RFR: Improved, more flexible heap region size settings
Zhengyu Gu
zgu at redhat.com
Tue Nov 15 15:22:56 UTC 2016
Some format issues in shenandoah_globals.hpp. Look good otherwise.
-Zhengyu
On 11/15/2016 10:07 AM, Roman Kennke wrote:
> Am Dienstag, den 15.11.2016, 09:36 -0500 schrieb Zhengyu Gu:
>> Hi Roman,
>>
>> Good in general. A couple of things to consider:
>>
>> 1. Could you not hard code the sizes? use enum instead.
>>
>> 2. Could you also specific unit of the sizes? I can see they are in
>> MB, but just make them clear.
> Like this?
>
> http://cr.openjdk.java.net/~rkennke/heapregionsize/webrev.01/
>
> Roman
>
>
>> Thanks,
>>
>> -Zhengyu
>>
>>
>>
>> On 11/15/2016 05:07 AM, Roman Kennke wrote:
>>> Ping?
>>>
>>> Am Freitag, den 11.11.2016, 18:55 +0100 schrieb Roman Kennke:
>>>> Currently, the heap region size is determined using hardcoded min
>>>> (1M),
>>>> max (32M) and target number (2048). We also have an option to
>>>> explicitely set regions size, but it will still not go below or
>>>> over
>>>> the hardcoded bounds.
>>>>
>>>> This change turns the 3 hardcoded values into proper cmd line
>>>> args,
>>>> and
>>>> also changes -XX:ShenandoahHeapRegionSize to override whatever
>>>> bounds
>>>> are set (default or not). All 3 arguments are checked for sanity.
>>>> I'm
>>>> including a jtreg test for the args checking.
>>>>
>>>> http://cr.openjdk.java.net/~rkennke/heapregionsize/webrev.00/
>>>>
>>>> Ok?
>>>>
>>>> Roman
More information about the shenandoah-dev
mailing list