RFR(S)[16]: 8246477: add whitebox support for deflating idle monitors

Daniel D. Daugherty daniel.daugherty at oracle.com
Fri Jun 19 16:58:15 UTC 2020


Ping!

And a testing update:

Mach5 Tier[1-8] testing:

   Tier[1-3] - done, 5 unrelated, known failures
   Tier4     - done - 1 unrelated, known failure
   Tier5     - done - no failures
   Tier6     - done, 1 unrelated, known failure
   Tier7     - almost done, 1 unrelated, known failure
   Tier8     - 56% done, 3 unrelated, known failures (so far)

The Mach5 testing is taking longer than usual due to resource limitations.
So far all failures are known to be in the baseline. There have been no
test failures related to not deflating an idle monitor in a timely fashion
(so far).

Thanks, in advance, for comments, questions or suggestions.

Dan

On 6/17/20 12:30 PM, Daniel D. Daugherty wrote:
> Greetings,
>
> I have a fix for cleaning up testing support for deflating idle monitors.
>
> JDK-8246477 add whitebox support for deflating idle monitors
> https://bugs.openjdk.java.net/browse/JDK-8246477
>
> This project is based on jdk-16+1 and is targeted to JDK16.
>
> Here's the webrev URL:
>
>     http://cr.openjdk.java.net/~dcubed/8246477-webrev/0-for-jdk16/
>
> Summary of the changes:
>
> - Add whitebox support for deflating idle monitors including
>   ObjectSynchronizer::request_deflate_idle_monitors(); includes
>   a new whitebox test.
> - Drop ObjectSynchronizer::_is_special_deflation_requested flag,
>   functions and uses.
> - Switch to ObjectSynchronizer::request_deflate_idle_monitors() as 
> needed.
> - bug fix: _last_async_deflation_time_ns should be set at the end of
>   async deflation.
>
> Because this fix is removing support for special deflation requests,
> I'm doing Mach5 Tier[1-8] testing:
>
>   Tier[1-3] - almost done, 5 unrelated, known failures
>   Tier4     - done - 1 unrelated, known failure
>   Tier5     - done - no failures
>   Tier6     - almost done, 1 unrelated, known failure
>   Tier7     - almost done, 1 unrelated, known failure
>   Tier8     - > half done, 3 unrelated, known failures (so far)
>
> The Mach5 testing is taking longer than usual due to resource 
> limitations.
> So far all failures are known to be in the baseline. There have been no
> test failures related to not deflating an idle monitor in a timely 
> fashion
> (so far).
>
> Thanks, in advance, for comments, questions or suggestions.
>
> Dan



More information about the hotspot-runtime-dev mailing list