How IntelliJ IDEA Performance Improved: Past, Present, and Future
summary
Recap of concrete changes and results: startup drops blocking constructor-based components in favor of lazy service lookups and parallel classloading, raising startup MegaApdex 0.85 → 0.94; indexing becomes multithreaded with a recorded 25% multithreaded indexing speedup and disabling automatic Gradle source downloads yields ~50% faster indexing on Spring Boot; parsing JSON/YAML switches to Jackson parser: inspections 18s → 1.5s; completion ranking uses an ML model to improve relevance, and K2 integration shows K2: up to 4× faster completion in many cases. Tools: IGPerf, async profiler, OpenTelemetry-style metrics.
AI-written summary · about
chapters (18)
- 0:00Intro
- 0:47Metrics
- 6:09MegaApdex
- 11:48Startup
- 15:16Components and Dependency Injection
- 18:16Coroutines
- 20:48Classloading
- 23:38Indexing
- 27:39Actions available during indexing
- 30:03Completion
- 31:48Highlighting
- 37:46JSON Schema
- 41:40Freezes and Lags
- 46:11Import of projects
- 47:43Word on K2
- 49:34Tools
- 50:39Thank you
- 51:30Q&A
related talks
46mHow we’re improving performance of IntelliJ IDEA Kotlin plugin
23mKotlin 1.4 in IntelliJ-based IDEs: Boosting Quality and Performance
31mLet's focus on IDE Indexing
35mIntelliJ IDEA Tips & Tricks
45mAndroid Tools and Testing Fireside Chat
52mIntelliJ IDEA Tips and Tricks
06h 30mJetBrains Plugin Developer Conf 2025
01h 00mIndexing, or How We Made Indexes Shared and Fast
