Details
Description
We use "passive" for the differential performance analysis against concurrent GC. There is no reason to enable pacing when all GC cycles are STW.
Disabling pacing automatically gives us a few benefits:
a) "Passive" becomes even more passive;
b) Avoids unnecessary stalls for rendezvous with control thread, which might be the diagnosable source of stalls;
c) Lists pacing options as ergonomically set with "passive", so users know how "passive" is different;
d) Tests -ShenandoahPacing paths in the course of "passive" configuration testing.
Disabling pacing automatically gives us a few benefits:
a) "Passive" becomes even more passive;
b) Avoids unnecessary stalls for rendezvous with control thread, which might be the diagnosable source of stalls;
c) Lists pacing options as ergonomically set with "passive", so users know how "passive" is different;
d) Tests -ShenandoahPacing paths in the course of "passive" configuration testing.
Attachments
Issue Links
- relates to
-
JDK-8233387 Shenandoah: passive mode should disable pacing ergonomically
-
- Resolved
-