[jdk17u-dev] RFR: 8331405: Shenandoah: Optimize ShenandoahLock with TTAS

[Xiaolong Peng]

On Thu, 12 Sep 2024 08:25:55 GMT, Goetz Lindenmaier <goetz at openjdk.org> wrote:

>> Hi, 
>>     This PR is a Backport of [JDK-8331405](https://bugs.openjdk.org/browse/JDK-8331405): Shenandoah: Optimize ShenandoahLock with TTAS, the original commit was authored by Aleksey Shipilev on 2 May 2024 and was reviewed by Zhengyu Gu and Y. S. Ramakrishna, and it had already been backported to jdk21.
>>     It is a clean merge w/o any conflicts, overall risk should be very minimal given it is single line change to apply TTAS. 
>> 
>>    Here is the series of improvements for ShenandoahLock I'm going to backport to JDK17:
>> 
>> | Bug         | Title                                          | PR         |
>> | ----------- | ---------------------------------------------- |------------|
>> | [JDK-8325587](https://bugs.openjdk.org/browse/JDK-8325587) | ShenandoahLock should allow blocking in VM     | [2797](https://github.com/openjdk/jdk17u-dev/pull/2797)     |
>> | [JDK-8331405](https://bugs.openjdk.org/browse/JDK-8331405) | Optimize ShenandoahLock with TTAS              | [2845](https://github.com/openjdk/jdk17u-dev/pull/2845) |
>> | [JDK-8331411](https://bugs.openjdk.org/browse/JDK-8331411) | Reconsider spinning duration in ShenandoahLock |    |
>> | [JDK-8335904](https://bugs.openjdk.org/browse/JDK-8335904) | Fix invalid comment in ShenandoahLock          |     |
>> 
>> ### Additional test
>> - [x] Linux AArch64 server fastdebug, hotspot_gc_shenandoah
>
> GHA failure: unrelated, well known ClhsdbFindPC issue.
> 
> @pengxiaolong please check the GHA tests before applying for approval.

@GoeLin Thanks for the review and approval, appreciate it! I did checked the GHA tests result and verified that the failures are not related this this change, should had mentioned it in the approval request or comments.
 
I understand that we prefer not to backport non-bug-fix changes to jdk17. As Aleksey said, we have seem the issue is causing problems in real services running in our production environment which will be in jdk17 for few more years I believe, also jdk17 will is the most widely adopted release, hence we decided to do these backports.

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

PR Comment: https://git.openjdk.org/jdk17u-dev/pull/2845#issuecomment-2345867743