RFR: Remove obsolete and unused reduce-storeval-barrier optimization code

Aleksey Shipilev shade at redhat.com
Mon Sep 18 13:41:45 UTC 2017


On 09/18/2017 03:30 PM, Roman Kennke wrote:
> Am 18.09.2017 um 11:58 schrieb Aleksey Shipilev:
>> On 09/16/2017 11:37 AM, Roman Kennke wrote:
>>> Ok. I fixed one place in library_call.cpp in the cmpxchg intrinsics. Everything else should be ok:
>>>
>>> http://cr.openjdk.java.net/~rkennke/remove-reduce-storeval/webrev.01/
>> *) I wonder if instead of putting shenandoah_storeval_barrier prior to store_oop_* calls, we are
>> better putting the barriers calls inside of them? Under UseShenandoahGC flag? Otherwise, I see
>> missing shenandoah_storeval_barrier in GraphKit::builtin_throw and Parse::expand_multianewarray.
>>
>> Seems good otherwise.
> Good point. Like this?
> 
> http://cr.openjdk.java.net/~rkennke/remove-reduce-storeval/webrev.02/

Looks better.

So, the shenandoah_storeval_barrier from LibraryCallKit::LibraryCallKit::inline_unsafe_load_store is
now subsumed by something else? Can't see by what.

-Aleksey



More information about the shenandoah-dev mailing list