RFR(S): 8231961: [lworld] C1 compiled code returns incorrect value for flattened array access

Tobias Hartmann tobias.hartmann at oracle.com
Thu Oct 10 06:11:05 UTC 2019


Thanks Ioi!

Best regards,
Tobias

On 10.10.19 08:01, Ioi Lam wrote:
> Looks good to me.
> 
> Thanks
> - Ioi
> 
> On 10/9/19 5:43 AM, Tobias Hartmann wrote:
>> Hi,
>>
>> please review the following patch:
>> https://bugs.openjdk.java.net/browse/JDK-8231961
>> http://cr.openjdk.java.net/~thartmann/8231961/webrev.00/
>>
>> In C1, flattened array loads are connected to a buffer allocation (NewValueTypeInstance) to write
>> the result into. If there are multiple loads from the same array element, local/global value
>> numbering should not remove them because they are initializing different buffers. The fix is to
>> simply make the hashing code aware of the _vt field of the LoadIndexed instruction.
>>
>> Thanks,
>> Tobias
> 


More information about the valhalla-dev mailing list