RFR: 8287384: Speed up jdk.test.lib.util.ForceGC

[Roger Riggs]

On Tue, 31 May 2022 17:41:08 GMT, Mandy Chung <mchung at openjdk.org> wrote:

> Hmm... one benefit of Cleaner is the ease of use to avoid the need of managing the reference queue. If the performance of the Cleaner API is a concern, perhaps we should look into reducing its overhead?

The code using a Cleaner here creates a CountDownLatch to wait on; that along with the Cleaner seems very similar to the complexity of the suggested alternative directly using a ReferenceQueue.  But I'm fine with it either way

Each Cleaner creates a new OS thread, used for a single instance of ForceGC.  
The caching of the Cleaner in a Static mostly reduces the per ForceGC overhead.

When Loom is a feature, a Virtual Thread would be sufficient.

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

PR: https://git.openjdk.java.net/jdk/pull/8907