Comparing Rapid Type Analysis with Points-To Analysis in GraalVM Native Image
summary
GraalVM Native Image replaces heavyweight points-to analysis with a prototype parallel, incremental RTA that integrates heap snapshotting and uses serialized method summaries plus atomic checks and task scheduling to scale across cores. The prototype cuts RTA analysis time on Spring Petclinic by 64% and total build time by 35% while increasing image size ~15% and marking more methods reachable; it wins on medium/large apps but can slow tiny programs.
AI-written summary · about
chapters (15)
- 0:00GraalVM Native Image intro
- 1:24Call graph challenges
- 3:09Motivation for Rapid Type Analysis alternative
- 4:12Call graph methods overview
- 4:55How RTA works with equations
- 6:23Paper contributions
- 7:04RTA core algorithm
- 8:28RTA implementation summaries
- 8:50Resolving virtual methods
- 11:19RTA simple example
- 12:26Incremental analysis
- 13:51Incremental analysis challenges
- 14:31Experimental results
- 15:54Conclusion and future work
- 16:37Q&A
related talks
55mNative Image: Large-scale static analysis for Java
20mScaling Type-Based Points-to Analysis with Saturation
29mGraalVM in a Nutshell
26mAn Analysis of Compiled Code Reusability in Dynamic Compilation
33mNew Opportunities for Developers with GraalVM
23mAdopting GraalVM
36mThe Call of Ctooling: The Secrets Behind Native Image Building
46mGraalVM Native Image — Faster, Smarter, Leaner