Benefits of type inference for an object-oriented real-time language

Abstract

RealTimeTalk is a framework and an object-oriented language for distributed hard real-time systems. This paper presents past, present, and future work in the RealTimeTalk project. It concentrates on the language and programming part of RealTimeTalk, and the problems a dynamically typed language impose on a real-time system.We look at how some of these problems can be solved by type inference and how type information can assist the compiler in accepting larger classes of real-time programs, generate more efficient and deterministic code, provide better estimation of execution times, and avoid run-time type errors. With type inference, one can have the best of two worlds -efficiency and type safety, without the need for manual type declarations. An example is used to illustrate these ideas.

An Overview of RealTimeTalk, a Design Framework for Real-Time Systems

“…Instead, we are currently in the process of designing a type inference system for RTT [10]. The purpose of this system is to annotate each variable occurrence in the program with a type.…”

“…This in turn enables other optimizations, e.g., inlining. Efficiency can sometimes also be improved even if the receiver could be an instance of more than one class [10].…”