LoopOracle : Speculative hotspot detection using machine learning in web browsing environments
Author(s)
Startsev, Julia
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
The performance of modern Just-In-Time (JIT) compilers, particularly in Web Virtual Machines (WebVMs) like SpiderMonkey, depends heavily on efficient profiling and optimization strategies. This thesis explores predictive and adaptive techniques to reduce profiling overhead, with a focus on loop speculation and machine learning-based methods. By addressing challenges in tiering decisions and loop prediction, this work aims to enhance performance, reduce delays, and provide a foundation for more efficient JIT compilation in dynamic browser environments.
Sponsor
Date
2025-05-14
Extent
Resource Type
Text
Resource Subtype
Thesis (Masters Degree)