Adaptive static analysis for high-level languages applied to abstract profiling of programs

点击翻译按钮获取中文摘要

The main axis of the proposed research consists in continuing research on adaptive static analyses.While the exact set of abstract entities used by a conventional static analysis is selected at implementation time by the developer of the compiler, that of an adaptive static analysis can be changed at compile time.More than that, an adaptive static analysis is able to determine by itself relevant changes that ought to be applied to the abstract entities.An adaptive static analysis has the potential to be very precise thanks to its ability to select a set of abstract entities that fit the program at hand while keeping the costs under control by avoiding superfluous increases in precision.The main project consists in performing static profiling of functional programs.Static profiling has strong similarities with static analysis in that the particular choice of abstract entities has a great influence on the cost and the precision of the analysis.We intend first to realize a working version of static profiling and then to apply the idea of adaptiveness to improve the accuracy of the results produced using static profiling.The research that we have conducted in the past and the research that we intend to conduct contribute to improve the implementation of high-level programming languages, and that of functional languages in particular.