RFR 8251359: Shenandoah: filter null oops before calling enqueue/SATB barrier
zgu at redhat.com
Tue Aug 11 18:43:12 UTC 2020
Updated and pushed.
On 8/11/20 12:54 PM, Aleksey Shipilev wrote:
> On 8/11/20 5:53 PM, Zhengyu Gu wrote:
>> On 8/11/20 10:22 AM, Aleksey Shipilev wrote:
>>> On 8/11/20 4:16 PM, Zhengyu Gu wrote:
>>>> Please review this small patch that ensures null oops are filtered out
>>>> before calling enqueue/SATB barrier. Enqueuing null oops create extra
>>>> works and only get thrown out in later phase.
>>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8251359
>>>> Webrev: http://cr.openjdk.java.net/~zgu/JDK-8251359/webrev.00/index.html
>>> I don't think you need CompressedOops::is_null(obj) for guaranteed "oop", and you can just use "obj
>>> != NULL"? That is what ShenandoahBarrierSet::storeval_barrier does (you might add parentheses there)...
>>> 1328 if (is_concurrent_mark_in_progress() && (obj != NULL))
>> Updated: http://cr.openjdk.java.net/~zgu/JDK-8251359/webrev.01/
> Sorry, I was not clear. shenandoahBarrierSet.inline.hpp can do the same != check. No need for
> another review, consider changing it before pushing.
More information about the hotspot-gc-dev