NESL: A Nested Data-Parallel Language

摘要: This report describes NESL, a strongly-typed, applicative, data-parallel language. NESL is intended to be used as portable interface for programming variety of parallel and vector supercomputers, basis teaching algorithms. Parallelism supplied through simple set constructs based on vectors, including mechanism applying any function over the elements in rich functions that manipulate vectors. fully supports nested vectors parallelism--the ability take apply it multiple instances parallel. Nested parallelism important implementing algorithms with complex dynamically changing data structures, such required many graph sparse matrix also provides calculating asymptotic running time program various machine models, random access (PRAM). useful estimating times actual machines and, when algorithms, supplying close correspondence between code theoretical complexity. defines several examples coded The include median finding, sorting, string searching, finding prime numbers, planar convex hull. currently compiles an intermediate language called VCODE, which runs Cray Y-MP, Connection Machine CM-2, Encore Multimax. For current implementation gives performance optimized machine-specific these machines.