This appears to improve cpu usage since launching a new thread is more expensive than performing a small amount of inline logic.
However, needs testing at scale.
When Melanie added the web fetch inventory throttle to core, she made the long poll requests (EQs) effectively be handled on an active loop. All those requests, if they existed, were being constantly dequeued, checked for events (which most often they didn't have), and requeued again. This was an active loop thread on a 100ms cycle!
This fixes the issue. Now the inventory requests, if they aren't ready to be served, are placed directly back in the queue, but the long poll requests aren't placed there until there are events ready to be sent or timeout has been reached.
This puts the LongPollServiceWatcherThread back to 1sec cycle, as it was before.
This reverts commit 21a09ad3ad.
After more analysis and discussion, it is apparant that the Count(), Contains() and GetQueueArray() cannot be made thread-safe anyway without external locking
And this change appears to have a positive impact on performance.
I still believe that Monitor.Exit() will not release any thread for Monitor.Wait(), as per http://msdn.microsoft.com/en-gb/library/vstudio/system.threading.monitor.exit%28v=vs.100%29.aspx
so this should in theory make no difference, though mono implementation issues could possibly be coming into play.
This reverts commit 42e2a0d66e
Reverting because unfortunately this introduces race conditions because Contains(), Count() and GetQueueArray() may now end up returning the wrong result if another thread performs a simultaneous update on m_queue.
Code such as PollServiceRequestManager.Stop() relies on the count being correct otherwise a request may be lost.
Also, though some of the internal queue methods do not affect state, they are not thread-safe and could return the wrong result generating the same problem
lock() generates Monitor.Enter() and Monitor.Exit() under the covers. Monitor.Exit() does not cause Monitor.Wait() to exist, only Pulse() and PulseAll() will do this
Reverted with agreement.
This adds explicit cap poll handler supporting to the Caps classes rather than relying on callers to do the complicated coding.
Other refactoring was required to get logic into the right places to support this.
17:14:28 - [APPLICATION]:
APPLICATION EXCEPTION DETECTED: System.UnhandledExceptionEventArgs
Exception: System.InvalidOperationException: Operation is not valid due to the current state of the object
at System.Collections.Generic.Queue`1[OpenSim.Region.ClientStack.Linden.WebFetchInvDescModule+aPollRequest].Peek () [0x00011] in /root/install/mono-3.1.0/mono/mcs/class/System/System.Collections.Generic/Queue.cs:158
at System.Collections.Generic.Queue`1[OpenSim.Region.ClientStack.Linden.WebFetchInvDescModule+aPollRequest].Dequeue () [0x00000] in /root/install/mono-3.1.0/mono/mcs/class/System/System.Collections.Generic/Queue.cs:140
at OpenSim.Framework.DoubleQueue`1[OpenSim.Region.ClientStack.Linden.WebFetchInvDescModule+aPollRequest].Dequeue (TimeSpan wait, OpenSim.Region.ClientStack.Linden.aPollRequest& res) [0x0004e] in /home/avacon/opensim_2013-07-14/OpenSim/Framework/Util.cs:2297
Revert "Trying to reduce CPU usage on logins and TPs: trying radical elimination of all FireAndForgets throughout CompleteMovement. There were 4."
This reverts commit 6825377380.
BestAvatarResponsiveness introduces the region rez delay in cases where the region is full of avatars with lots of attachments, which is the case in CC load tests. In that case, the inworld prims are sent only after all avatar attachments are sent. Not recommended for regions with heavy avatar traffic!
Justin, if you read this, there's a long story here. Some time ago you placed SendInitialDataToMe at the very beginning of client creation (in LLUDPServer). That is problematic, as we discovered relatively recently: on TPs, as soon as the client starts getting data from child agents, it starts requesting resources back *from the simulator where its root agent is*. We found this to be the problem behind meshes missing on HG TPs (because the viewer was requesting the meshes of the receiving sim from the departing grid). But this affects much more than meshes and HG TPs. It may also explain cloud avatars after a local TP: baked textures are only stored in the simulator, so if a child agent receives a UUID of a baked texture in the destination sim and requests that texture from the departing sim where the root agent is, it will fail to get that texture.
Bottom line: we need to delay sending the new simulator data to the viewer until we are absolutely sure that the viewer knows that its main agent is in a new sim. Hence, moving it to CompleteMovement.
Now I am trying to tune the initial rez delay that we all experience in the CC. I think that when I fixed the issue described above, I may have moved SendInitialDataToMe to much later than it should be, so now I'm moving to earlier in CompleteMovement.
A separate PhysicsActor variable is used in case some other thread removes the PhysicsActor whilst this code is executing.
If this is now impossible please revert - just adding this now whilst I remember.
Also makes method comment into proper method doc.
This was probably the mistake.
The other handlers are named RenderMaterials as well but this actully has no affect apart from on stats, due to a (counterintuitive) disconnect between the registration name and the name of the request handler.
Will be tested very soon and reverted if this still does not work.
This was a regression since commit 831e4c3 (Thu Apr 4 00:36:15 2013)
This commit also adds a regression test for this case, though this currently only works with Mono
This aims to address http://opensimulator.org/mantis/view.php?id=6704
reception thread to sleep for 30ms if the number of available user worker
threads got low. It doesn't look like any of the UDP packet types are
marked async so this check is 1) unnecessary and 2) really crazy since
it stops up the reception thread under heavy load without any indication.
Justin Clark-Casey (justincc)