Concurrent marking of root objects

[Zhengyu Gu]

Hi Ralph,

I am working on moving some root marking into concurrent phase.
https://bugs.openjdk.java.net/browse/JDK-8245961

ZGC also has concurrent thread stack marking 
(https://github.com/openjdk/zgc/commit/d9abea0cc6df148a3781600dfe6bfcc1a4a1873c#diff-a6bc76531aecc703472f887d4e162b94), 
hopefully, will mature in near future.

After that, remaining STW roots are quite small.

-Zhengyu

On 5/28/20 8:17 PM, raell at web.de wrote:
> Hi,
> 
> this isn't a real question, but more kind of an idea. After Shenandoah has made evacuation
> of root objects concurrent in JDK 14, I wonder if it shouldn't be possible to mark root
> objects concurrently, as well? To my understanding, the SATB barrier is already able
> to detect reference changes on root objects. So, at least from the SATB barrier point of view,
> it shouldn't be much of a difference if root objects are marked in a STW phase or concurrently.
> 
> Regards
> Ralph
>