AOT or JIT: Faster Startup or Faster Code?
summary
Compares AOT and JIT with concrete numbers and tools: Graal native-image gives up to 50x faster startup and ~5x smaller footprints but loses aggressive inlining and speculative optimizations and is platform specific; JIT gains much higher steady throughput because speculative optimizations deliver roughly half of JIT’s benefit; Azul’s approach uses ReadyNow profiles, a compile stashing cache and the LLVM-based Falcon compiler to hit ~98% warmed performance immediately while cutting compile time up to 80% and CPU compile load up to 60% and avoids long GC pauses with C4 pauseless GC.
AI-written summary · about
chapters (15)
- 0:00Intro
- 1:17JVM Performance Challenges
- 4:46Microservice-Based Application
- 9:41Microservice Performance Challenges
- 12:53Java Compilers: Ahead of Time (AOT)
- 18:54Speculative Optimisations
- 22:30GraalVM High performance runtime for polyglot applications
- 25:22Graal VM For Java Applications
- 27:15The Azul Prime JVM
- 28:23ReadyNow! Solution
- 31:19ReadyNow! Start-up Time
- 33:02Falcon Pipeline
- 33:53Deterministic Compiler
- 35:36Compile Stashing Effect
- 36:46Faster Startup or Faster Code
related talks
50mAOT or JIT: Faster Startup or Faster Code?
27mHow to Eliminate Java Performance Warmup
55mJIT and AOT in the JVM
56mBaselining AArch64 Java AOT Performance
51mAOT: Everything you need to know about GraalVM for Java applications
51mGoing AOT: Everything you need to know about GraalVM for Java applications
01h 03mKeeping Your Java Hot.
46mAhead Of Time (AOT) Internals