RFR: 8290892: C2: Intrinsify Reference.reachabilityFence [v8]

[Emanuel Peter]

On Tue, 9 Sep 2025 21:51:31 GMT, Vladimir Ivanov <vlivanov at openjdk.org> wrote:

>> I'm also not sure yet why there is a difference between incremental inlining and regular inlining.
>> Do you think it would make sense to explain that here, or is it explained elsewhere?
>
> There are no safepoint-attached reachability edges present during normal parsing. For incremental inlining, JVMS from the original call is taken and extended with callee state. If there are reachability edges present, they have to be treated specially and carried over to all safepoints produced during incremental inlining attempt. There's no such support in place yet.

@iwanowww Ok, sounds a bit complicated. Maybe that is what we have to do, at least for now. But please make sure that this is documented, maybe right here or elsewhere. Because it is only half-clear to me now.

Ok, so if the outer scope has RF edges, we need to make sure the inner scope has those RF edges too, right?
Ah, you are saying we are not doing that yet? Are you keeping track of that information for later?

Could we now create a reproducer that would fail in incremental inlining with a missing RF edge? Probably tricky, but very valuable ;)

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/25315#discussion_r2344201256