SHF: Small: RUI: Characterizing, Detecting, and Fixing Performance Bugs That Have Non-Intrusive Fixes

Florida State University, Tallahassee FL

Investigators

Abstract

Software performance is critical for the success of a software project. Performance bugs are programming errors that slow down execution. Many recent techniques have been proposed to detect various performance bugs. However, there are still many performance bugs that cannot be detected by existing techniques. Furthermore, a crucial and practical aspect of performance bugs has not received the attention it deserves: How likely are developers to fix a detected performance bug? To significantly improve software performance, this project will develop a set of novel techniques that focus on a class of performance bugs that are very likely to be fixed by developers, specifically on performance bugs that have non-intrusive fixes. Performance bugs that have non-intrusive fixes are very likely to be fixed by developers because the benefits of the fix (i.e., code speedup) clearly outweigh the drawbacks of the fix (e.g., introducing new correctness bugs, breaking good software engineering practices, development time and effort, etc). This project will address three fundamental challenges: (1) What performance bugs have non-intrusive fixes and what are their defining characteristics? (2) How to automatically detect performance bugs and how to establish that their fixes will be non-intrusive? (3) How to automatically fix some of the detected performance bugs? This work will broaden our understanding of an understudied yet important aspect of software performance and will provide automated solutions to improve software performance.