Better default for ParallelGCThreads and ConcGCThreads by using number of physical cores and CPU mask.

Jungwoo Ha jwha at google.com
Fri Jan 10 01:04:02 UTC 2014


>
>
>
>> os_linux_x86.cpp
>>
>> This is where the important part of your change is. I don't know enough
>> about Linux internals to validate the changes you have made there. I will
>> leave that out of my review and let others (Runtime team?) with more Linux
>> knowledge review it.
>>
>>
>> Some code comments:
>>
>> Since core_count() returns 0 for all platforms except linux_x86, couldn't
>> the following code be moved to a platform independent place?
>>
>>  758 int os::active_core_count() {
>>  759   if (os::core_count() <= 0) {
>>  760     os::set_core_count(os::active_processor_count());
>>  761   }
>>  762   return os::core_count();
>>  763 }
>>
>> That way this code would not have to be duplicated in all other platforms:
>>
>>  726 int os::active_core_count() {
>>  727   return os::active_processor_count();
>>  728 }
>>
>> We could just always use os::active_core_count() and rely on the value it
>> returns.
>>
>
>  Sounds good to me.
>
>
> Good.
>
>
I am now working on this update. If we pull os::active_core_count() to
os.cpp, where should the linux_x86 version be?


Jungwoo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/hotspot-gc-dev/attachments/20140109/02ed4fb4/attachment.htm>


More information about the hotspot-gc-dev mailing list