The Renaissance for Big Data and Parallelism with GraalVM
summary
A new benchmark collection targets modern parallel JVM workloads and ships with Renaissance suite (21 benchmarks) that measure synchronized blocks, atomic ops, parking, invokedynamic and cache misses. Results drive real optimizations — e.g. escape analysis for atomics, loop‑wide and invokedynamic simplifications — and measured a >6% geomean speedup on the Renaissance set and ~3% on older suites. The work informs compiler, GC and tooling changes used to improve GraalVM Community Edition ~6% vs OpenJDK and Enterprise edition deployments.
AI-written summary · about
chapters (12)
- 0:00Start
- 4:24The Renaissance benchmark suite
- 4:52On benchmark suites...
- 5:55Time does not wait for anyone...
- 8:20Renaissance Benchmark Suite (1)
- 14:54Analyzing diversity
- 16:11Principal Component Analysis
- 18:41Optimization opportunities
- 20:09Impact of individual optimizations in GraalVM
- 22:06Compiler optimizations: conclusions
- 22:41Performance of GraalVM vs HotSpot (C2)
- 24:44Renaissance Summary
related talks
50mGraalVM for Java developers
47mGraalVM for Endless Performance
37mTurning GraalVM from Research to Product
59mGraalVM: Run Programs Faster Anywhere
52mGraalVM: Run Programs Faster Anywhere
03h 07mEverything you need to know about GraalVM
49mGraalVM: Run Programs Faster Anywhere
40mThe need for speed. What is GraalVM and how can it help you?