[13] RFR (S): 8218163: C2: Continuous deoptimization w/ Reason_speculate_class_check and Action_none
Vladimir Ivanov
vladimir.x.ivanov at oracle.com
Thu Jan 31 19:46:08 UTC 2019
http://cr.openjdk.java.net/~vlivanov/8218163/webrev.00/
https://bugs.openjdk.java.net/browse/JDK-8218163
Guarded inlining doesn't take recompilation count into account when
decided whether to issue an uncommon trap or virtual call on slow path.
But it may end up as uncommon trap with Action_none when recompilation
count is too high.
The fix is to consider both trap & recompilation counts when making the
decision.
Testing: hs-precheckin-comp, tier1-5
Best regards,
Vladimir Ivanov
More information about the hotspot-compiler-dev
mailing list