Exploiting parallelism in primitive operations on bulk data types

Abstract

Structures based on tuples may be used to represent the common bulk data types found in symbolic applications. Few languages support primitive operations that handle such structures explicitly, and fewer support them as persistent objects. Even when relevant data types are accommodated at the language level, implementations are very inefficient because of the lack of underlying systems architecture support. The result is over-complex software and missed opportunities to exploit the inherent data-parallelism present in these structures. In this paper we propose a representational framework, a repertoire of primitive operations, and a systems architecture for handling the bulk data types relevant to (deductive) databases and knowledge-based applications. We indicate the opportunities for exploiting inherent parallelism, and briefly report on prototype novel parallel hardware which supports the structure primitives.