RFR: 8273605: VM Exit does not abort concurrent mark

Kim Barrett kbarrett at openjdk.java.net
Mon Sep 13 16:13:52 UTC 2021


On Mon, 13 Sep 2021 14:38:43 GMT, Thomas Schatzl <tschatzl at openjdk.org> wrote:

> Hi,
> 
>   can I have reviews for this change that adds (best-effort) notification of the concurrent mark thread about that the VM is exiting. In particular, the change reuses the existing abort mechanism.
> 
> Best-effort because the change just flips a global flag that is periodically polled by all threads (particularly the concurrent mark thread) to see whether we should abort.
> 
> In that test case it reduces shutdown time from a wait until the whole marking cycle finished to "immediate".
> 
> There is no test case because I could not think of anything that's reproducable: after all to have a very long delay for shutdown we need to have a *huge* heap (tested on a 50gb heap with 25gb live data), and additionally measuring shutdown time in CI is just asking for spurious errors when that process does not get enough cpu time for any reason.
> 
> Testing: test case that found that issue; tier1-3

Looks good.

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

Marked as reviewed by kbarrett (Reviewer).

PR: https://git.openjdk.java.net/jdk/pull/5493



More information about the hotspot-gc-dev mailing list