RFR[T]: 8242003: Remove CallInfo::_selected_klass

John Rose john.r.rose at oracle.com
Wed Apr 1 20:13:06 UTC 2020

On Apr 1, 2020, at 10:48 AM, coleen.phillimore at oracle.com wrote:
> This looks good to me.  I looked through the history as far back as ~2007 and it's never been used.

That’s a head-scratcher for me too.  The whole idea of performing
selection during link resolution is painful to think about.  I think
there are corner cases where the JVM chooses to re-run link logic
on every call, where this might help.  There’s also (IIRC) a possible
connection to the management of inline caches, where in some
sense the first call “wins” over all subsequent calls.

I have an old SCCS snapshot which shows a use of selected_klass
in opto/runtime.cpp.  It was removed in 1998, when a distinction
between resolve_opt_virtual and resolve_virtual was introduced.
The deleted code was handling the case of an “optimized” (CHA
limited, effectively final) virtual

— John

More information about the hotspot-runtime-dev mailing list