RFR: fix loop unswitching with -XX:-ShenandoahWriteBarrierMemBar

[Roman Kennke]

Am 1. Februar 2018 11:25:15 MEZ schrieb Roland Westrelin <rwestrel at redhat.com>:
>
>>>-ShenandoahWriteBarrierMemBar
>>>Compiler.compiler  thrpt   30  378.604 ± 18.569  ops/min
>>>
>>>+ShenandoahWriteBarrierMemBar
>>>Compiler.compiler  thrpt   30  398.564 ± 23.585  ops/min
>>>
>>>current repo with a product build.
>>
>> Yes. But without membars, I would expect it to be faster or equal
>compared to with membars. But not slower... Or am I missing something?
>
>Those membars are noop on x86 so perf should be the same. 

But it may affect compiler optimizations? Or if something went wrong, inhibit them?

Given how
>much
>compiler numbers vary, it's hard to tell there's a regression from the
>numbers above, AFAICT.

True. But we have both observed a noticable regression. Better to check it? No-membars will likely become the default/only path, so it better be good ;-)

Roman

-- 
Diese Nachricht wurde von meinem Android-Gerät mit K-9 Mail gesendet.