RFR: 8148376: Test for PLAB behavior at evacuation failure.

Thomas Schatzl thomas.schatzl at oracle.com
Tue Apr 5 10:41:06 UTC 2016


On Fri, 2016-03-18 at 18:47 +0300, Michail Chernov wrote:
> Hi,
> Could I have a review for this change, please:
> Bug: https://bugs.openjdk.java.net/browse/JDK-8148376
> Webrev: http://cr.openjdk.java.net/~mchernov/8148376/webrev.00/
> This test provokes Evacuation Failure and checks PLAB statistics for 
> this event.
> This fix is also related to 
> http://mail.openjdk.java.net/pipermail/hotspot-gc-dev/2016-March/0171
> 97.html 
> because uses LogParser.java.

-   56  * that both fields 'falure_waste' and 'failure_used' for
Evacuation Failure statistic

Should be "failure_waste"

- I think the test should specify 
-XX:InitiatingHeapOccupancyPercent=100 and -XX:-G1UseAdaptiveIHOP to
make sure that evacuation failures happen always.

Otherwise some future ultra-fast computer might be able to catch up
(then again, the test never releases any objects)


More information about the hotspot-gc-dev mailing list