How JDK 26 Improves G1's Throughput
summary
JDK 26 reduces G1 write-barrier overhead by replacing fine-grained per-write synchronization with two card tables that the app threads mark and a concurrent scanner consumes via a handshaking switch, eliminating expensive atomic work on each write. The change comes from JEP 522, trades about 0.2% memory overhead (one byte per ~512 bytes) for measurably higher throughput, removes card-queue bookkeeping, and yields better default G1 performance without configuration.
AI-written summary · about
chapters (3)
related talks
43mGarbage Collection in Java: The Performance Benefits of Upgrading
51mJava Performance Update: From JDK 21 to JDK 25
41mJava 24, Faster Than Ever
53mJava Performance Update 2024
01h 01mOpenJDK: In the new age of Concurrent Garbage Collectors
50mGarbage Collection in Java: The progress since JDK 8
46mMastering GC: tame the beast and make it your best ally
48mJava 9 Performance