RFR: 8230851 Interpreter could avoid calling the runtime for defaultvalue

Dmitry Samersoff dms at samersoff.net
Thu Sep 12 07:56:33 UTC 2019


Frederic,

> I’ve made the CR x86 specific, and also added [x86] to the summary.
> Do you need a more specific label?

What I need is a clear indication that the fix contains platform
specific code and therefor have to be ported (or validated) to other
platforms like AArch64.

Label seems to me the most natural way to provide such indication, but
any other approach will work for me as soon as I can easy select all
changes for period that requires my attention.

-Dmitry

On 11.09.19 17:13, Frederic Parain wrote:
> Dmitry,
> 
> Thank you for the review.
> 
> I’ve moved the verify_oop() as suggested:
> 
> http://cr.openjdk.java.net/~fparain/opt_int/webrev.01/index.html
> 
> I’ve made the CR x86 specific, and also added [x86] to the summary.
> Do you need a more specific label?
> 
> Regards,
> 
> Fred
> 
> 
>> On Sep 11, 2019, at 09:16, Dmitry Samersoff <dms at samersoff.net> wrote:
>>
>> Hello Frederic,
>>
>> Looks good to me.
>>
>> You may consider to move
>>
>> __ verify_oop(rax);
>>
>> after done, to save few instructions.
>>
>> PS:
>>
>> Could you add a label (something like x86_specific) to the CR to
>> simplify further porting.
>>
>> We discussed this approach with IoI and it may be a time to start
>> adopting it.
>>
>> -Dmitry
>>
>> On 11.09.19 15:20, Frederic Parain wrote:
>>> Please this review this small change in the interpreter adding
>>> a fast path to defaultvalue.
>>>
>>> CR: https://bugs.openjdk.java.net/browse/JDK-8230851
>>> Webrev: http://cr.openjdk.java.net/~fparain/opt_int/webrev.00/index.html
>>>
>>> Thank you,
>>>
>>> Fred
>>>
> 


More information about the valhalla-dev mailing list