Abstract

As increasingly complex real-time systems are created, program representations must improve to meet the demands of more sophisticated programs. In realtime systems, when an answer is produced is part of its correctness. Proper run-time management in a real-time system thus depends on accurately predicting program execution behavior, which must be based on information either specified in the program source or derived from it during translation. This paper discusses our method for deriving behavioral predictions while translating between the programming and run-time representations used by the Spring system. A graph representation of the program is derived from the intermediate representation the compiler uses to emit code. This graph is then reduced and analyzed to make behavioral predictions. We present the basic translation method, and give examples of how it translates programming language constructs for critical sections and synchronous communication.

Citations