Data compatibility in a heterogeneous processor environment

Abstract

Because the Ada language does not require a particular representation of data at the machine level, Ada compilers targeted to different computer architectures can and do generate different object sizes and composite object layouts (i.e., field ordering and alignment) for the same type declarations. This, coupled with the possibility of differing floating point representations or inconsistent treatment of integer byte sex, presents a data compatibility problem for distributed applications targeted to a heterogeneous processor environment.

This paper describes three techniques that have been used to address the problem of data compatibility on the Advanced Automation System (AAS), a large air-traffic control application being developed in Ada, and discusses trade-offs and difficulties associated with each. The three techniques are representation clauses, a type dictionary service, and a portable character-based data representation.