A multiprocessor real-time Ada-based environment is becoming increasingly necessary to support the growing class of complex simulation and training applications. The Bare Machine Ada runtime, or BMA, project at Gould C.S.D. is an effort to produce such an environment. The results of this development effort are providing valuable insights into the mechanisms required to adequately and efficiently support Ada on multiple processors.

Variations of a distributed real-time environment are evaluated with respect to very tightly coupled processors running with a single (common) local memory, tightly coupled processors equipped with a range of common memory, and loosely coupled processors containing no common memory. Finally, we consider methods for obtaining a processing system able to distribute Ada tasks freely over multiple processors. Each processor in the system maintains its own local memory, but entire ranges of local memories shadow each other's contents, forming a single common address range with a minimum of contention.