-
Enhancement
-
Resolution: Fixed
-
P3
-
9, 11, 17, 21, 22
-
b20
| Issue | Fix Version | Assignee | Priority | Status | Resolution | Resolved In Build |
|---|---|---|---|---|---|---|
| JDK-8342877 | 21.0.7-oracle | Ivan Walulya | P3 | Resolved | Fixed | b01 |
| JDK-8335293 | 21.0.5 | Xiaolong Peng | P3 | Resolved | Fixed | b01 |
| JDK-8335343 | 17.0.13 | Xiaolong Peng | P3 | Resolved | Fixed | b01 |
http://mail.openjdk.java.net/pipermail/shenandoah-dev/2016-December/001275.html
In short, returning MinTLABSize from unsafe_max_tlab_alloc when there is no space in the region makes the allocation mechanics to allocate MinTLABSize-sized TLAB in the next free region, fragmenting it. Shenandoah fixed this by returning the size of the next region, which we know is free (http://hg.openjdk.java.net/shenandoah/jdk9/hotspot/rev/7009fc6f74b3), not sure the same can be done for G1.
- backported by
-
JDK-8335293 G1: Returning MinTLABSize from unsafe_max_tlab_alloc causes TLAB flapping
-
- Resolved
-
-
-
- Resolved
-
-
-
- Resolved
-
- links to
-
Commit
openjdk/jdk17u-dev/10701005
-
Commit
openjdk/jdk21u-dev/6b4fa865
-
Commit
openjdk/jdk/4c79e7d5
-
Review
openjdk/jdk17u-dev/2639
-
Review
openjdk/jdk21u-dev/774
-
Review
openjdk/jdk/16102