RFR: 8258414: OldObjectSample events too expensive [v2]
Jaroslav Bachorik
jbachorik at openjdk.java.net
Thu Mar 11 14:29:09 UTC 2021
On Thu, 11 Mar 2021 10:55:54 GMT, Markus Grönlund <mgronlun at openjdk.org> wrote:
>> Thanks, Jaroslav - yes, indeed.
>
> I also attempted to have the artefact tagging happen lazily, e.g. methods and classes, making it a part of the stack trace resolution process. There is a problem with that approach concerning class unloading because a referenced class can unload, but since it does not yet have the proper tag, the framework does not intercept it. I elaborated on setting an "interest" tag to handle this, but that approach opens additional challenges related to visibility and concurrency. Hard problem to solve; I was unable to do it (yet).
Probably can live without lazy tagging in the first iteration - still it will be much better than the current state.
And then we can keep on adding incremental improvements based on cost-benefit weighing.
-------------
PR: https://git.openjdk.java.net/jdk/pull/2780
More information about the hotspot-jfr-dev
mailing list