RFR 8204947: Port ShenandoahTaskTerminator to mainline and make it default

Thomas Schatzl thomas.schatzl at oracle.com
Fri Nov 30 14:45:56 UTC 2018


On Thu, 2018-11-29 at 13:15 -0500, Zhengyu Gu wrote:
> Hi Thomas,
> > > I did not port it to CMS, as it is deprecated.
> > 
> > I submitted a job to push it through our testing infrastructure. I
> > will also run it through our perf benchmarks checking for issues.
> > 
> Thanks!

Passed hs-tier1-7, jdk-tier1-3. I started another run with latest
changes as CMS support (sans diagnostic flag change) has been added.

> > As for CMS support, I would prefer to provide it - I need to see
> > the reasoning in the CR - while CMS is deprecated, we may not be
> > able to remove it in 13, while I would like to remove this flag
> > independently and earlier :)
> > 
> > With no CMS support for this task terminator we tie flag removal to
> > CMS removal probably needlessly.
> Okay, ported to CMS.
> Webrev: 
> http://cr.openjdk.java.net/~zgu/JDK-8204947/webrev.02/index.html
> and passed hotspot_gc on Linux 64. I will run more tests.

I think the change is good from a functionality POV, however I think we
can reduce the added boilerplate code (TaskTerminator with assignment
operators, etc.) a bit. Give me a few hours to try out something.

In the meantime I collected typos and formatting issues in 
http://cr.openjdk.java.net/~tschatzl/8204947/webrev/ . Nothing
interesting to see there, but fixing up typos that I do not think are
worth listing one by one here again.

(It also changes the flag from experimental to diagnostic, but you
already did that).


More information about the hotspot-gc-dev mailing list