Review Request (M) 8032379: Remove the is_scavenging flag to process_strong_roots (original) (raw)
Mikael Gerdin mikael.gerdin at oracle.com
Tue Jan 21 13:13:25 UTC 2014
- Previous message (by thread): JDK7u45 Young GC pause time increasing with increasing total heap, but it is not the case with JDK6
- Next message (by thread): Review Request (M) 8032379: Remove the is_scavenging flag to process_strong_roots
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Hi all,
as a part of implementing JEP-156 (G1 class unloading) we are doing some cleanups in order to refactor the strong root processing code.
Currently there are two "dimensions" in which SharedHeap::process_strong_roots is configured. One is the ScanOption enum, the other is the is_scavenging booelan. The semantic meaning of is_scavenging can easily be folded into the ScanOption enum by:
- Introducing a SO_AllCodeCache/SO_ScavengeCodeCache distinction (in a way similar to SO_AllClasses/SO_SystemClasses)
- Noting that passing a CLD closure to Threads::oops_do is only needed when we want to determine the precise liveness of classes, this is already signaled by SO_SystemClasses.
Bug link: https://bugs.openjdk.java.net/browse/JDK-8032379 Webrev: http://cr.openjdk.java.net/~mgerdin/8032379/webrev.0
Thanks /Mikael
- Previous message (by thread): JDK7u45 Young GC pause time increasing with increasing total heap, but it is not the case with JDK6
- Next message (by thread): Review Request (M) 8032379: Remove the is_scavenging flag to process_strong_roots
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]