RFR: 8356556: AArch64: No need for acquire fence in safepoint poll in FFM

David Holmes dholmes at openjdk.org
Fri Jul 11 05:58:41 UTC 2025


On Thu, 10 Jul 2025 09:40:04 GMT, Samuel Chee <duke at openjdk.org> wrote:

>> Removes the unnecessary acquire to reduce memory ordering requirement since it is not longer needed as threads always disarm their own poll value.
>
> What seems to have happened, is that:
> 
> 1. I created this GitHub PR and someone else created the corresponding JBS ticket since I don't have the ability to do so.
> 2. This PR and the JBS ticket didn't automatically link together due to my lack of OCA access.
> 3. In the time it took for to get my OCA access (1.5 months), someone else saw the unassigned, unresolved JBS ticket, did it themselves and got it upstreamed before I ever got my OCA access.
> 4. I got my OCA access and realised what just happened.
> 
> What should happen now? Should I just close this PR?
> 
> And is there something both sides can do to prevent this from happening again? Since its work which has almost identically unnecessarily been done twice.

@spchee I would say that when @stooart-mon created the JBS issue he should have assigned it to himself and added a comment that you were working on it. That way it would have not been seen as available for picking up. Please just close this PR. If there is anything further to do in this area that your PR covered and the integrated one did not then a new JBS issue can be filed for that.

Also the OCA process should not have taken 1.5 months, but I can't comment on what may have happened there - @robilad?

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

PR Comment: https://git.openjdk.org/jdk/pull/25211#issuecomment-3060696128


More information about the hotspot-dev mailing list