Request for reviews (S): 7190310: Inlining WeakReference.get(), and hoisting $referent may lead to non-terminating loops

Christian Thalinger christian.thalinger at oracle.com
Wed Aug 15 17:44:59 PDT 2012


On Aug 15, 2012, at 4:26 PM, Vladimir Kozlov <vladimir.kozlov at oracle.com> wrote:

> I updated webrev:
> 
> http://cr.openjdk.java.net/~kvn/7190310/webrev.01
> 
> During testing I found several problem with code for unsafe.getObject() intrinsic code in both: C1 and C2. I moved some checks in C1 from G1UnsafeGetObjSATBBarrierStub stub into shared code and also added static klass analysis to avoid generation of G1/membar code if klass is not Reference or Object. In C2 I hit bad graph problem because If node's projections in insert_pre_barrier() were not put on IGVN worklist.
> 
> I added jtreg tests (thanks to John Cuthbertson for unsafe test which I extended).
> 
> Tested with CTW and Kitchensink with C1, C2, tiered.

Looks good.

-- Chris

> 
> Thanks,
> Vladimir
> 
> Vladimir Kozlov wrote:
>> Christian Thalinger wrote:
>>> On Aug 9, 2012, at 7:15 PM, Vladimir Kozlov <vladimir.kozlov at oracle.com> wrote:
>>> 
>>>> Thanks, David
>>>> 
>>>> pre_barrier() methods don't generate code for other GCs. But I forgot to put UseG1GC guard around "new G1UnsafeGetObjSATBBarrierStub" in c1_LIRGenerator.cpp.
>>> 
>>> Did you update the webrev?  -- Chris
>> Not yet. I want to rewrite that code (which use G1UnsafeGetObjSATBBarrierStub).
>> Vladimir
>>> 
>>>> Thanks,
>>>> Vladimir
>>>> 
>>>> On 8/9/12 6:52 PM, David Holmes wrote:
>>>>> On 10/08/2012 11:34 AM, Vladimir Kozlov wrote:
>>>>>> http://cr.openjdk.java.net/~kvn/7190310/webrev
>>>>>> 
>>>>>> 7190310: Inlining WeakReference.get(), and hoisting $referent may lead
>>>>>> to non-terminating loops
>>>>>> 
>>>>>> Add acquire membar after load from Reference.referent field to prevent
>>>>>> commoning of loads across safepoint since GC can change its value.
>>>>> Wow that was quick! :)
>>>>> 
>>>>> My only comment (not being a compiler expert) is that it would appear that we are now inserting read barriers with all
>>>>> GC's not just G1. What kind of impact will that have on performance under the other GC's?
>>>>> 
>>>>> Thanks,
>>>>> David
>>>>> 
>>>>>> Thanks,
>>>>>> Vladimir



More information about the hotspot-compiler-dev mailing list