[9] RFR(S): JDK-8029436: CICompilerCount is not updated when the number of compiler threads is adjusted to the number of CPUs

Vladimir Kozlov vladimir.kozlov at oracle.com
Fri Apr 11 08:29:37 UTC 2014


Looks good.

Thanks,
Vladimir

On 4/11/14 12:27 AM, Tobias Hartmann wrote:
> Vladimir, Christian, thanks for the reviews.
>
> Please see comments inline.
>
> On 04/10/2014 08:52 PM, Vladimir Kozlov wrote:
>> There was long discussion with Tobias on IM last night and I agreed to update CICompilerCount. I updated bug report.
>>
>> I thought to do that when it is default only:
>>
>> +    if (FLAG_IS_DEFAULT(CICompilerCount)) {
>> +      FLAG_SET_DEFAULT(CICompilerCount, count);
>> +    }
>>
>> If we reset it in all case (specifically when both flags are specified on command line) then we need to use
>> FLAG_SET_ERGO. yes, it is good compromise.
>
> Okay, I think so too. If you are okay with that, I now use FLAG_SET_ERGO and update it in all cases.
>
> New webrev: http://cr.openjdk.java.net/~anoll/8029436/webrev.02/ <http://cr.openjdk.java.net/%7Eanoll/8029436/webrev.02/>
>
>>
>> Thanks,
>> Vladimir
>>
>> On 4/10/14 11:31 AM, Christian Thalinger wrote:
>>> Vladimir, your comment in the bug says you don’t want CICompilerCount to
>>> be overwritten.  I’m unsure what to suggest but FLAG_SET_ERGO seems to
>>> be a good compromise.
>>>
>>> On Apr 10, 2014, at 7:31 AM, Vladimir Kozlov <vladimir.kozlov at oracle.com
>>> <mailto:vladimir.kozlov at oracle.com>> wrote:
>>>
>>>> So you decided to override CICompilerCount always (not when it is
>>>> default only). In such case you have to use FLAG_SET_ERGO() macro.
>>>>
>
> Yes, I thought this is necessary to be consistent. Otherwise CICompilerCount is not updated if both CICompilerCount and
> CICompilerCountPerCPU are set by the user.
>
> Thanks,
> Tobias
>
>>>> Thanks,
>>>> Vladimir
>>>>
>>>> On 4/10/14 4:37 AM, Tobias Hartmann wrote:
>>>>> Hi,
>>>>>
>>>>> please review the following patch for bug 8029436.
>>>>>
>>>>> *Problem:*
>>>>> The JVM option CICompilerCount holds the number of compiler threads.
>>>>> If CICompilerCountPerCPU is enabled, the number of
>>>>> compiler threads is adjusted at runtime by the policies. The value
>>>>> CICompilerCount is not updated.
>>>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8029436
>>>>>
>>>>> *Solution:*
>>>>> CICompilerCount is updated in AdvancedThresholdPolicy::initialize,
>>>>> SimpleThresholdPolicy::initialize and
>>>>> NonTieredCompPolicy::initialize. A warning is printed if the user
>>>>> both sets CICompilerCount and CICompilerCountPerCPU.
>>>>> Webrev:http://cr.openjdk.java.net/~anoll/8029436/webrev.01/<http://cr.openjdk.java.net/%7Eanoll/8029436/webrev.01/>
>>>>>
>>>>> *Tests:*
>>>>> JPRT
>>>>>
>>>>> Thanks,
>>>>>
>>>>> Tobias
>>>
>


More information about the hotspot-compiler-dev mailing list