[14] RFR(XS): 8223769: Assert triggers with -XX:+StressReflectiveCode

[Christian Hagedorn]

Hi Vladimir

Thanks for the review and your advise. The test took less than a minute 
locally with a slowdebug build using Graal, just slightly longer than 
without Graal.

Best regards,
Christian

On 27.07.19 03:21, Vladimir Kozlov wrote:
> Good.
> 
> Christian, please when you add a test with -Xcomp you also look on test 
> time when Graal is used as JIT.
> 
> Thanks,
> Vladimir
> 
> On 7/25/19 3:55 AM, Tobias Hartmann wrote:
>> Hi Christian,
>>
>> looks good to me.
>>
>> Best regards,
>> Tobias
>>
>> On 25.07.19 12:51, Christian Hagedorn wrote:
>>> Hi
>>>
>>> Please review the following patch:
>>> https://bugs.openjdk.java.net/browse/JDK-8223769
>>> http://cr.openjdk.java.net/~thartmann/8223769/webrev.00/
>>>
>>> The assertion is too strong when enabling the StressReflectiveCode 
>>> flag.  While the type system
>>> knows the size of an array is constant, the ArrayCopyNode::Length 
>>> input node itself is not. The
>>> layout helper is not used with StressReflectiveCode and therefore a 
>>> load node is inserted as part of
>>> the entire input to the length node
>>> (http://hg.openjdk.java.net/jdk/jdk/file/3307a6ded22d/src/hotspot/share/opto/graphKit.cpp#l3455). 
>>>
>>> This load node is not optimized away with StressReflectiveCode
>>> (http://hg.openjdk.java.net/jdk/jdk/file/3307a6ded22d/src/hotspot/share/opto/memnode.cpp#l1853). 
>>>
>>>
>>> Thanks!
>>>
>>> Best regards,
>>> Christian