Timeline for BiasedLocking removal?
Roman Kennke
rkennke at redhat.com
Tue May 11 13:50:00 UTC 2021
Hi Hotspot devs,
what is the timeline for removing BiasedLocking altogether? Or have we
even agreed that we are going to remove it?
I am asking because this affects how I'd move forward with Lilliput. If
I can ignore BiasedLocking because it'd be removed in, say, JDK 18
anyway, then I believe I can come up with a relatively simple
implementation for 64bit header that should be possible to integrate
into JDK 18. That is, I put the (compressed) Klass* in the upper bits of
the header, also keep the hashcode in the header, and keep everything
else (locking, GC bits) as they are.
With biased-locking, the Klass* would clash with the Thread* in the
upper bits. If we are not going to remove BiasedLocking anytime soon, I
would have to come up with a different plan.
(Also, note that I expect the performance improvement of Lilliput to
outweight possible - and hopefully rare - biased-locking removal
regressions)
WDYT?
Cheers,
Roman
More information about the hotspot-dev
mailing list