RFR(M): 8204615: [lworld] C2 support for java.lang.Object methods on value types
tobias.hartmann at oracle.com
Fri Jul 6 07:22:41 UTC 2018
On 05.07.2018 20:56, Roland Westrelin wrote:
> The monitorenter debug info doesn't include the object being locked so
> that can't work.
Yes, I've tried again and I get all kinds of crashes. I think we could make it work by changing the
debug info for monitorenter but it's definitely more complex than my webrev.00.
We already deoptimize with a pending exception for other runtime calls, for example,
OptoRuntime::new_instance_C -> deoptimize_caller_frame. The difference here is that monitorenter is
usually not expected to throw an exception in the slow path.
Since this is one of the last missing pieces for LW1, I would like to push the patch and enhance it
later if we find a better way. What do you think?
More information about the valhalla-dev