Baselining AArch64 Java AOT Performance
summary
Comparison uses jaotc (Graal) AOT front-ending OpenJDK HotSpot across microbenchmarks and SpecJVM startup test, measuring 100 loop warmups and 15s runs. Results are mixed: some AArch64 cases outperform x86 for specific hot-path optimizations, but many spec workloads show tiered-AOT slower or up to ~62% worse versus non-AOT because loading .so files and startup overhead cancels native startup gains. AOT avoids interpreter fallback on deoptimization and cuts background JIT threads, but benefits depend on which methods are AOT-compiled and code-cache/memory behavior.
AI-written summary · about
chapters (8)
- 0:00Introduction and background
- 1:13What is Java AOT
- 8:05AOT compilation methods
- 10:16Initial five benchmarks
- 28:55SPEC JVM benchmarks
- 39:04Startup tests analysis
- 42:18Real-world AOT benefits
- 45:33Q&A session
related talks
38mAOT or JIT: Faster Startup or Faster Code?
55mJIT and AOT in the JVM
41mJava Goes AOT
50mAOT or JIT: Faster Startup or Faster Code?
46mAhead Of Time (AOT) Internals
27mHow to Eliminate Java Performance Warmup
51mGoing AOT: Everything you need to know about GraalVM for Java applications
51mAOT: Everything you need to know about GraalVM for Java applications