RFR(L) 8153224 Monitor deflation prolong safepoints (CR12/v2.12/15-for-jdk15)
Erik Österlund
erik.osterlund at oracle.com
Wed May 20 11:28:25 UTC 2020
Hi Dan,
I'm glad we are in agreement about the problem domain. I will focus my
reply on the solution domain.
On 2020-05-20 03:05, Daniel D. Daugherty wrote:
>> In the solution space, I would like to make it so that there really
>> is a linearization point for committing
>> the deflation when the counter is decremented to a negative value.
>
> I've previously said that this line of code defines the linearization
> point:
>
> (owner_is_DEFLATER_MARKER() && contentions() < 0)
>
> After the analysis of this race (thank you for finding it), I have to
> amend
> that to this:
>
> (owner_is_DEFLATER_MARKER() && contentions() < 0 &&
> owner_is_DEFLATER_MARKER())
>
> so that the check for the linearization point matches the logic
> implemented by
> the three part async deflation protocol in deflate_monitor_using_JT().
First of all, a linearization point is when a single atomic write
logically commits a whole operation, such that all concurrent threads
that observe this write, will know that the operation has logically been
committed. A correct concurrent algorithm may or may not have a
linearization point. What you are proposing does not have a
linearization point, while the change I proposed did have a
linearization point, (the write of _contentions to -max_jint). In my
proposed changes, anyone observing that store making _contentions < 0
knows that deflation has been logically committed. In what you are
proposing it's more of a dance where you check for the owner, check for
contentions, and then check owners again, assuming a total order, and
reasoning about what could happen between these events. That doesn't
automatically mean that it's bad to not have a linearization point. I'd
just like to point out that in what you are proposing, there is indeed
no linearization point for deflating the monitor. No individual write
publishes that now deflation is a fact. We can live with that though,
but I find it less intuitive to reason about, and would prefer to have a
linearization point.
> I think the solution that I'm proposing with the revised version of
> ObjectMonitor::is_being_async_deflated() above is way simpler than what
> is described in the next two paragraphs below. Obviously, this is the
> key piece of this reply and I need to know if you agree that the above
> closes the race (on TSO machines!).
I agree that what you propose will probably fix the issue on TSO
machines. But naturally it won't on non-TSO machines, which is where
things get messy. In fact the reason I did not propose the same solution
is that it's quite tricky to reason about what fencing is required. In
this case, the _owner and _contentions are written by different threads,
but we need to observe them in a total order, requiring at least a
loadload(). That also makes it seem like an IRIW scenario, but it's not
that "simple". The threads that writes _owner and _contentions know
about each other through causality. Causality is when a write on one
thread is read on a second thread, before a write on that thread, which
is observed on a third thread. The third thread is guaranteed to be able
to subsequently observe the write from the first thread if it observed
the write of the second thread, iff it supports causality of accesses.
Managing to flip _contentions to negative implies knowledge about the
subsequent counter decrement from a JT that aborted the deflation.
Therefore, if a third observer thread can read the negative counter,
their memory should be sufficiently up to date to guarantee the
subsequent load of the owner can't have a value written before the
counter transitioned to negative. This works with causality rules.
Unfortunately, causality is not ensured on PPC. So unless there is more
to it and causality is otherwise ensured with fencing in the causality
chain, there is an IRIW situation again. The causality link we rely on
is only ensured if the read that observes the _contentions == 0 value
written by the aborting java thread, and the write that makes it
negative are interleaved with a full fence. This causality link is
inside of the guts of the CAS that flips contentions from 0 to negative.
The implementation of the CAS on PPC in pseudo code is:
fence()
for (;;)
temp = load_link(addr);
if (temp != expected_value) return false;
if (store_conditional(addr, new_value)) {
break;
}
fence()
return true;
This load_link and store_conditional is the weak link in our causality
chain. There is no full fence between them, and that unfortunately means
that causality is not ensured in our algorithm. And because the
causality chain is not respected, we find ourselves in yet another IRIW
situation, but this time even more tricky to reason about, due to its
existence being conditional on a lack of causality, which is very
implicit and tricky to see here and reason about.
Comparision of your approach to my approch:
1) We get into even more tricky fencing decision. It's an IRIW situation
iff causality is not ensured by the HW memory model, which it is not on
PPC. Not sure what to make out of that, and how easy others will find it
to reason about the correctness of the algorithm. In my solution, the
fencing is straight forward.
2) There is still no intuitive linearization point. We are reasoning
about reading two atomic variables in different orders, trying to figure
out what has happened. Neither of the two atomic variables transition
monotonically. It is rather unintuitive, and could be more intuitive
like my solution (with a linearization point). By making the key
transition of the _contentions counter to negative a monotonic
transition that will never flip back, it becomes easier to reason about
correctness. That becomes the only transition that observers need to
care about, and it can't be undone. I definitely find that more
intuitive and simple to reason about, compared to reasoning about all
the ways that the two non-monotonic variables can have their state flip
back and forth in different interleavings with other threads, in
different skewed local versions of different total orders of events.
3) The is_async_deflated() function becomes more heavy weight,
especially on nMCA machines. It could be a single load, as it is in my
solution.
I have included a patch so that you can see what my solution might look
like, for your consideration:
http://cr.openjdk.java.net/~eosterlund/monitors/webrev.00/
Do you still find your solution easier to reason about?
>
>> The idea is to let JT1 that aborts the deflation with a successful
>> _owner flip to leave its stake in the
>> _contentions counter there when it exits, so that the linearization
>> point is blocked appropriately. The
>> responsibility to decrement the _contentions counter falls to the
>> ServiceThread. Once it tries to CAS
>> the _contentions from 0 to negative, it is guaranteed to fail if
>> another thread has flipped back the _owner.
>> When the service thread fails to make the counter negative (and hence
>> commit the deflation operation as definitive),
>> it tries to CAS the _owner back to NULL. If that succeeds, then the
>> ServiceThread aborted deflation. If that does not
>> succeed, another thread aborted deflation, and then the ServiceThread
>> decrements _contentions by 1 to
>> balance out the counter (as whichever thread left 1 stake in the
>> _contentions counter).
>>
>> Now we have a real linearization point. If anyone observes a negative
>> counter, the deflation is definitive.
>> If not, it has not yet been decided whether to commit or abort
>> deflation.
>> The is_async_deflated() function then simply becomes: return
>> Atomic::load(&_contentions) < 0.
>>
>> The second issue I would like to highlight is that all places where
>> we help out with installing back
>> the object markWord from the ObjectMonitor if a monitor is found to
>> be deflated, must have more rigorous
>> fencing between the load of _contentions in the is_async_deflated()
>> check and the load of the _header in
>> the install_displaced_markword_in_object() function. This is another
>> place where there exists IRIW issues.
>> Therefore, I propose to put another fancy if (nMCA) loadload else
>> loadload() right before loading the _header
>> in install_displaced_markword_in_object. If we don't do this then the
>> two loads can be reordered w.r.t. the
>> total order. This can result in a racy installation of a new
>> hashCode, that some threads start observing
>> and using, not making it to the deflated object header. This would
>> cause inconsistencies. It is once again a
>> situation where the value of _contentions to negative is updated by
>> one thread, and the update of the _header
>> to have a hashCode is performed by another thread, and all observers
>> must have a total ordering w.r.t. which
>> one happened-before the other, or we are in trouble. I'm okay if you
>> want to wait with commenting on this
>> until after our off-list IRIW discussion has cooled down though.
>
> I'm going to have to mull on the need for a memory barrier (loadload()
> for TSO) in between the is_being_async_deflated() call and the
> install_displaced_markword_in_object() call. If we decide we need it
> for all install_displaced_markword_in_object() calls, then it would
> make sense to put the logic in the function itself rather than in the
> callers.
Okay.
Thanks,
/Erik
>> Otherwise this looks good.
>
> Thanks.
>
>
>> And sorry for more headache.
>
> It's worth discussing this in gory detail before we integrate it than
> trying to hunt down elusive failures due to a race later.
>
> Thanks for your thorough reviews!
>
> Dan
>
>
>>
>> Thanks,
>> /Erik
>>
>> On 2020-05-14 23:40, Daniel D. Daugherty wrote:
>>> Greetings,
>>>
>>> I have made changes to the Async Monitor Deflation code in response to
>>> the CR11/v2.11/14-for-jdk15 code review cycle. Thanks to David H.,
>>> Erik O.,
>>> and Robbin for their OpenJDK reviews in the v2.11 round!
>>>
>>> I have attached the change list from CR11 to CR12 and I've also added a
>>> link to the CR11-to-CR12-changes file to the webrevs so it should be
>>> easy
>>> to find.
>>>
>>> Main bug URL:
>>>
>>> JDK-8153224 Monitor deflation prolong safepoints
>>> https://bugs.openjdk.java.net/browse/JDK-8153224
>>>
>>> The project is currently baselined on jdk-15+23.
>>>
>>> Here's the full webrev URL for those folks that want to see all of the
>>> current Async Monitor Deflation code in one go (v2.12 full):
>>>
>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/15-for-jdk15%2b23.v2.12.full/
>>>
>>>
>>> Some folks might want to see just what has changed since the last
>>> review
>>> cycle so here's a webrev for that (v2.12 inc):
>>>
>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/15-for-jdk15%2b23.v2.12.inc/
>>>
>>>
>>>
>>> The OpenJDK wiki is currently at v2.11 and might require minor
>>> tweaks for v2.12:
>>>
>>> https://wiki.openjdk.java.net/display/HotSpot/Async+Monitor+Deflation
>>>
>>> The jdk-15+23 based v2.12 version of the patch is going thru the usual
>>> Mach5 testing right now.
>>>
>>> Thanks, in advance, for any questions, comments or suggestions.
>>>
>>> Dan
>>>
>>>
>>> On 5/7/20 1:08 PM, Daniel D. Daugherty wrote:
>>>> Greetings,
>>>>
>>>> I have made changes to the Async Monitor Deflation code in response to
>>>> the CR10/v2.10/13-for-jdk15 code review cycle and DaCapo-h2 perf
>>>> testing.
>>>> Thanks to Erik O., Robbin and David H. for their OpenJDK reviews in
>>>> the
>>>> v2.10 round! Thanks to Eric C. for his help in isolating the DaCapo-h2
>>>> performance regression.
>>>>
>>>> With the removal of ref_counting and the ObjectMonitorHandle class,
>>>> the
>>>> Async Monitor Deflation project is now closer to Carsten's original
>>>> prototype. While ref_counting gave us ObjectMonitor* safety
>>>> enforced by
>>>> code, I saw a ~22.8% slow down with -XX:-AsyncDeflateIdleMonitors
>>>> ("off"
>>>> mode). The slow down with "on" mode -XX:+AsyncDeflateIdleMonitors
>>>> is ~17%.
>>>>
>>>> I have attached the change list from CR10 to CR11 instead of
>>>> putting it in
>>>> the body of this email. I've also added a link to the
>>>> CR10-to-CR11-changes
>>>> file to the webrevs so it should be easy to find.
>>>>
>>>> Main bug URL:
>>>>
>>>> JDK-8153224 Monitor deflation prolong safepoints
>>>> https://bugs.openjdk.java.net/browse/JDK-8153224
>>>>
>>>> The project is currently baselined on jdk-15+21.
>>>>
>>>> Here's the full webrev URL for those folks that want to see all of the
>>>> current Async Monitor Deflation code in one go (v2.11 full):
>>>>
>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/14-for-jdk15%2b21.v2.11.full/
>>>>
>>>>
>>>> Some folks might want to see just what has changed since the last
>>>> review
>>>> cycle so here's a webrev for that (v2.11 inc):
>>>>
>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/14-for-jdk15%2b21.v2.11.inc/
>>>>
>>>>
>>>> Because of the removal of ref_counting and the ObjectMonitorHandle
>>>> class, the
>>>> incremental webrev is a bit noisier than I would have preferred.
>>>>
>>>>
>>>> The OpenJDK wiki has NOT YET been updated for this round of changes:
>>>>
>>>> https://wiki.openjdk.java.net/display/HotSpot/Async+Monitor+Deflation
>>>>
>>>> The jdk-15+21 based v2.11 version of the patch has been thru Mach5
>>>> tier[1-6]
>>>> testing on Oracle's usual set of platforms. Mach5 tier[78] are
>>>> still running.
>>>> I'm running the v2.11 patch through my usual set of stress testing on
>>>> Linux-X64 and macOSX.
>>>>
>>>> I'm planning to do a SPECjbb2015, DaCapo-h2 and volano round on the
>>>> CR11/v2.11/14-for-jdk15 bits.
>>>>
>>>> Thanks, in advance, for any questions, comments or suggestions.
>>>>
>>>> Dan
>>>>
>>>>
>>>> On 2/26/20 5:22 PM, Daniel D. Daugherty wrote:
>>>>> Greetings,
>>>>>
>>>>> I have made changes to the Async Monitor Deflation code in
>>>>> response to
>>>>> the CR9/v2.09/12-for-jdk14 code review cycle. Thanks to Robbin and
>>>>> Erik O.
>>>>> for their comments in this round!
>>>>>
>>>>> With the extraction and push of {8235931,8236035,8235795} to
>>>>> JDK15, the
>>>>> Async Monitor Deflation code is back to "just" async deflation
>>>>> changes!
>>>>>
>>>>> I have attached the change list from CR9 to CR10 instead of
>>>>> putting it in
>>>>> the body of this email. I've also added a link to the
>>>>> CR9-to-CR10-changes
>>>>> file to the webrevs so it should be easy to find.
>>>>>
>>>>> Main bug URL:
>>>>>
>>>>> JDK-8153224 Monitor deflation prolong safepoints
>>>>> https://bugs.openjdk.java.net/browse/JDK-8153224
>>>>>
>>>>> The project is currently baselined on jdk-15+11.
>>>>>
>>>>> Here's the full webrev URL for those folks that want to see all of
>>>>> the
>>>>> current Async Monitor Deflation code in one go (v2.10 full):
>>>>>
>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/13-for-jdk15+11.v2.10.full/
>>>>>
>>>>>
>>>>> Some folks might want to see just what has changed since the last
>>>>> review
>>>>> cycle so here's a webrev for that (v2.10 inc):
>>>>>
>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/13-for-jdk15+11.v2.10.inc/
>>>>>
>>>>>
>>>>> Since we backed out the HandshakeAfterDeflateIdleMonitors option
>>>>> and the
>>>>> C2 ref_count changes and updated the copyright years, the "inc"
>>>>> webrev has
>>>>> a bit more noise in it than usual. Sorry about that!
>>>>>
>>>>> The OpenJDK wiki has been updated for this round of changes:
>>>>>
>>>>> https://wiki.openjdk.java.net/display/HotSpot/Async+Monitor+Deflation
>>>>>
>>>>> The jdk-15+11 based v2.10 version of the patch has been thru Mach5
>>>>> tier[1-7]
>>>>> testing on Oracle's usual set of platforms. Mach5 tier8 is still
>>>>> running.
>>>>> I'm running the v2.10 patch through my usual set of stress testing on
>>>>> Linux-X64 and macOSX.
>>>>>
>>>>> I'm planning to do a SPECjbb2015 round on the
>>>>> CR10/v2.20/13-for-jdk15 bits.
>>>>>
>>>>> Thanks, in advance, for any questions, comments or suggestions.
>>>>>
>>>>> Dan
>>>>>
>>>>>
>>>>> On 2/4/20 9:41 AM, Daniel D. Daugherty wrote:
>>>>>> Greetings,
>>>>>>
>>>>>> This project is no longer targeted to JDK14 so this is NOT an
>>>>>> urgent code
>>>>>> review request.
>>>>>>
>>>>>> I've extracted the following three fixes from the Async Monitor
>>>>>> Deflation
>>>>>> project code:
>>>>>>
>>>>>> JDK-8235931 add OM_CACHE_LINE_SIZE and use smaller size on
>>>>>> SPARCv9 and X64
>>>>>> https://bugs.openjdk.java.net/browse/JDK-8235931
>>>>>>
>>>>>> JDK-8236035 refactor ObjectMonitor::set_owner() and _owner
>>>>>> field setting
>>>>>> https://bugs.openjdk.java.net/browse/JDK-8236035
>>>>>>
>>>>>> JDK-8235795 replace monitor list
>>>>>> mux{Acquire,Release}(&gListLock) with spin locks
>>>>>> https://bugs.openjdk.java.net/browse/JDK-8235795
>>>>>>
>>>>>> Each of these has been reviewed separately and will be pushed to
>>>>>> JDK15
>>>>>> in the near future (possibly by the end of this week). Of course,
>>>>>> there
>>>>>> were improvements during these review cycles and the purpose of this
>>>>>> e-mail is to provided updated webrevs for this fix
>>>>>> (CR9/v2.09/12-for-jdk14)
>>>>>> within the revised context provided by {8235931, 8236035, 8235795}.
>>>>>>
>>>>>> Main bug URL:
>>>>>>
>>>>>> JDK-8153224 Monitor deflation prolong safepoints
>>>>>> https://bugs.openjdk.java.net/browse/JDK-8153224
>>>>>>
>>>>>> The project is currently baselined on jdk-14+34.
>>>>>>
>>>>>> Here's the full webrev URL for those folks that want to see all
>>>>>> of the
>>>>>> current Async Monitor Deflation code along with {8235931,
>>>>>> 8236035, 8235795}
>>>>>> in one go (v2.09b full):
>>>>>>
>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/12-for-jdk14.v2.09b.full/
>>>>>>
>>>>>>
>>>>>> Compare the open.patch file in 12-for-jdk14.v2.09.full and
>>>>>> 12-for-jdk14.v2.09b.full
>>>>>> using your favorite file comparison/merge tool to see how Async
>>>>>> Monitor Deflation
>>>>>> evolved due to {8235931, 8236035, 8235795}.
>>>>>>
>>>>>> Some folks might want to see just the Async Monitor Deflation
>>>>>> code on top of
>>>>>> {8235931, 8236035, 8235795} so here's a webrev for that (v2.09b
>>>>>> inc):
>>>>>>
>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/12-for-jdk14.v2.09b.inc/
>>>>>>
>>>>>>
>>>>>> These webrevs have gone thru several Mach5 Tier[1-8] runs along with
>>>>>> my usual stress testing and SPECjbb2015 testing and there aren't any
>>>>>> surprises relative to CR9/v2.09/12-for-jdk14.
>>>>>>
>>>>>> Thanks, in advance, for any questions, comments or suggestions.
>>>>>>
>>>>>> Dan
>>>>>>
>>>>>>
>>>>>> On 12/11/19 3:41 PM, Daniel D. Daugherty wrote:
>>>>>>> Greetings,
>>>>>>>
>>>>>>> I have made changes to the Async Monitor Deflation code in
>>>>>>> response to
>>>>>>> the CR8/v2.08/11-for-jdk14 code review cycle. Thanks to David
>>>>>>> H., Robbin
>>>>>>> and Erik O. for their comments!
>>>>>>>
>>>>>>> This project is no longer targeted to JDK14 so this is NOT an
>>>>>>> urgent code
>>>>>>> review request. The primary purpose of this webrev is simply to
>>>>>>> close the
>>>>>>> CR8/v2.08/11-for-jdk14 code review loop and to let folks see how
>>>>>>> I resolved
>>>>>>> the code review comments from that round.
>>>>>>>
>>>>>>> Most of the comments in the CR8/v2.08/11-for-jdk14 code review
>>>>>>> cycle were
>>>>>>> on the monitor list changes so I'm going to take a look at
>>>>>>> extracting those
>>>>>>> changes into a standalone patch. Switching from
>>>>>>> Thread::muxAcquire(&gListLock)
>>>>>>> and Thread::muxRelease(&gListLock) to finer grained internal
>>>>>>> spin locks needs
>>>>>>> to be thoroughly reviewed and the best way to do that is
>>>>>>> separately from the
>>>>>>> Async Monitor Deflation changes. Thanks to Coleen for suggesting
>>>>>>> doing this
>>>>>>> extraction earlier.
>>>>>>>
>>>>>>> I have attached the change list from CR8 to CR9 instead of
>>>>>>> putting it in
>>>>>>> the body of this email. I've also added a link to the
>>>>>>> CR8-to-CR9-changes
>>>>>>> file to the webrevs so it should be easy to find.
>>>>>>>
>>>>>>> Main bug URL:
>>>>>>>
>>>>>>> JDK-8153224 Monitor deflation prolong safepoints
>>>>>>> https://bugs.openjdk.java.net/browse/JDK-8153224
>>>>>>>
>>>>>>> The project is currently baselined on jdk-14+26.
>>>>>>>
>>>>>>> Here's the full webrev URL for those folks that want to see all
>>>>>>> of the
>>>>>>> current Async Monitor Deflation code in one go (v2.09 full):
>>>>>>>
>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/12-for-jdk14.v2.09.full/
>>>>>>>
>>>>>>>
>>>>>>> Some folks might want to see just what has changed since the
>>>>>>> last review
>>>>>>> cycle so here's a webrev for that (v2.09 inc):
>>>>>>>
>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/12-for-jdk14.v2.09.inc/
>>>>>>>
>>>>>>>
>>>>>>> The OpenJDK wiki has NOT yet been updated for this round of
>>>>>>> changes:
>>>>>>>
>>>>>>> https://wiki.openjdk.java.net/display/HotSpot/Async+Monitor+Deflation
>>>>>>>
>>>>>>>
>>>>>>> The jdk-14+26 based v2.09 version of the patch has been thru
>>>>>>> Mach5 tier[1-7]
>>>>>>> testing on Oracle's usual set of platforms. Mach5 tier8 is still
>>>>>>> running.
>>>>>>> A slightly older version of the v2.09 patch has also been
>>>>>>> through my usual
>>>>>>> set of stress testing on Linux-X64 and macOSX with the addition
>>>>>>> of Robbin's
>>>>>>> "MoCrazy 1024" test running in parallel on Linux-X64 with the
>>>>>>> other tests in
>>>>>>> my lab. The "MoCrazy 1024" has been going for > 5 days and 6700+
>>>>>>> iterations
>>>>>>> without any failures.
>>>>>>>
>>>>>>> I'm planning to do a SPECjbb2015 round on the
>>>>>>> CR9/v2.09/12-for-jdk14 bits.
>>>>>>>
>>>>>>> Thanks, in advance, for any questions, comments or suggestions.
>>>>>>>
>>>>>>> Dan
>>>>>>>
>>>>>>>
>>>>>>> On 11/4/19 4:03 PM, Daniel D. Daugherty wrote:
>>>>>>>> Greetings,
>>>>>>>>
>>>>>>>> I have made changes to the Async Monitor Deflation code in
>>>>>>>> response to
>>>>>>>> the CR7/v2.07/10-for-jdk14 code review cycle. Thanks to David
>>>>>>>> H., Robbin
>>>>>>>> and Erik O. for their comments!
>>>>>>>>
>>>>>>>> JDK14 Rampdown phase one is coming on Dec. 12, 2019 and the
>>>>>>>> Async Monitor
>>>>>>>> Deflation project needs to push before Nov. 12, 2019 in order
>>>>>>>> to allow
>>>>>>>> for sufficient bake time for such a big change. Nov. 12 is
>>>>>>>> _next_ Tuesday
>>>>>>>> so we have 8 days from today to finish this code review cycle
>>>>>>>> and push
>>>>>>>> this code for JDK14.
>>>>>>>>
>>>>>>>> Carsten and Roman! Time for you guys to chime in again on the
>>>>>>>> code reviews.
>>>>>>>>
>>>>>>>> I have attached the change list from CR7 to CR8 instead of
>>>>>>>> putting it in
>>>>>>>> the body of this email. I've also added a link to the
>>>>>>>> CR7-to-CR8-changes
>>>>>>>> file to the webrevs so it should be easy to find.
>>>>>>>>
>>>>>>>> Main bug URL:
>>>>>>>>
>>>>>>>> JDK-8153224 Monitor deflation prolong safepoints
>>>>>>>> https://bugs.openjdk.java.net/browse/JDK-8153224
>>>>>>>>
>>>>>>>> The project is currently baselined on jdk-14+21.
>>>>>>>>
>>>>>>>> Here's the full webrev URL for those folks that want to see all
>>>>>>>> of the
>>>>>>>> current Async Monitor Deflation code in one go (v2.08 full):
>>>>>>>>
>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/11-for-jdk14.v2.08.full
>>>>>>>>
>>>>>>>>
>>>>>>>> Some folks might want to see just what has changed since the
>>>>>>>> last review
>>>>>>>> cycle so here's a webrev for that (v2.08 inc):
>>>>>>>>
>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/11-for-jdk14.v2.08.inc/
>>>>>>>>
>>>>>>>>
>>>>>>>> The OpenJDK wiki did not need any changes for this round:
>>>>>>>>
>>>>>>>> https://wiki.openjdk.java.net/display/HotSpot/Async+Monitor+Deflation
>>>>>>>>
>>>>>>>>
>>>>>>>> The jdk-14+21 based v2.08 version of the patch has been thru
>>>>>>>> Mach5 tier[1-8]
>>>>>>>> testing on Oracle's usual set of platforms. It has also been
>>>>>>>> through my usual
>>>>>>>> set of stress testing on Linux-X64, macOSX and Solaris-X64 with
>>>>>>>> the addition
>>>>>>>> of Robbin's "MoCrazy 1024" test running in parallel with the
>>>>>>>> other tests in
>>>>>>>> my lab. Some testing is still running, but so far there are no
>>>>>>>> new regressions.
>>>>>>>>
>>>>>>>> I have not yet done a SPECjbb2015 round on the
>>>>>>>> CR8/v2.08/11-for-jdk14 bits.
>>>>>>>>
>>>>>>>> Thanks, in advance, for any questions, comments or suggestions.
>>>>>>>>
>>>>>>>> Dan
>>>>>>>>
>>>>>>>>
>>>>>>>> On 10/17/19 5:50 PM, Daniel D. Daugherty wrote:
>>>>>>>>> Greetings,
>>>>>>>>>
>>>>>>>>> The Async Monitor Deflation project is reaching the end game.
>>>>>>>>> I have no
>>>>>>>>> changes planned for the project at this time so all that is
>>>>>>>>> left is code
>>>>>>>>> review and any changes that results from those reviews.
>>>>>>>>>
>>>>>>>>> Carsten and Roman! Time for you guys to chime in again on the
>>>>>>>>> code reviews.
>>>>>>>>>
>>>>>>>>> I have attached the list of fixes from CR6 to CR7 instead of
>>>>>>>>> putting it
>>>>>>>>> in the main body of this email.
>>>>>>>>>
>>>>>>>>> Main bug URL:
>>>>>>>>>
>>>>>>>>> JDK-8153224 Monitor deflation prolong safepoints
>>>>>>>>> https://bugs.openjdk.java.net/browse/JDK-8153224
>>>>>>>>>
>>>>>>>>> The project is currently baselined on jdk-14+19.
>>>>>>>>>
>>>>>>>>> Here's the full webrev URL for those folks that want to see
>>>>>>>>> all of the
>>>>>>>>> current Async Monitor Deflation code in one go (v2.07 full):
>>>>>>>>>
>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/10-for-jdk14.v2.07.full
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Some folks might want to see just what has changed since the
>>>>>>>>> last review
>>>>>>>>> cycle so here's a webrev for that (v2.07 inc):
>>>>>>>>>
>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/10-for-jdk14.v2.07.inc/
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> The OpenJDK wiki has been updated to match the
>>>>>>>>> CR7/v2.07/10-for-jdk14 changes:
>>>>>>>>>
>>>>>>>>> https://wiki.openjdk.java.net/display/HotSpot/Async+Monitor+Deflation
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> The jdk-14+18 based v2.07 version of the patch has been thru
>>>>>>>>> Mach5 tier[1-8]
>>>>>>>>> testing on Oracle's usual set of platforms. It has also been
>>>>>>>>> through my usual
>>>>>>>>> set of stress testing on Linux-X64, macOSX and Solaris-X64
>>>>>>>>> with the addition
>>>>>>>>> of Robbin's "MoCrazy 1024" test running in parallel with the
>>>>>>>>> other tests in
>>>>>>>>> my lab.
>>>>>>>>>
>>>>>>>>> The jdk-14+19 based v2.07 version of the patch has been thru
>>>>>>>>> Mach5 tier[1-3]
>>>>>>>>> test on Oracle's usual set of platforms. Mach5 tier[4-8] are
>>>>>>>>> in process.
>>>>>>>>>
>>>>>>>>> I did another round of SPECjbb2015 testing in Oracle's Aurora
>>>>>>>>> Performance lab
>>>>>>>>> using using their tuned SPECjbb2015 Linux-X64 G1 configs:
>>>>>>>>>
>>>>>>>>> - "base" is jdk-14+18
>>>>>>>>> - "v2.07" is the latest version and includes C2
>>>>>>>>> inc_om_ref_count() support
>>>>>>>>> on LP64 X64 and the new
>>>>>>>>> HandshakeAfterDeflateIdleMonitors option
>>>>>>>>> - "off" is with -XX:-AsyncDeflateIdleMonitors specified
>>>>>>>>> - "handshake" is with
>>>>>>>>> -XX:+HandshakeAfterDeflateIdleMonitors specified
>>>>>>>>>
>>>>>>>>> hbIR hbIR
>>>>>>>>> (max attempted) (settled) max-jOPS critical-jOPS runtime
>>>>>>>>> --------------- --------- -------- ------------- -------
>>>>>>>>> 34282.00 30635.90 28831.30 20969.20 3841.30 base
>>>>>>>>> 34282.00 30973.00 29345.80 21025.20 3964.10 v2.07
>>>>>>>>> 34282.00 31105.60 29174.30 21074.00 3931.30
>>>>>>>>> v2.07_handshake
>>>>>>>>> 34282.00 30789.70 27151.60 19839.10 3850.20
>>>>>>>>> v2.07_off
>>>>>>>>>
>>>>>>>>> - The Aurora Perf comparison tool reports:
>>>>>>>>>
>>>>>>>>> Comparison max-jOPS critical-jOPS
>>>>>>>>> ---------------------- --------------------
>>>>>>>>> --------------------
>>>>>>>>> base vs 2.07 +1.78% (s, p=0.000) +0.27%
>>>>>>>>> (ns, p=0.790)
>>>>>>>>> base vs 2.07_handshake +1.19% (s, p=0.007) +0.58%
>>>>>>>>> (ns, p=0.536)
>>>>>>>>> base vs 2.07_off -5.83% (ns, p=0.394) -5.39%
>>>>>>>>> (ns, p=0.347)
>>>>>>>>>
>>>>>>>>> (s) - significant (ns) - not-significant
>>>>>>>>>
>>>>>>>>> - For historical comparison, the Aurora Perf comparision tool
>>>>>>>>> reported for v2.06 with a baseline of jdk-13+31:
>>>>>>>>>
>>>>>>>>> Comparison max-jOPS critical-jOPS
>>>>>>>>> ---------------------- --------------------
>>>>>>>>> --------------------
>>>>>>>>> base vs 2.06 -0.32% (ns, p=0.345) +0.71%
>>>>>>>>> (ns, p=0.646)
>>>>>>>>> base vs 2.06_off +0.49% (ns, p=0.292) -1.21%
>>>>>>>>> (ns, p=0.481)
>>>>>>>>>
>>>>>>>>> (s) - significant (ns) - not-significant
>>>>>>>>>
>>>>>>>>> Thanks, in advance, for any questions, comments or suggestions.
>>>>>>>>>
>>>>>>>>> Dan
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On 8/28/19 5:02 PM, Daniel D. Daugherty wrote:
>>>>>>>>>> Greetings,
>>>>>>>>>>
>>>>>>>>>> The Async Monitor Deflation project has rebased to JDK14 so
>>>>>>>>>> it's time
>>>>>>>>>> for our first code review in that new context!!
>>>>>>>>>>
>>>>>>>>>> I've been focused on changing the monitor list management
>>>>>>>>>> code to be
>>>>>>>>>> lock-free in order to make SPECjbb2015 happier. Of course
>>>>>>>>>> with a change
>>>>>>>>>> like that, it takes a while to chase down all the new and
>>>>>>>>>> wonderful
>>>>>>>>>> races. At this point, I have the code back to the same
>>>>>>>>>> stability that
>>>>>>>>>> I had with CR5/v2.05/8-for-jdk13.
>>>>>>>>>>
>>>>>>>>>> To lay the ground work for this round of review, I pushed the
>>>>>>>>>> following
>>>>>>>>>> two fixes to jdk/jdk earlier today:
>>>>>>>>>>
>>>>>>>>>> JDK-8230184 rename, whitespace, indent and comments
>>>>>>>>>> changes in preparation
>>>>>>>>>> for lock free Monitor lists
>>>>>>>>>> https://bugs.openjdk.java.net/browse/JDK-8230184
>>>>>>>>>>
>>>>>>>>>> JDK-8230317 serviceability/sa/ClhsdbPrintStatics.java
>>>>>>>>>> fails after 8230184
>>>>>>>>>> https://bugs.openjdk.java.net/browse/JDK-8230317
>>>>>>>>>>
>>>>>>>>>> I have attached the list of fixes from CR5 to CR6 instead of
>>>>>>>>>> putting
>>>>>>>>>> in the main body of this email.
>>>>>>>>>>
>>>>>>>>>> Main bug URL:
>>>>>>>>>>
>>>>>>>>>> JDK-8153224 Monitor deflation prolong safepoints
>>>>>>>>>> https://bugs.openjdk.java.net/browse/JDK-8153224
>>>>>>>>>>
>>>>>>>>>> The project is currently baselined on jdk-14+11 plus the
>>>>>>>>>> fixes for
>>>>>>>>>> JDK-8230184 and JDK-8230317.
>>>>>>>>>>
>>>>>>>>>> Here's the full webrev URL for those folks that want to see
>>>>>>>>>> all of the
>>>>>>>>>> current Async Monitor Deflation code in one go (v2.06 full):
>>>>>>>>>>
>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/9-for-jdk14.v2.06.full/
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> The primary focus of this review cycle is on the lock-free
>>>>>>>>>> Monitor List
>>>>>>>>>> management changes so here's a webrev for just that patch
>>>>>>>>>> (v2.06c):
>>>>>>>>>>
>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/9-for-jdk14.v2.06c.inc/
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> The secondary focus of this review cycle is on the bug fixes
>>>>>>>>>> that have
>>>>>>>>>> been made since CR5/v2.05/8-for-jdk13 so here's a webrev for
>>>>>>>>>> just that
>>>>>>>>>> patch (v2.06b):
>>>>>>>>>>
>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/9-for-jdk14.v2.06b.inc/
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> The third and final bucket for this review cycle is the
>>>>>>>>>> rename, whitespace,
>>>>>>>>>> indent and comments changes made in preparation for lock free
>>>>>>>>>> Monitor list
>>>>>>>>>> management. Almost all of that was extracted into JDK-8230184
>>>>>>>>>> for the
>>>>>>>>>> baseline so this bucket now has just a few comment changes
>>>>>>>>>> relative to
>>>>>>>>>> CR5/v2.05/8-for-jdk13. Here's a webrev for the remainder
>>>>>>>>>> (v2.06a):
>>>>>>>>>>
>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/9-for-jdk14.v2.06a.inc/
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Some folks might want to see just what has changed since the
>>>>>>>>>> last review
>>>>>>>>>> cycle so here's a webrev for that (v2.06 inc):
>>>>>>>>>>
>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/9-for-jdk14.v2.06.inc/
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Last, but not least, some folks might want to see the code
>>>>>>>>>> before the
>>>>>>>>>> addition of lock-free Monitor List management so here's a
>>>>>>>>>> webrev for
>>>>>>>>>> that (v2.00 -> v2.05):
>>>>>>>>>>
>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/9-for-jdk14.v2.05.inc/
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> The OpenJDK wiki will need minor updates to match the CR6
>>>>>>>>>> changes:
>>>>>>>>>>
>>>>>>>>>> https://wiki.openjdk.java.net/display/HotSpot/Async+Monitor+Deflation
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> but that should only be changes to describe per-thread list
>>>>>>>>>> async monitor
>>>>>>>>>> deflation being done by the ServiceThread.
>>>>>>>>>>
>>>>>>>>>> (I did update the OpenJDK wiki for the CR5 changes back on
>>>>>>>>>> 2019.08.14)
>>>>>>>>>>
>>>>>>>>>> This version of the patch has been thru Mach5 tier[1-8]
>>>>>>>>>> testing on
>>>>>>>>>> Oracle's usual set of platforms. It has also been through my
>>>>>>>>>> usual set
>>>>>>>>>> of stress testing on Linux-X64, macOSX and Solaris-X64.
>>>>>>>>>>
>>>>>>>>>> I did a bunch of SPECjbb2015 testing in Oracle's Aurora
>>>>>>>>>> Performance lab
>>>>>>>>>> using using their tuned SPECjbb2015 Linux-X64 G1 configs.
>>>>>>>>>> This was using
>>>>>>>>>> this patch baselined on jdk-13+31 (for stability):
>>>>>>>>>>
>>>>>>>>>> hbIR hbIR
>>>>>>>>>> (max attempted) (settled) max-jOPS critical-jOPS runtime
>>>>>>>>>> --------------- --------- -------- ------------- -------
>>>>>>>>>> 34282.00 28837.20 27905.20 19817.40 3658.10 base
>>>>>>>>>> 34965.70 29798.80 27814.90 19959.00 3514.60 v2.06d
>>>>>>>>>> 34282.00 29100.70 28042.50 19577.00 3701.90
>>>>>>>>>> v2.06d_off
>>>>>>>>>> 34282.00 29218.50 27562.80 19397.30 3657.60
>>>>>>>>>> v2.06d_ocache
>>>>>>>>>> 34965.70 29838.30 26512.40 19170.60 3569.90 v2.05
>>>>>>>>>> 34282.00 28926.10 27734.00 19835.10 3588.40
>>>>>>>>>> v2.05_off
>>>>>>>>>>
>>>>>>>>>> The "off" configs are with -XX:-AsyncDeflateIdleMonitors
>>>>>>>>>> specified and
>>>>>>>>>> the "ocache" config is with 128 byte cache line sizes instead
>>>>>>>>>> of 64 byte
>>>>>>>>>> cache lines sizes. "v2.06d" is the last set of changes that I
>>>>>>>>>> made before
>>>>>>>>>> those changes were distributed into the "v2.06a", "v2.06b"
>>>>>>>>>> and "v2.06c"
>>>>>>>>>> buckets for this review recycle.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Thanks, in advance, for any questions, comments or suggestions.
>>>>>>>>>>
>>>>>>>>>> Dan
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> On 7/11/19 3:49 PM, Daniel D. Daugherty wrote:
>>>>>>>>>>> Greetings,
>>>>>>>>>>>
>>>>>>>>>>> I've been focused on chasing down and fixing the rare test
>>>>>>>>>>> failures
>>>>>>>>>>> that only pop up rarely. So this round is primarily fixes
>>>>>>>>>>> for races
>>>>>>>>>>> with a few additional fixes that came from Karen's review of
>>>>>>>>>>> CR4.
>>>>>>>>>>> Thanks Karen!
>>>>>>>>>>>
>>>>>>>>>>> I have attached the list of fixes from CR4 to CR5 instead of
>>>>>>>>>>> putting
>>>>>>>>>>> in the main body of this email.
>>>>>>>>>>>
>>>>>>>>>>> Main bug URL:
>>>>>>>>>>>
>>>>>>>>>>> JDK-8153224 Monitor deflation prolong safepoints
>>>>>>>>>>> https://bugs.openjdk.java.net/browse/JDK-8153224
>>>>>>>>>>>
>>>>>>>>>>> The project is currently baselined on jdk-13+29. This will
>>>>>>>>>>> likely be
>>>>>>>>>>> the last JDK13 baseline for this project and I'll roll to
>>>>>>>>>>> the JDK14
>>>>>>>>>>> (jdk/jdk) repo soon...
>>>>>>>>>>>
>>>>>>>>>>> Here's the full webrev URL:
>>>>>>>>>>>
>>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/8-for-jdk13.full/
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Here's the incremental webrev URL:
>>>>>>>>>>>
>>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/8-for-jdk13.inc/
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> I have not yet checked the OpenJDK wiki to see if it needs
>>>>>>>>>>> any updates
>>>>>>>>>>> to match the CR5 changes:
>>>>>>>>>>>
>>>>>>>>>>> https://wiki.openjdk.java.net/display/HotSpot/Async+Monitor+Deflation
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> (I did update the OpenJDK wiki for the CR4 changes back on
>>>>>>>>>>> 2019.06.26)
>>>>>>>>>>>
>>>>>>>>>>> This version of the patch has been thru Mach5 tier[1-3]
>>>>>>>>>>> testing on
>>>>>>>>>>> Oracle's usual set of platforms. Mach5 tier[4-6] is running
>>>>>>>>>>> now and
>>>>>>>>>>> Mach5 tier[78] will follow. I'll kick off the usual stress
>>>>>>>>>>> testing
>>>>>>>>>>> on Linux-X64, macOSX and Solaris-X64 as those machines
>>>>>>>>>>> become available.
>>>>>>>>>>> Since I haven't made any performance changes in this round,
>>>>>>>>>>> I'll only
>>>>>>>>>>> be running SPECjbb2015 to gather the latest monitorinflation
>>>>>>>>>>> logs.
>>>>>>>>>>>
>>>>>>>>>>> Next up:
>>>>>>>>>>>
>>>>>>>>>>> - We're still seeing 4-5% lower performance with SPECjbb2015 on
>>>>>>>>>>> Linux-X64 and we've determined that some of that comes from
>>>>>>>>>>> contention on the gListLock. So I'm going to investigate
>>>>>>>>>>> removing
>>>>>>>>>>> the gListLock. Yes, another lock free set of changes is
>>>>>>>>>>> coming!
>>>>>>>>>>> - Of course, going lock free often causes new races and new
>>>>>>>>>>> failures
>>>>>>>>>>> so that's a good reason for make those changes isolated in
>>>>>>>>>>> their
>>>>>>>>>>> own round (and not holding up CR5/v2.05/8-for-jdk13 anymore).
>>>>>>>>>>> - I finally have a potential fix for the Win* failure with
>>>>>>>>>>> gc/g1/humongousObjects/TestHumongousClassLoader.java
>>>>>>>>>>> but I haven't run it through Mach5 yet so it'll be in the
>>>>>>>>>>> next round.
>>>>>>>>>>> - Some RTM tests were recently re-enabled in Mach5 and I'm
>>>>>>>>>>> seeing some
>>>>>>>>>>> monitor related failures there. I suspect that I need to
>>>>>>>>>>> go take a
>>>>>>>>>>> look at the C2 RTM macro assembler code and look for
>>>>>>>>>>> things that might
>>>>>>>>>>> conflict if Async Monitor Deflation. If you're interested
>>>>>>>>>>> in that kind
>>>>>>>>>>> of issue, then see the macroAssembler_x86.cpp sanity check
>>>>>>>>>>> that I
>>>>>>>>>>> added in this round!
>>>>>>>>>>>
>>>>>>>>>>> Thanks, in advance, for any questions, comments or suggestions.
>>>>>>>>>>>
>>>>>>>>>>> Dan
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> On 5/26/19 8:30 PM, Daniel D. Daugherty wrote:
>>>>>>>>>>>> Greetings,
>>>>>>>>>>>>
>>>>>>>>>>>> I have a fix for an issue that came up during performance
>>>>>>>>>>>> testing.
>>>>>>>>>>>> Many thanks to Robbin for diagnosing the issue in his
>>>>>>>>>>>> SPECjbb2015
>>>>>>>>>>>> experiments.
>>>>>>>>>>>>
>>>>>>>>>>>> Here's the list of changes from CR3 to CR4. The list is a bit
>>>>>>>>>>>> verbose due to the complexity of the issue, but the changes
>>>>>>>>>>>> themselves are not that big.
>>>>>>>>>>>>
>>>>>>>>>>>> Functional:
>>>>>>>>>>>> - Change SafepointSynchronize::is_cleanup_needed() from
>>>>>>>>>>>> calling
>>>>>>>>>>>> ObjectSynchronizer::is_cleanup_needed() to calling
>>>>>>>>>>>> ObjectSynchronizer::is_safepoint_deflation_needed():
>>>>>>>>>>>> - is_safepoint_deflation_needed() returns the result of
>>>>>>>>>>>> monitors_used_above_threshold() for safepoint based
>>>>>>>>>>>> monitor deflation (!AsyncDeflateIdleMonitors).
>>>>>>>>>>>> - For AsyncDeflateIdleMonitors, it only returns true if
>>>>>>>>>>>> there is a special deflation request, e.g., System.gc()
>>>>>>>>>>>> - This solves a bug where there are a bunch of Cleanup
>>>>>>>>>>>> safepoints that simply request async deflation which
>>>>>>>>>>>> keeps the async JavaThreads from making progress on
>>>>>>>>>>>> their async deflation work.
>>>>>>>>>>>> - Add AsyncDeflationInterval diagnostic option. Description:
>>>>>>>>>>>> Async deflate idle monitors every so many
>>>>>>>>>>>> milliseconds when
>>>>>>>>>>>> MonitorUsedDeflationThreshold is exceeded (0 is off).
>>>>>>>>>>>> - Replace
>>>>>>>>>>>> ObjectSynchronizer::gOmShouldDeflateIdleMonitors() with
>>>>>>>>>>>> ObjectSynchronizer::is_async_deflation_needed():
>>>>>>>>>>>> - is_async_deflation_needed() returns true when
>>>>>>>>>>>> is_async_cleanup_requested() is true or when
>>>>>>>>>>>> monitors_used_above_threshold() is true (but no more often
>>>>>>>>>>>> than
>>>>>>>>>>>> AsyncDeflationInterval).
>>>>>>>>>>>> - if AsyncDeflateIdleMonitors Service_lock->wait() now
>>>>>>>>>>>> waits for
>>>>>>>>>>>> at most GuaranteedSafepointInterval millis:
>>>>>>>>>>>> - This allows is_async_deflation_needed() to be
>>>>>>>>>>>> checked at
>>>>>>>>>>>> the same interval as GuaranteedSafepointInterval.
>>>>>>>>>>>> (default is 1000 millis/1 second)
>>>>>>>>>>>> - Once is_async_deflation_needed() has returned true, it
>>>>>>>>>>>> generally cannot return true for
>>>>>>>>>>>> AsyncDeflationInterval.
>>>>>>>>>>>> This is to prevent async deflation from swamping the
>>>>>>>>>>>> ServiceThread.
>>>>>>>>>>>> - The ServiceThread still handles async deflation of the
>>>>>>>>>>>> global
>>>>>>>>>>>> in-use list and now it also marks JavaThreads for async
>>>>>>>>>>>> deflation
>>>>>>>>>>>> of their in-use lists.
>>>>>>>>>>>> - The ServiceThread will check for async deflation work
>>>>>>>>>>>> every
>>>>>>>>>>>> GuaranteedSafepointInterval.
>>>>>>>>>>>> - A safepoint can still cause the ServiceThread to
>>>>>>>>>>>> check for
>>>>>>>>>>>> async deflation work via is_async_deflation_requested.
>>>>>>>>>>>> - Refactor code from
>>>>>>>>>>>> ObjectSynchronizer::is_cleanup_needed() into
>>>>>>>>>>>> monitors_used_above_threshold() and remove
>>>>>>>>>>>> is_cleanup_needed().
>>>>>>>>>>>> - In addition to System.gc(), the VM_Exit VM op and the
>>>>>>>>>>>> final
>>>>>>>>>>>> VMThread safepoint now set the
>>>>>>>>>>>> is_special_deflation_requested
>>>>>>>>>>>> flag to reduce the in-use monitor population that is
>>>>>>>>>>>> reported by
>>>>>>>>>>>> ObjectSynchronizer::log_in_use_monitor_details() at VM exit.
>>>>>>>>>>>>
>>>>>>>>>>>> Test update:
>>>>>>>>>>>> - test/hotspot/gtest/oops/test_markOop.cpp is updated to
>>>>>>>>>>>> work with
>>>>>>>>>>>> AsyncDeflateIdleMonitors.
>>>>>>>>>>>>
>>>>>>>>>>>> Collateral:
>>>>>>>>>>>> - Add/clarify/update some logging messages.
>>>>>>>>>>>>
>>>>>>>>>>>> Cleanup:
>>>>>>>>>>>> - Updated comments based on Karen's code review.
>>>>>>>>>>>> - Change 'special cleanup' -> 'special deflation' and
>>>>>>>>>>>> 'async cleanup' -> 'async deflation'.
>>>>>>>>>>>> - comment and function name changes
>>>>>>>>>>>> - Clarify MonitorUsedDeflationThreshold description;
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> Main bug URL:
>>>>>>>>>>>>
>>>>>>>>>>>> JDK-8153224 Monitor deflation prolong safepoints
>>>>>>>>>>>> https://bugs.openjdk.java.net/browse/JDK-8153224
>>>>>>>>>>>>
>>>>>>>>>>>> The project is currently baselined on jdk-13+22.
>>>>>>>>>>>>
>>>>>>>>>>>> Here's the full webrev URL:
>>>>>>>>>>>>
>>>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/7-for-jdk13.full/
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> Here's the incremental webrev URL:
>>>>>>>>>>>>
>>>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/7-for-jdk13.inc/
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> I have not updated the OpenJDK wiki to reflect the CR4
>>>>>>>>>>>> changes:
>>>>>>>>>>>>
>>>>>>>>>>>> https://wiki.openjdk.java.net/display/HotSpot/Async+Monitor+Deflation
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> The wiki doesn't say a whole lot about the async deflation
>>>>>>>>>>>> invocation
>>>>>>>>>>>> mechanism so I have to figure out how to add that content.
>>>>>>>>>>>>
>>>>>>>>>>>> This version of the patch has been thru Mach5 tier[1-8]
>>>>>>>>>>>> testing on
>>>>>>>>>>>> Oracle's usual set of platforms. My Solaris-X64 stress kit
>>>>>>>>>>>> run is
>>>>>>>>>>>> running now. Kitchensink8H on product, fastdebug, and
>>>>>>>>>>>> slowdebug bits
>>>>>>>>>>>> are running on Linux-X64, MacOSX and Solaris-X64. I still
>>>>>>>>>>>> have to run
>>>>>>>>>>>> my stress kit on Linux-X64. I still have to run the
>>>>>>>>>>>> SPECjbb2015
>>>>>>>>>>>> baseline and CR4 runs on Linux-X64, MacOSX and Solaris-X64.
>>>>>>>>>>>>
>>>>>>>>>>>> Thanks, in advance, for any questions, comments or
>>>>>>>>>>>> suggestions.
>>>>>>>>>>>>
>>>>>>>>>>>> Dan
>>>>>>>>>>>>
>>>>>>>>>>>> On 5/6/19 11:52 AM, Daniel D. Daugherty wrote:
>>>>>>>>>>>>> Greetings,
>>>>>>>>>>>>>
>>>>>>>>>>>>> I had some discussions with Karen about a race that was in
>>>>>>>>>>>>> the
>>>>>>>>>>>>> ObjectMonitor::enter() code in CR2/v2.02/5-for-jdk13. This
>>>>>>>>>>>>> race was
>>>>>>>>>>>>> theoretical and I had no test failures due to it. The fix
>>>>>>>>>>>>> is pretty
>>>>>>>>>>>>> simple: remove the special case code for async deflation
>>>>>>>>>>>>> in the
>>>>>>>>>>>>> ObjectMonitor::enter() function and rely solely on the
>>>>>>>>>>>>> ref_count
>>>>>>>>>>>>> for ObjectMonitor::enter() protection.
>>>>>>>>>>>>>
>>>>>>>>>>>>> During those discussions Karen also floated the idea of
>>>>>>>>>>>>> using the
>>>>>>>>>>>>> ref_count field instead of the contentions field for the
>>>>>>>>>>>>> Async
>>>>>>>>>>>>> Monitor Deflation protocol. I decided to go ahead and code
>>>>>>>>>>>>> up that
>>>>>>>>>>>>> change and I have run it through the usual stress and
>>>>>>>>>>>>> Mach5 testing
>>>>>>>>>>>>> with no issues. It's also known as v2.03 (for those for
>>>>>>>>>>>>> with the
>>>>>>>>>>>>> patches) and as webrev/6-for-jdk13 (for those with webrev
>>>>>>>>>>>>> URLs).
>>>>>>>>>>>>> Sorry for all the names...
>>>>>>>>>>>>>
>>>>>>>>>>>>> Main bug URL:
>>>>>>>>>>>>>
>>>>>>>>>>>>> JDK-8153224 Monitor deflation prolong safepoints
>>>>>>>>>>>>> https://bugs.openjdk.java.net/browse/JDK-8153224
>>>>>>>>>>>>>
>>>>>>>>>>>>> The project is currently baselined on jdk-13+18.
>>>>>>>>>>>>>
>>>>>>>>>>>>> Here's the full webrev URL:
>>>>>>>>>>>>>
>>>>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/6-for-jdk13.full/
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> Here's the incremental webrev URL:
>>>>>>>>>>>>>
>>>>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/6-for-jdk13.inc/
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> I have also updated the OpenJDK wiki to reflect the CR3
>>>>>>>>>>>>> changes:
>>>>>>>>>>>>>
>>>>>>>>>>>>> https://wiki.openjdk.java.net/display/HotSpot/Async+Monitor+Deflation
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> This version of the patch has been thru Mach5 tier[1-8]
>>>>>>>>>>>>> testing on
>>>>>>>>>>>>> Oracle's usual set of platforms. My Solaris-X64 stress kit
>>>>>>>>>>>>> run had
>>>>>>>>>>>>> no issues. Kitchensink8H on product, fastdebug, and
>>>>>>>>>>>>> slowdebug bits
>>>>>>>>>>>>> had no failures on Linux-X64; MacOSX fastdebug and
>>>>>>>>>>>>> slowdebug and
>>>>>>>>>>>>> Solaris-X64 release had the usual "Too large time diff"
>>>>>>>>>>>>> complaints.
>>>>>>>>>>>>> 12 hour Inflate2 runs on product, fastdebug and slowdebug
>>>>>>>>>>>>> bits on
>>>>>>>>>>>>> Linux-X64, MacOSX and Solaris-X64 had no failures. My
>>>>>>>>>>>>> Linux-X64
>>>>>>>>>>>>> stress kit is running right now.
>>>>>>>>>>>>>
>>>>>>>>>>>>> I've done the SPECjbb2015 baseline and CR3 runs. I need to
>>>>>>>>>>>>> gather
>>>>>>>>>>>>> the results and analyze them.
>>>>>>>>>>>>>
>>>>>>>>>>>>> Thanks, in advance, for any questions, comments or
>>>>>>>>>>>>> suggestions.
>>>>>>>>>>>>>
>>>>>>>>>>>>> Dan
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> On 4/25/19 12:38 PM, Daniel D. Daugherty wrote:
>>>>>>>>>>>>>> Greetings,
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> I have a small but important bug fix for the Async
>>>>>>>>>>>>>> Monitor Deflation
>>>>>>>>>>>>>> project ready to go. It's also known as v2.02 (for those
>>>>>>>>>>>>>> for with the
>>>>>>>>>>>>>> patches) and as webrev/5-for-jdk13 (for those with webrev
>>>>>>>>>>>>>> URLs). Sorry
>>>>>>>>>>>>>> for all the names...
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> JDK-8222295 was pushed to jdk/jdk two days ago so that
>>>>>>>>>>>>>> baseline patch
>>>>>>>>>>>>>> is out of our hair.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Main bug URL:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> JDK-8153224 Monitor deflation prolong safepoints
>>>>>>>>>>>>>> https://bugs.openjdk.java.net/browse/JDK-8153224
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> The project is currently baselined on jdk-13+17.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Here's the full webrev URL:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/5-for-jdk13.full/
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Here's the incremental webrev URL (JDK-8153224):
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/5-for-jdk13.inc/
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> I still have to update the OpenJDK wiki to reflect the
>>>>>>>>>>>>>> CR2 changes:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> https://wiki.openjdk.java.net/display/HotSpot/Async+Monitor+Deflation
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> This version of the patch has been thru Mach5 tier[1-6]
>>>>>>>>>>>>>> testing on
>>>>>>>>>>>>>> Oracle's usual set of platforms. Mach5 tier[7-8] is
>>>>>>>>>>>>>> running now.
>>>>>>>>>>>>>> My stress kit is running on Solaris-X64 now.
>>>>>>>>>>>>>> Kitchensink8H is running
>>>>>>>>>>>>>> now on product, fastdebug, and slowdebug bits on
>>>>>>>>>>>>>> Linux-X64, MacOSX
>>>>>>>>>>>>>> and Solaris-X64. 12 hour Inflate2 runs are running now on
>>>>>>>>>>>>>> product,
>>>>>>>>>>>>>> fastdebug and slowdebug bits on Linux-X64, MacOSX and
>>>>>>>>>>>>>> Solaris-X64.
>>>>>>>>>>>>>> I'll start my my stress kit on Linux-X64 sometime on
>>>>>>>>>>>>>> Sunday (after
>>>>>>>>>>>>>> my jdk-13+18 stress run is done).
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> I'll do SPECjbb2015 baseline and CR2 runs after all the
>>>>>>>>>>>>>> stress
>>>>>>>>>>>>>> testing is done.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Thanks, in advance, for any questions, comments or
>>>>>>>>>>>>>> suggestions.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Dan
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> On 4/19/19 11:58 AM, Daniel D. Daugherty wrote:
>>>>>>>>>>>>>>> Greetings,
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> I finally have CR1 for the Async Monitor Deflation
>>>>>>>>>>>>>>> project ready to
>>>>>>>>>>>>>>> go. It's also known as v2.01 (for those for with the
>>>>>>>>>>>>>>> patches) and as
>>>>>>>>>>>>>>> webrev/4-for-jdk13 (for those with webrev URLs). Sorry
>>>>>>>>>>>>>>> for all the
>>>>>>>>>>>>>>> names...
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Main bug URL:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> JDK-8153224 Monitor deflation prolong safepoints
>>>>>>>>>>>>>>> https://bugs.openjdk.java.net/browse/JDK-8153224
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Baseline bug fixes URL:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> JDK-8222295 more baseline cleanups from Async
>>>>>>>>>>>>>>> Monitor Deflation project
>>>>>>>>>>>>>>> https://bugs.openjdk.java.net/browse/JDK-8222295
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> The project is currently baselined on jdk-13+15.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Here's the webrev for the latest baseline changes
>>>>>>>>>>>>>>> (JDK-8222295):
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/4-for-jdk13.8222295
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Here's the full webrev URL (JDK-8153224 only):
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/4-for-jdk13.full/
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Here's the incremental webrev URL (JDK-8153224):
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/4-for-jdk13.inc/
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> So I'm looking for reviews for both JDK-8222295 and the
>>>>>>>>>>>>>>> latest version
>>>>>>>>>>>>>>> of JDK-8153224...
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> I still have to update the OpenJDK wiki to reflect the
>>>>>>>>>>>>>>> CR changes:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> https://wiki.openjdk.java.net/display/HotSpot/Async+Monitor+Deflation
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> This version of the patch has been thru Mach5 tier[1-3]
>>>>>>>>>>>>>>> testing on
>>>>>>>>>>>>>>> Oracle's usual set of platforms. Mach5 tier[4-6] is
>>>>>>>>>>>>>>> running now and
>>>>>>>>>>>>>>> Mach5 tier[78] will be run later today. My stress kit on
>>>>>>>>>>>>>>> Solaris-X64
>>>>>>>>>>>>>>> is running now. Linux-X64 stress testing will start on
>>>>>>>>>>>>>>> Sunday. I'm
>>>>>>>>>>>>>>> planning to do Kitchensink runs, SPECjbb2015 runs and my
>>>>>>>>>>>>>>> monitor
>>>>>>>>>>>>>>> inflation stress tests on Linux-X64, MacOSX and
>>>>>>>>>>>>>>> Solaris-X64.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Thanks, in advance, for any questions, comments or
>>>>>>>>>>>>>>> suggestions.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Dan
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> On 3/24/19 9:57 AM, Daniel D. Daugherty wrote:
>>>>>>>>>>>>>>>> Greetings,
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Welcome to the OpenJDK review thread for my port of
>>>>>>>>>>>>>>>> Carsten's work on:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> JDK-8153224 Monitor deflation prolong safepoints
>>>>>>>>>>>>>>>> https://bugs.openjdk.java.net/browse/JDK-8153224
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Here's a link to the OpenJDK wiki that describes my port:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> https://wiki.openjdk.java.net/display/HotSpot/Async+Monitor+Deflation
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Here's the webrev URL:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> http://cr.openjdk.java.net/~dcubed/8153224-webrev/3-for-jdk13/
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Here's a link to Carsten's original webrev:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> http://cr.openjdk.java.net/~cvarming/monitor_deflate_conc/0/
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Earlier versions of this patch have been through
>>>>>>>>>>>>>>>> several rounds of
>>>>>>>>>>>>>>>> preliminary review. Many thanks to Carsten, Coleen,
>>>>>>>>>>>>>>>> Robbin, and
>>>>>>>>>>>>>>>> Roman for their preliminary code review comments. A
>>>>>>>>>>>>>>>> very special
>>>>>>>>>>>>>>>> thanks to Robbin and Roman for building and testing the
>>>>>>>>>>>>>>>> patch in
>>>>>>>>>>>>>>>> their own environments (including specJBB2015).
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> This version of the patch has been thru Mach5 tier[1-8]
>>>>>>>>>>>>>>>> testing on
>>>>>>>>>>>>>>>> Oracle's usual set of platforms. Earlier versions have
>>>>>>>>>>>>>>>> been run
>>>>>>>>>>>>>>>> through my stress kit on my Linux-X64 and Solaris-X64
>>>>>>>>>>>>>>>> servers
>>>>>>>>>>>>>>>> (product, fastdebug, slowdebug).Earlier versions have
>>>>>>>>>>>>>>>> run Kitchensink
>>>>>>>>>>>>>>>> for 12 hours on MacOSX, Linux-X64 and Solaris-X64
>>>>>>>>>>>>>>>> (product, fastdebug
>>>>>>>>>>>>>>>> and slowdebug). Earlier versions have run my monitor
>>>>>>>>>>>>>>>> inflation stress
>>>>>>>>>>>>>>>> tests for 12 hours on MacOSX, Linux-X64 and Solaris-X64
>>>>>>>>>>>>>>>> (product,
>>>>>>>>>>>>>>>> fastdebug and slowdebug).
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> All of the testing done on earlier versions will be
>>>>>>>>>>>>>>>> redone on the
>>>>>>>>>>>>>>>> latest version of the patch.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Thanks, in advance, for any questions, comments or
>>>>>>>>>>>>>>>> suggestions.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Dan
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> P.S.
>>>>>>>>>>>>>>>> One subtest in
>>>>>>>>>>>>>>>> gc/g1/humongousObjects/TestHumongousClassLoader.java
>>>>>>>>>>>>>>>> is currently failing in -Xcomp mode on Win* only. I've
>>>>>>>>>>>>>>>> been trying
>>>>>>>>>>>>>>>> to characterize/analyze this failure for more than a
>>>>>>>>>>>>>>>> week now. At
>>>>>>>>>>>>>>>> this point I'm convinced that Async Monitor Deflation
>>>>>>>>>>>>>>>> is aggravating
>>>>>>>>>>>>>>>> an existing bug. However, I plan to have a better
>>>>>>>>>>>>>>>> handle on that
>>>>>>>>>>>>>>>> failure before these bits are pushed to the jdk/jdk repo.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>
More information about the hotspot-runtime-dev
mailing list