RFR (S): 8189355: Cleanup of BarrierSet barrier functions

[Erik Österlund]

Hi Aleksey,

New full webrev:
http://cr.openjdk.java.net/~eosterlund/8189355/webrev.01/

New incremental webrev:
http://cr.openjdk.java.net/~eosterlund/8189355/webrev.00_01/


On 2017-10-17 21:37, Aleksey Shipilev wrote:
> Hi Erik,
>
> On 10/17/2017 12:27 PM, Erik Österlund wrote:
>> Webrev:
>> http://cr.openjdk.java.net/~eosterlund/8189355/webrev.00/
> Good patch, I scratched my head about the meaning about many of those methods when building the most
> trivial BS for Epsilon:
>
> http://hg.openjdk.java.net/jdk10/sandbox/hotspot/file/2e4b568252ea/src/share/vm/gc/epsilon/epsilonBarrierSet.hpp

:)

> Which also raises one observations:
>
>   *) gen_write_ref_array_pre_barrier in stubGenerator_arm.cpp now has unconditional call to
> BarrierSet::static_write_ref_array_pre, right? Epsilon returns "false" for
> has_write_ref_pre_barrier. Is this something to be addressed by GC interface later?

Yes this will indeed be addressed later on and will be solved in a much 
better way (if I say so myself). But since it was mentioned by both you 
and Kim in this round, I have a new version of the patch that fixes that 
so that the pre barrier runtime calls for arraycopy stubs are only 
generated if G1 was selected.

> Otherwise a very welcome cleanup!

Thank you for the review.

/Erik

> Thanks,
> -Aleksey
>
>