• Explore
    • Contact Us
  • Faculty
  • Research
    • Research Areas
    • Research Centers
  • Graduate Degrees
    • Computer Science Programs
    • Current Graduate Students
  • Undergraduate Degrees
  • News & Events
    • News
    • Seminar Series
    • Distinguished Lecture Series
    • Research Showcase
  • Apply Now
    • Undergraduate Admissions
    • Graduate Admissions
    • Faculty Candidates

Xu tackles software bloat for Office of Naval Research

March 23, 2016

Assistant Professor of Computer Science Harry Xu is developing a novel approach to counter software runtime redundancies and performance-blocking inefficiencies—known as “software bloat”—thanks to funding from the Office of Naval Research.


“Much evidence shows that seemingly harmless performance problems can lead to severe scalability reductions and financial losses. Even mature software written by expert engineers has had performance problems that caused serious and highly publicized incidents,” Xu says. “In fact, the levels of inefficiencies in real-world software grow surprisingly fast. There is an ever-increasing demand for performance optimization in modern software, especially when processor-core clock speed has reached a practical limit and the performance ‘free lunch’ the software community has enjoyed so far from hardware advances will not exist anymore.”

In response, Xu has contributed to state-of-the-art static and dynamic program analysis techniques that can identify bloat patterns in program source code or during execution. Xu notes, “It remains largely unknown whether and how the penalty of over-general implementations can be automatically optimized away and at what level techniques should be developed.” He proposes instead an optimization technique called Library Auto-Selection (LAS).

LAS combines both the developer’s insights and compiler and runtime system support to automatically remove inefficiencies in software. It targets a broad class of performance issues in object-oriented programs resulting from generalized implementations of libraries. Acording to Xu, using developers’ domain insights and runtime support in tandem, LAS will allow users to “dynamically determine the optimal implementation for a data structure or algorithm” and “safely switch to this new implementation at run time for increased efficiency.”

« Harris organizing design automation workshop in Germany
Circuit Cellar: “The Future of IoT Security” by Ian Harris »

Latest news

  • Wired: “Google Has a Plan to Stop Its New AI From Being Dirty and Rude” (Sameer Singh quoted) May 12, 2022
  • NSF Announces 2022 Graduate Research Fellows May 2, 2022
  • Alumni Spotlight: Ludovico Verniani ’21 Rewards Adventure With Quadra App April 28, 2022
  • UCI Celebration of Teaching Honors Three ICS Faculty Members April 26, 2022
  • UCI Machine Learning Repository to Host 2022 ML Hackathon April 26, 2022
  • © 2022 UC Regents
  • Feedback
  • Privacy Policy