RFR(S): 8181742: Load that bypasses arraycopy has wrong memory state
rwestrel at redhat.com
Fri Jun 16 09:15:17 UTC 2017
Thanks for reviewing this change.
> this looks good to me but can you elaborate a bit on why this also fixes 8179678?
With 8179678, the load from the arraycopy source is added after the
arraycopy so we have to make sure the arraycopy doesn't modify the
source element that is loaded otherwise the value loaded is
incorrect. With this change, the load is moved above the arraycopy
(i.e. happens before the arraycopy) so whether the arraycopy modifies
the source element or not becomes irrelevant.
> I think Vladimir should have a look as well. I'll run some pre-integration testing in the meantime.
More information about the hotspot-compiler-dev