Request for review:8013067:Zero builds are broken after 8010862

Jiangli Zhou jiangli.zhou at oracle.com
Mon May 6 15:04:50 PDT 2013


Hi Vladimir, Coleen and Christian,

Here is the updated webrev: 
http://cr.openjdk.java.net/~jiangli/8013067/webrev.01/. I left the 
GET_METHOD_COUNTERS macro unchanged due to usage of goto in CALL_VM. I 
think that's what Coleen tried to point out in her email.

Thanks,
Jiangli

On 05/06/2013 11:03 AM, Vladimir Kozlov wrote:
> On 5/6/13 10:40 AM, Jiangli Zhou wrote:
>> Hi Vladimir,
>>
>> I was considering whether to switch to functions, but decided not to
>
> Functions could be not inlined (in debug VM, for example) so you will 
> get performance regression. That is why we use macros sometimes but it 
> is not the case in this code.
>
>> alter the existing code too much. Thanks a lot for the review and
>> feedback. I will replace the macros.
>
> Thanks, in this code there is no big win from using macro:
>
> INCR_INVOCATION_COUNT(mcs)
>
> instead of:
>
> mcs->invocation_counter()->increment();
>
> Thanks,
> Vladimir
>
>>
>> Jiangli
>>
>> On 05/06/2013 10:13 AM, Vladimir Kozlov wrote:
>>> Jiangli,
>>>
>>> I understand that you followed existing code pattern using macros. But
>>> it is very difficult to debug, especially nested macros.
>>> I don't think we need separate macros just for 1 or 2 uses.
>>>
>>> I would like to see all these macros gone (GET_METHOD_COUNTERS,
>>> *_COUNT(mcs), INCR_*_COUNT(mcs)).
>>>
>>> Thanks,
>>> Vladimir
>>>
>>> On 5/6/13 9:19 AM, Jiangli Zhou wrote:
>>>> Christian has reviewed the change. Could I have one more review?
>>>>
>>>> Thanks,
>>>> Jiangli
>>>>
>>>> On 04/30/2013 04:31 PM, Jiangli Zhou wrote:
>>>>> Hi,
>>>>>
>>>>> Please review the following change for 8013067:
>>>>>
>>>>>   http://cr.openjdk.java.net/~jiangli/8013067/webrev.00/
>>>>>
>>>>> Thanks,
>>>>>
>>>>> Jiangli
>>>>
>>



More information about the hotspot-runtime-dev mailing list