COMPUTER OPERATING SYSTEM INSTRUCTION SET PRIMITIVES FOR ADA

THE ADA LANGUAGE HAS BEEN DESIGNED FOR THE DOD AS A PRIMARY LANGUAGE FOR EMBEDDED SYSTEMS. MANY FEATURES OF ADA TEND TO MAKE EXECUTION OPERATING SYSTEM (OS)-INTENSIVE. THESE ARE I/O, MULTITASKING, EXCEPTION HANDLING, TIMESLICING, AND DYNAMIC STORAGE MANAGEMENT. IN MANY SYSTEMS THESE FUNCTIONS ARE HANDLED BY THE OS COOPERATING WITH THE LANGUAGE RUNTIME SYSTEM (LRS). CALLS TO THE OS GENERALLY FORCE A SEQUENCE OF NESTED CONTEXT SWITCHES, WHICH IMPLIES A LARGE OVERHEAD ON OPERATIONS WHICH MAY BE SIMPLE IN THEMSELVES. EMBEDDED SYSTEMS FOR REALTIME CONTROL, SUCH AS AN INTELLIGENT AVIONIC SYSTEM MANAGER, REQUIRE INTENSIVE I/O OPERATIONS (SUCH AS MONITORING INSTRUMENTS) AND EVENT-DRIVEN MULTIPROCESSING. THERE ARE THREE OBJECTIVES FOR THE PROPOSED RESEARCH EFFORT: FIRST IS TO DEFINE COMPLEX INSTRUCTIONS FOR OS PRIMITIVES (CISC), COMPATIBLE WITH MICROCODED ARCHITECTURES, AND DESCRIBE THE NECESSARY SOFTWARE INTERFACE. THE SECOND IS TO DEFINE SIMPLE INSTRUCTIONS TO ENHANCE SOFTWARE-BASED SOLUTIONS, COMPATIBLE WITH COMPACT REDUCED INSTRUCTION SET ARCHITECTURES (RISC) WHICH IS MORE PRACTICAL IN SOME EMBEDDED SYSTEMS AND DESCRIBE THE NECESSARY SOFTWARE INTERFCE. THE THIRD IS TO IDENTIFY CASES WHERE THE OPERATING SYSTEM MUST INTERVENE IN EXECUTION, TO DEFINE THE LIMITS OF OUR REORGANIZATION. OUR ULTIMATE GOAL IS TO REMOVE AS MANY FEATURES AS POSSIBLE FROM THE OS DOMAIN. WE CANNOT HOWEVER REMOVE THEM ALL AND IT IS NECESSARY TO IDENTIFY THOSE WE CAN FOR SYSTEM OPTIMIZATION.