Reproduce “OOPredictor: Predicting Object-Oriented Accesses using Static Analysis” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Hassan Arafat, David Bremner, Kenneth B. Kent, Julian Wang
The OOPredictor framework predicts object-oriented memory access patterns using compile-time static analysis and Markov chains within the OpenJ9 JVM, addressing performance bottlenecks caused by pointer chasing without runtime profiling overhead. It demonstrates high prediction accuracy for a substantial portion of methods across various workloads.
Key insights
- Compile-time static analysis within a JIT compiler can effectively predict object-oriented memory access patterns for a significant portion of methods, offering a profile-free alternative to dynamic approaches.
- Markov chains provide a robust model for probabilistic transitions of field accesses, and compression techniques help manage the complexity of these static models.
- Affinity graphs derived from static predictions show strong structural and relational similarity to those built from runtime profiling, indicating their utility for guiding object and field layout optimizations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.