Adaptive Optimization in the Jalapeño JVM: The Controller's Analytic Model

Authors: Matthew Arnold, Stephen Fink, David Grove, Michael Hind, and Peter F. Sweeney.
Citation: 3rd ACM Workshop on Feedback-Directed and Dynamic Optimization (FDDO-3)

This paper provides details of the component of the Jalapeño adaptive optimization system that determines what methods to optimize. This component, called the <em>controller</em>, can choose from one of several optimization levels. In the current implementation, the controller uses a simple cost/benefit analysis to drive adaptive compilation decisions. It has been demonstrated that even this simple analytic model can achieve reasonable performance compared to various JIT compilation scenarios in both startup and steady-state program regimes.

This paper outlines several open questions in developing a more accurate controller model. We present two experiments that study the effects of how the current model predicts future execution from the past, a limited experimental evaluation of stability of the current model across applications, and describe our ongoing efforts to improve the Jalapeño controller.

PDF