SISAL 1.2: high-performance applicative computing

The acquisition of parallel processors in the scientific community is increasing, but the difficulties of parallel programming persist. Three approaches have emerged: automatic parallelizing compilers for extant languages; extended languages; and new languages that provide a cleaner and easier-to-use parallel programming model. One such new language is SISAL 1.2, a general-purpose applicative language. Regrettably, applicative languages have acquired a reputation for inefficiency because of their single-assignment semantics, and dynamic creation of aggregate objects. The authors show that a set of powerful yet simple optimization techniques can reduce the overhead of applicative semantics without sacrificing parallelism. Optimized SISAL codes can achieve execution speeds comparable to FORTRAN, and effectively exploit shared-memory multiprocessors