Applicative parallelism on a shared memory multiprocessor

The authors discuss how software for applicative language programs can automatically detect and exploit parallelism to take advantage of current and future parallel processing systems. They use an applicative language, SISAL, to show the sources of parallelism. A prototype SISAL system is available to execute some sample programs and help examine how well the implementation uses potential parallelism. An outstanding problem in applicative language design and implementation is copy optimization. The single assignment semantics of these languages requires frequent data structure copying in some programs, The authors exemplify the importance of minimizing copying, and show that other optimizations, including high-quality code generation, are also important for applicative languages.< >