archive/jdk
1
0
Fork 0
mirror of https://github.com/openjdk/jdk.git synced 2025-09-17 17:44:40 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

8146987: Improve Parallel GC Full GC by caching results of live_words_in_range()

Browse source 
A large part of time in the parallel scavenge collector is spent finding out the amount of live words within memory ranges to find out where to move an object to. Try to incrementally calculate this value.

Reviewed-by: tschatzl, mgerdin, jmasa
This commit is contained in:
Ray Alex 2016-01-28 13:30:12 +01:00 • committed by Thomas Schatzl
parent ce491c9057
commit 4f42f17d9e
17 changed files with 180 additions and 81 deletions
Download patch file Download diff file Expand all files Collapse all files

4
hotspot/src/share/vm/oops/oop.inline.hpp
View file

@ -650,11 +650,11 @@ void oopDesc::pc_follow_contents(ParCompactionManager* cm) {
klass()->oop_pc_follow_contents(this, cm);
}
void oopDesc::pc_update_contents() {
void oopDesc::pc_update_contents(ParCompactionManager* cm) {
Klass* k = klass();
if (!k->is_typeArray_klass()) {
// It might contain oops beyond the header, so take the virtual call.
k->oop_pc_update_pointers(this);
k->oop_pc_update_pointers(this, cm);
}
// Else skip it. The TypeArrayKlass in the header never needs scavenging.
}