Request for Review (s) - 8135318: CMS wrong max_eden_size for check_gc_overhead_limit

[Volker Simonis]

Hi John,

could you please add noreg-hard label to the bug. I think that should
be fine for this simple fix.

The problem is that for the bug to manifest, all the young spaces must
have exactly the same size. But that's hard to achieve because on the
command line we can only give the ratio between the various regions.
In reality, the actual size is computed from the ratio but rounded
such that exact equality is not guaranteed.

Thank you and best regards,
Volker



On Wed, Nov 4, 2015 at 6:10 PM, Jon Masamitsu <jon.masamitsu at oracle.com> wrote:
> Ivan,
>
> Do you have a regression test for this?  I'm expected to push
> a regression test with the fix or explain why it's hard to write a
> regression
> test.
>
> Jon
>
>
> On 11/4/2015 1:30 AM, Galkin, Ivan wrote:
>>
>> Hello Jon,
>>
>> Looks good! Thank you!
>>
>> Kind regards,
>> Ivan
>>
>> -----Original Message-----
>> From: hotspot-gc-dev [mailto:hotspot-gc-dev-bounces at openjdk.java.net] On
>> Behalf Of Jon Masamitsu
>> Sent: Dienstag, 3. November 2015 18:54
>> To: hotspot-gc-dev at openjdk.java.net openjdk.java.net
>> Subject: Request for Review (s) - 8135318: CMS wrong max_eden_size for
>> check_gc_overhead_limit
>>
>> This fixes a bug in the CMS calculation that is used to
>> check the GC overhead limit.  The max size of eden was
>> incorrectly calculated as the young gen size - 2 * survivor
>> space size.  The young gen size only includes 1 survivor space
>> so only 1 should have been subtracted.  The fix uses the young
>> generation max_eden_size() that does the calculation correctly.
>>
>> Fix was contributed by ivan.galkin at sap.com
>>
>> Ivan, please check that the patch is correct.
>>
>>
>> CR: https://bugs.openjdk.java.net/browse/JDK-8135318
>> webrev: http://cr.openjdk.java.net/~jmasa/8135318/webrev.00/
>>
>> Thanks.
>>
>> Jon
>>
>