| Issue | Fix Version | Assignee | Priority | Status | Resolution | Resolved In Build |
|---|---|---|---|---|---|---|
| JDK-8197696 | 10u-cpu | Thomas Schatzl | P4 | Resolved | Fixed | master |
| JDK-8200940 | 10.0.2 | Unassigned | P4 | Resolved | Fixed | b01 |
| JDK-8197517 | 10.0.1 | Thomas Schatzl | P4 | Resolved | Fixed | b05 |
| JDK-8197414 | 10 | Thomas Schatzl | P4 | Resolved | Fixed | b44 |
G1RemSet::refine_card_concurrently tests whether the region is in the collection set, skipping refinement if so, with an explanation that objects in the cset are about to move. But the cset is a collection pause artifact, and is clear when not in a GC pause. The test was useful when there was a single refine_card function that was used both concurrently and during a GC pause, but JDK-8177707 split the two cases into separate functions. It is no longer useful, and should be removed.
The hot card cache handling also has a comment mentioning that it doesn't need to check the cset, since hcc is only used during concurrent processing. That comment isn't needed anymore either.
Note that refine_card_during_gc already implicitly performs an in_cset check via the scan_top check. scan_top is initialized in part based on the collection set.
The hot card cache handling also has a comment mentioning that it doesn't need to check the cset, since hcc is only used during concurrent processing. That comment isn't needed anymore either.
Note that refine_card_during_gc already implicitly performs an in_cset check via the scan_top check. scan_top is initialized in part based on the collection set.
- backported by
-
JDK-8197414 G1RemSet::refine_card_concurrently doesn't need to check for cards in collection set
-
- Resolved
-
-
-
- Resolved
-
-
-
- Resolved
-
-
-
- Resolved
-