A programming environment for real-time multi-threaded systems.

Abstract

In recent years, the rapid progress in the functionality, performance, and cost of VLSI technology has made multiprocessor computer platforms feasible and cost-effective. The key benefit of these platforms is their ability to offer true concurrency at the physical layer, through the replication of processor, memory and I/O facilities, and at the logical layer, through the support for multi-tasking and multi-threading. For many applications, such as those found in the real-time problem domain, the concurrent nature of these platforms bring numerous advantages. To this date however, because of the lack of real-time concurrent programming tools, delivering the potential benefits of multiprocessing directly to these applications has proven elusive. This thesis proposes the Multi-Activity Programming Environment (MAPE) which allows developers to exploit the concurrent capabilities of multiprocessor platforms while remaining consistent with the requirements of the real-time problem domain. MAPE provides a programming model and a set of supporting tools which facilitate the decomposition of an application into cooperating entities and allow for the specification of deadline-based time constraints. The MAPE components also support the allocation of the platform's resources and the modification of the application's functionality based on run-time urgency considerations.