Exploiting the data parallelism of subset equational languages

Abstract

The authors propose an alternative approach to the parallel implementation of declarative languages, the data parallel approach. The focus is on large collections of data and the core idea is to parallelize the execution of elementwise operations. The target language is SEL, a subset equational language. An abstract machine for the language, the SAM (subset abstract machine) is outlined, which from a certain point of view belongs to the WAM family. The data parallel structure of the SAM and of its instruction set is discussed, and some examples are given.<>