Data Parallel Programming in Scala
Bruce Lester

To more fully utilize the potential offered by multi-core processors, programming languages must have features for expressing parallelism. Data parallel programming features allow high-level collection-oriented operations to be easily expressed by the programmer, and then implemented by the runtime system in a parallel fashion for improved performance. This paper describes a data parallel class library, for the Scala language. This data parallel library includes a new type of data structure called a Parallel Vector, and fifteen basic operations on this data structure. The library and its implementation in Scala have several novel aspects including Where-Masks for conditional execution and vertical integration of data parallel operations in the runtime library implementation.