to be a valid approach for increasing performance through the execution of parallel programs. In particular, nondeterminism and partially instantiated data structures give logic programming expressive power beyond that of functional programming. However, functional programming often provides convenient syntactic features, such as having a designated implicit output argument, which allow function call nesting and sometimes results in more compact code, and also sometimes a more direct encoding of lazy evaluation, with its ability to deal with infinite data structures. The high-level nature of these languages, in addition to their relatively simple semantics and the use of logic variables, preserves more of the original parallelism to be uncovered by an automatic parallelization. Different alternatives for performing automatic goal-level, unrestricted independent and-parallelization of logic programs through source-to-source transformations are studied in this work, which use as targets new parallel execution primitives which are simpler and more flexible than the well-known fork-join parallel operator, in order to generate better parallel expressions by exposing more potential parallelism among the literals of a particular program. An alternative approach for implementing and-parallel logic programming languages has also been explored, which tames the complexity of the low-level machinery required by most of the previous implementations by raising core parts to the source language level. A significant portion of the implementation mechanisms of parallel execution is handled directly at the Prolog level with the help of a comparatively small number of concurrency-related primitives that take care of simpler low-level tasks such as locking, and thread and stack set management. Moreover, in order to extend this work to different paradigms, a syntactic functional extension to ISO-standard Prolog systems has been developed, which covers function application, predefined evaluable functors, functional definitions, quoting, and lazy evaluation, and is composable with higher-order and other extensions to ISO-Prolog, such as constraints.
展开▼
