8076177: Remove usage of stack.inline.hpp functions from taskqueue.hpp

Stefan Karlsson stefan.karlsson at oracle.com
Mon Apr 27 07:53:17 UTC 2015


On 2015-04-27 09:37, Bengt Rutisson wrote:
>
> Hi Stefan,
>
> On 2015-04-24 13:03, Stefan Karlsson wrote:
>> Hi Goetz,
>>
>> On 2015-04-24 11:47, Lindenmaier, Goetz wrote:
>>> Hi Stefan,
>>>
>>> I appreciate your changes to taskqueue, especially the introduction of
>>> a .inline.hpp file.  On a first view they look good.
>>>
>>> I tried to push the change this depends on, but it does not apply, 
>>> and, if fixed, does not build:
>>> http://cr.openjdk.java.net/~stefank/8078345/webrev.01/hotspot.changeset
>>> The code contains a new call to mark_and_push.  I just commented it 
>>> out for now.
>>
>> That patch was in turn built upon the two patches that I thought was 
>> unrelated, but they were not. All the related patches have been 
>> pushed to hotspot-gc, so the patch should apply now.
>>
>>>
>>> Then, I applied this patch, and started reviewing it, but it does 
>>> not contain all the
>>> changes in the diffs you mailed around.  E.g., 
>>> orderAccess.inline.hpp is not removed from taskqueue.hpp.
>>> http://cr.openjdk.java.net/~stefank/8076177/webrev.01/all/hotspot.changeset 
>>>
>>>
>>> Could you please post a complete patch, so I can test it?
>>
>> Sigh. I forgot about the webrev bug that creates incomplete/incorrect 
>> files:
>> https://bugs.openjdk.java.net/browse/CODETOOLS-7901115 - webrev -r 
>> generates broken patch files
>>
>> I've added the output from hg diff -r qparent at:
>> http://cr.openjdk.java.net/~stefank/8076177/webrev.01/all/hotspot.diff
>
> The hotspot.diff changes look good to me.

Yes, all other patches seems to be broken.

Thanks, Bengt.

StefanK

>
> Thanks,
> Bengt
>
>>
>> Hopefully that will apply without problems.
>>
>> Thanks,
>> StefanK
>>
>>>
>>> Best regards,
>>>    Goetz
>>>
>>> Rejected patch:
>>> --- psParallelCompact.hpp
>>> +++ psParallelCompact.hpp
>>> @@ -956,11 +947,8 @@
>>>       void do_klass(Klass* klass);
>>>     };
>>>   -  friend class FollowStackClosure;
>>>     friend class AdjustPointerClosure;
>>>     friend class AdjustKlassClosure;
>>> -  friend class FollowKlassClosure;
>>> -  friend class InstanceClassLoaderKlass;
>>>     friend class RefProcTaskProxy;
>>>      private:
>>>
>>>
>>> Build error:
>>>
>>> psParallelCompact.hpp: In member function void 
>>> PSParallelCompact::KeepAliveClosure::do_oop_work(T*) [with T = 
>>> oopDesc*]
>>> psParallelCompact.cpp:823:   instantiated from here
>>> psParallelCompact.hpp:1299: error: mark_and_push was not declared in 
>>> this scope
>>> psParallelCompact.hpp: In member function void 
>>> PSParallelCompact::KeepAliveClosure::do_oop_work(T*) [with T = 
>>> narrowOop]
>>> psParallelCompact.cpp:824:   instantiated from here
>>> psParallelCompact.hpp:1299: error: mark_and_push was not declared in 
>>> this scope
>>>
>>>
>>>
>>>
>>>
>>> -----Original Message-----
>>> From: hotspot-gc-dev 
>>> [mailto:hotspot-gc-dev-bounces at openjdk.java.net] On Behalf Of Stefan 
>>> Karlsson
>>> Sent: Donnerstag, 23. April 2015 13:44
>>> To: hotspot-gc-dev
>>> Subject: RFR: 8076177: Remove usage of stack.inline.hpp functions 
>>> from taskqueue.hpp
>>>
>>> Hi,
>>>
>>> Please review this patch to move the implementation of the push, pop,
>>> and steal functions out of taskqueue.hpp into a new 
>>> taskqueue.inline.hpp
>>> file.
>>>
>>> I've split the patch into three part, but intend to push them as one
>>> change, unless someone wants this to be pushed as three separate 
>>> changes.
>>>
>>> http://cr.openjdk.java.net/~stefank/8076177/webrev.01/00.movePush/
>>> http://cr.openjdk.java.net/~stefank/8076177/webrev.01/01.movePop/
>>> http://cr.openjdk.java.net/~stefank/8076177/webrev.01/02.moveSteal/
>>> http://cr.openjdk.java.net/~stefank/8076177/webrev.01/all/
>>>
>>> https://bugs.openjdk.java.net/browse/JDK-8076177
>>>
>>> The patch builds upon this patch that hasn't been pushed yet:
>>> http://mail.openjdk.java.net/pipermail/hotspot-gc-dev/2015-April/013039.html 
>>>
>>>
>>> Thanks,
>>> StefanK
>>
>



More information about the hotspot-gc-dev mailing list