RFR: Concurrent preclean

Roman Kennke rkennke at redhat.com
Wed May 31 12:25:13 UTC 2017


Am 31.05.2017 um 14:20 schrieb Aleksey Shipilev:
> On 05/31/2017 01:59 PM, Roman Kennke wrote:
>> - what's that commented part in referenceProcessor.cpp ?
> Ah, leftover I had from performance experiments that exacerbate the issue to
> extreme. That commented block is the initial filtering on reference discovery.
>
> Removed:
>  http://cr.openjdk.java.net/~shade/shenandoah/preclean/webrev.02/
>
>
>> - is precleaning single-threaded? if yes, can it be made multithreaded?
> RP has only single-threaded precleaning. But it can be multi-threaded if RP
> supported it, like it supports MT reference processing. Trying to do this
> outside of RP would be incorrect.
>
>> - isn't experimental flag defaulting to true pointless? shouldn't it be
>> diagnostic-> default true, experimental -> default false? (not very
>> important to me...)
> I thought we mark all our kill-switches "experimental", regardless of their
> default value. Kept as is.
>
> -Aleksey
>
Ok good to go.

Just out of curiousity: would MT precleaning win us anything significant?

Roman



More information about the shenandoah-dev mailing list