Compiler optimizations for an asynchronous stream-oriented programming language

Abstract

Stream-oriented programs allow different opportunities for optimization than procedural programs. Moreover, as compared to purely synchronous stream-oriented programs, optimizing for asynchronous stream-based programs is difficult, owing to the latters' inherent unpredictability. In this thesis, we present several compiler optimizations for WaveScript, a high-level, functional, stream-oriented programming language. We also present a framework for using profiling of stream-graph execution to drive optimizations; two of the optimizations use this profiled information to generate noticeable performance benefits for real-world applications written in WaveScript. Thus, it is shown that profiling presents an important avenue by which to optimize asynchronous stream-based programs.