Self-Applicable online partial evaluation

Abstract

We propose a hybrid approach to partial evaluation to achieve self-application of realistic online partial evaluators. Whereas the offline approach to partial evaluation leads to efficient specializers and selfapplication, online partial evaluators perform better specialization at the price of efficiency. Moreover, no online partial evaluator for a realistic higher-order language has been successfully self-applied. We present a binding-time analysis for an online partial evaluator for a higher-order subset of Scheme. The analysis distinguishes between static, dynamic, and unknown binding times. Thus, it makes some reduce/residualize decisions offiine while leaving others to the specializer. We have implemented the binding-time analysis and an online specializer to go with it. After a standard binding-time improvement, our partial evaluator successfully self-applies. Our work confirms the practicality of an idea by Morry Katz.