A VLSI architecture for enhancing software reliability
As a solution to the software crisis, we propose an architecture that supports and encourages the use of programming techniques and mechanisms for enhancing software reliability. The proposed architecture provides efficient mechanisms for detecting a wide variety of run-time errors, for supporting data abstraction, module-based programming and encourages the use of small protection domains through a highly efficient capability mechanism. The proposed architecture also provides efficient support for user-specified exception handlers and both event-driven and trace-driven debugging mechanisms. The shortcomings of the existing capability-based architectures that were designed with a similar goal in mind are examined critically to identify their problems with regard to capability translation, domain switching, storage management, data abstraction and interprocess communication. Assuming realistic VLSI implementation constraints, an instruction set for the proposed architecture is designed. Performance estimates of the proposed system are then made from the microprograms corresponding to these instructions based on observed characteristics of similar systems and language usage. A comparison of the proposed architecture with similar ones, both in terms of functional characteristics and low-level performance indicates the proposed design to be superior.