Ahead of Time Computation
summary
Project Leyden ships ahead-of-time class loading and linking in JDK 24 via JEP 483 and uses a three-step workflow: a training run that records loaded classes, an assembly phase that builds an AOT cache generation (now multi-threaded and measured in seconds), and deployment that reuses prelinked classes and lambdas. Leyden EA adds saved profiles and optional ahead-of-time compilation; results show 2.5–3× startup improvement for tiny framework samples and roughly 40% for a Spring Petclinic, with work ongoing to extend optimizations to user class loaders.
AI-written summary · about
chapters (7)
- 0:00Intro
- 0:40Leyden's Goals
- 4:49Leyden's Approach
- 10:17AOT Class Loading and Linking in JDK 24
- 14:09AOT Features in Leyden EA
- 17:04Some Performance Measurements
- 19:41AOT Feature Ideas
related talks
46mProject Leyden Update
46mAssembling Project Leyden
41mSupercharge your JVM performance with Project Leyden
51mSupercharge your JVM performance with Project Leyden
48mProject Leyden - Capturing Lightning in a Bottle
45mProject Leyden: Improving Java’s Startup Time
45mSupercharge your JVM performance with Project Leyden and Spring Boot
42mSupercharge your JVM performance with Project Leyden and Spring Boot