RFR: Consolidate assembly write barriers
Roman Kennke
rkennke at redhat.com
Tue Jul 10 14:10:49 UTC 2018
Am 10.07.2018 um 10:07 schrieb Aleksey Shipilev:
> On 07/09/2018 08:12 PM, Aleksey Shipilev wrote:
>> On 07/09/2018 08:08 PM, Roman Kennke wrote:
>>>> // The set of registers to be saved+restored is the same as in the write-barrier above.
>>>> // Those are the commonly used registers in the interpreter.
>>>
>>> The write-barrier goes through the stub, which saves the registers.
>>
>> Ah! Okay then.
>>
>>> Remaining patch:
>>> http://cr.openjdk.java.net/~rkennke/shenandoahwb/webrev.01/
>>
>> Looks good.
>
> On a second thought, I think it hides the bug in SBSA::write_barrier_impl: it only goes through read
> barrier if (EVAC|TRAVERSAL), which is not complete, and it should go via that path when
> HAS_FORWARDED is true as well. I am preparing the fix.
>
> -Aleksey
>
>
Ok. Adjusted and hopefully final patch:
http://cr.openjdk.java.net/~rkennke/shenandoahwb/webrev.02/
Please check carefully if logic is as intended. ;-)
Passes tier1_gc_shenandoah
Roman
More information about the shenandoah-dev
mailing list