Commit Graph

581 Commits (838bc80ab9273c2834794535886a86c7574bb0d3)

Author SHA1 Message Date
Adam Frisby 838bc80ab9 * Implemented some tweaks to monitoring module.
* Output is prettier & more useful.
* Added 'Alerts' to allow rules to be constructed using Monitors to detect for events such as deadlocks. This will be translated to SNMP Traps when I get SNMP implemented.
2009-11-02 00:05:49 +11:00
Adam Frisby 711dde34e4 * Implements new 'Monitoring' system for reporting performance.
* Mostly the same set as the StatsMonitor used for Viewer notification, but exposes some new frametimes - including EventMS, PhysicsUpdateMS, LandUpdateMS; new memory monitoring - both GC.TotalMemory and Process.PrivateWorkingMemory64; also exposes ThreadCount (using System.Diagnostics.Process)
* Type 'monitor report' on the console to see output.
* SNMP Implementation forthcoming.
2009-11-01 19:37:40 +11:00
John Hurliman 8a73dc0f8a * Fix for a potential race condition in ScenePresence.AbsolutePosition
* Unified the way region handles are stored and used in ScenePresence
* Fixed camera position for child agents
* CheckForSignificantMovement now checks avatar and camera position (both are important for scene prioritization)
* Removing debug code from the previous commit
2009-10-30 03:01:15 -07:00
John Hurliman aecaa51063 * Fixed a NullReferenceException in GetMovementAnimation() and added more protection against NREs in AddNewMovement()
* Removed the three second limit on ImprovedTerseObjectUpdate. With the latest fixes I don't think this is necessary, and it generates a lot of unnecessary updates in a crowded sim
2009-10-29 16:31:48 -07:00
John Hurliman 2913c24c8a * Commented out two noisy debug lines in the LLUDP server
* Misc. cleanup in ScenePresence.HandleAgentUpdate()
2009-10-29 15:24:31 -07:00
John Hurliman 5c894dac8b * Implemented foot collision plane for avatars
* Fixed a NullReferenceException regression
2009-10-29 05:34:40 -07:00
John Hurliman 3f2d6fe470 Ported the Simian avatar animation system to OpenSim. Landing is currently not working 2009-10-29 04:13:51 -07:00
John Hurliman fd2c99f184 Fixing NullReferenceException regression in the previous commit 2009-10-29 02:10:48 -07:00
John Hurliman 7132877075 * Log progress messages when loading OAR files with a lot of assets
* Change the PhysicsCollision callback for objects to send full contact point information. This will be used to calculate the collision plane for avatars
* Send the physics engine velocity in terse updates, not the current force being applied to the avatar. This should fix several issues including crouching through the floor and walking through walls
2009-10-29 01:46:58 -07:00
John Hurliman 59eb378d16 Small performance tweaks to code called by the heartbeat loop 2009-10-28 23:10:16 -07:00
John Hurliman 1c9696a9d2 Always send a time dilation of 1.0 while we debug rubberbanding issues 2009-10-28 15:11:01 -07:00
John Hurliman a069a1ee68 Limit physics time dilation to 1.0 2009-10-28 14:44:05 -07:00
John Hurliman b81c829576 * Standalone logins will now go through the sequence of "requested region, default region, any region" before giving up
* Hip offset should have been added not subtracted (it's a negative offset). This puts avatar feet closer to the ground
* Improved duplicate checking for terse updates. This should reduce bandwidth and walking through walls
2009-10-28 14:13:17 -07:00
John Hurliman a65c8cdc38 * Reduce the velocity tolerance on sending terse updates to avoid slowly drifting prims/avatars
* Added contacts_per_collision to the ODE config section. This allows you to reduce the maximum number of contact points ODE will generate per collision and reduce the size of the array that stores contact structures
2009-10-28 12:45:40 -07:00
John Hurliman cdbeb8b83b Track timestamps when terse updates were last sent for a prim or avatar to avoid floating away forever until a key is pressed (deviates from SL behavior in a hopefully good way) 2009-10-28 03:21:53 -07:00
John Hurliman 0c466b28bb Move the calculation of time dilation from the scene to the physics engine. The scene is still the one reporting dilation so this does not break the API or remove flexibility, but it gets the calculation happening in the right place for the normal OpenSim usage. The actual calculation of physics time dilation probably needs tweaking 2009-10-27 16:24:43 -07:00
John Hurliman fefe767476 Lowering the position tolerance of terse updates for ScenePresences to mitigate some of the rubberbanding issues while we are sending incorrect time dilation values 2009-10-27 13:16:58 -07:00
John Hurliman a718d7d56a Making the defaults for interest management variables match whether you have the [InterestManagement] section in your config or not 2009-10-27 10:17:20 -07:00
John Hurliman bcd7593dfb Forgot a line in the previous commit 2009-10-27 10:07:35 -07:00
John Hurliman 3a1ee79ee4 Finally hunted down the Parallel deadlock. Packets were being handled asynchronously (filling up the threadpool with handlers), which would turn around and try to do parallel operations on the starved threadpool. The solution for now is to disable Parallel.cs operations until we can gracefully handle parallel operations with a potentially starved threadpool 2009-10-27 02:36:57 -07:00
John Hurliman c75d415648 * Converts ClientManager.ForEach() (and as a result, Scene.ForEachClient()) to use a non-blocking parallel method when operating in async mode
* Minor code readability cleanup
2009-10-27 00:26:56 -07:00
John Hurliman d199767e69 Experimental change of PhysicsVector to Vector3. Untested 2009-10-26 18:23:43 -07:00
John Hurliman b6651ce790 * Double the priority on avatar bake texture requests to get avatars rezzing in faster than the surrounding scene
* Adds duplicate tracking for SceneObjectParts and ScenePresences to avoid sending out duplicate ImprovedTerseObjectUpdate packets
2009-10-26 18:22:32 -07:00
John Hurliman 0b1726b524 Removing the ClientManager reference from IScene and hiding it entirely inside Scene as an implementation detail. This will reduce programming error and make it easier to refactor the avatar vs client vs presence mess later on 2009-10-26 16:48:43 -07:00
John Hurliman 4847e62e9f * Switched all operations on the list of clients that could be either sync or async to use Scene.ForEachClient() instead of referencing ClientManager directly
* Added a new [Startup] config option called use_async_when_possible to signal how to run operations that could be either sync or async
* Changed Scene.ForEachClient to respect use_async_when_possible
* Fixing a potential deadlock in Parallel.ForEach by locking on a temporary object instead of the enumerator (which may be shared across multiple invocations on ForEach). Thank you diva
2009-10-26 16:33:04 -07:00
John Hurliman 119cf80e13 Added calls to GC.AddMemoryPressure() when unmanaged memory is allocated for ODE (helps the GC make better scheduling choices), and a call to GC.Collect() right before logins are enabled for a region. Although this doesn't change actual memory usage, it improves the reported usage from OpenSim and the operating system 2009-10-26 15:52:59 -07:00
John Hurliman ac7ccdf7d7 * Changed the watchdog timer to improve the speed of UpdateThread(), only track threads once the first call to UpdateThread() has been made, and allow re-tracking of threads that timed out but revived later
* Added a commented out call to Watchdog.UpdateThread() in OdeScene. If it turns out that loading a large OAR file or some other operation is timing out the heartbeat thread, we'll need to uncomment it
2009-10-26 14:41:27 -07:00
John Hurliman de25dcfb7b Minor tweaks to get OpenSim compiling against both the current libomv and the upcoming 0.8.0 2009-10-25 01:10:02 -07:00
John Hurliman 730930955a Changing Scene.ForEachClient to use the synchronous for loop instead of Parallel. This is quite possibly the source of some deadlocking, and at the very least the synchronous version gives better stack traces
* Lock the LLUDPClient RTO math * Add a helper function for backing off the RTO, and follow the optional advice in RFC 2988 to clear existing SRTT and RTTVAR values during a backoff

* Removing the unused PrimitiveBaseShape.SculptImage parameter * Improved performance of SceneObjectPart instantiation * ZeroMesher now drops SculptData bytes like Meshmerizer, to allow the texture data to be GCed * Improved typecasting speed in MySQLLegacyRegionData.BuildShape()

* Improved the instantiation of PrimitiveBaseShape
2009-10-25 00:40:21 -07:00
John Hurliman 2c34619aea * Changed various modules to not initialize timers unless the module is initialized. Ideally, the timers would not initialize unless the module was actually enabled, but Melanie's work on configuring module loading from a config file should make that unnecessary
* Wrapped the Bitmap class used to generate the world map tile in a using statement to dispose of it after the JPEG2000 data is created
2009-10-23 14:22:21 -07:00
John Hurliman a41cd1d069 * Unregister Mono.Addins event handlers in PluginLoader.Dispose() and always handle PluginLoader with the using pattern. This freed up 121,634,796 bytes on my system
* Avoid allocating an Action<IClientAPI> object every round of the OutgoingPacketHandler
* Removed unnecessary semi-colon endings from OpenSim.ini.example [InterestManagement] section
2009-10-23 13:14:29 -07:00
John Hurliman 62f1a5e36d Implemented a "FrontBack" prioritizer, using distance plus the plane equation to give double weight to prims/avatars in front of you 2009-10-23 02:38:59 -07:00
John Hurliman 588361e2a2 Experimental change to use an immutable array for iterating ScenePresences, avoiding locking and copying the list each time it is accessed 2009-10-23 01:02:36 -07:00
John Hurliman ba2972eaf6 * Send out m_bodyRot everywhere instead of m_rotation. Still have no clue which is right
* Fix WorldMapModule.process() to not trip the watchdog timer
2009-10-22 14:33:54 -07:00
John Hurliman 4ba3842d71 Forgot to hit save in the last commit 2009-10-22 13:24:24 -07:00
John Hurliman 167d8e39fa * Sending (position - hipoffset) instead of position
* Sending m_rotation instead of m_bodyRot in full updates to match terse updates (no idea which one is right!)
2009-10-22 13:14:41 -07:00
John Hurliman b2ed348aa2 Implemented a Watchdog class. Do not manually create Thread objects anymore, use Watchdog.StartThread(). While your thread is running call Watchdog.UpdateThread(). When it is shutting down call Watchdog.RemoveThread(). Most of the threads in OpenSim have been updated 2009-10-22 12:33:23 -07:00
Melanie c4969d47d9 Merge branch 'master' into vehicles 2009-10-22 07:12:10 +01:00
John Hurliman 45dc4e0a54 * Added a sanity check to GetScriptAssemblies() and GetScriptStates() for the case where no scripting engine is enabled
* Added TokenBucket.cs to OpenSim, with some fixes for setting a more accurate MaxBurst value and getting a more accurate Content value (by Drip()ing each get)
2009-10-20 18:19:17 -07:00
John Hurliman 1833c69568 * Removed the unused m_agentUpdates collection and some extra work that was being done for AgentUpdate packets
* Start LLUDPClients unpaused (this variable is not being used yet)
2009-10-20 15:19:19 -07:00
John Hurliman bd03cbd815 Merge branch 'master' of ssh://opensimulator.org/var/git/opensim into prioritization 2009-10-19 15:19:37 -07:00
John Hurliman 142008121e * Change Util.FireAndForget to use ThreadPool.UnsafeQueueUserWorkItem(). This avoids .NET remoting and a managed->unmanaged->managed jump. Overall, a night and day performance difference
* Initialize the LLClientView prim full update queue to the number of prims in the scene for a big performance boost
* Reordered some comparisons on hot code paths for a minor speed boost
* Removed an unnecessary call to the expensive DateTime.Now function (if you *have* to get the current time as opposed to Environment.TickCount, always use DateTime.UtcNow)
* Don't fire the queue empty callback for the Resend category
* Run the outgoing packet handler thread loop for each client synchronously. It seems like more time was being spent doing the execution asynchronously, and it made deadlocks very difficult to track down
* Rewrote some expensive math in LandObject.cs
* Optimized EntityManager to only lock on operations that need locking, and use TryGetValue() where possible
* Only update the attachment database when an object is attached or detached
* Other small misc. performance improvements
2009-10-19 15:19:09 -07:00
Melanie ffd59868f5 Merge branch 'master' into vehicles 2009-10-19 21:58:51 +01:00
John Hurliman 233e16b99c * Rewrote the methods that build ObjectUpdate and ImprovedTerseObjectUpdate packets to fill in the data more accurately and avoid allocating memory that is immediately thrown away
* Changed the Send*Data structs in IClientAPI to use public readonly members instead of private members and getters
* Made Parallel.ProcessorCount public
* Started switching over packet building methods in LLClientView to use Util.StringToBytes[256/1024]() instead of Utils.StringToBytes()
* More cleanup of the ScenePresences vs. ClientManager nightmare
* ScenePresence.HandleAgentUpdate() will now time out and drop incoming AgentUpdate packets after three seconds. This fixes a deadlock on m_AgentUpdates that was blocking up the LLUDP server
2009-10-18 20:24:20 -07:00
Jeff Ames 0d29614ca1 Formatting cleanup. 2009-10-19 08:58:03 +09:00
John Hurliman fdb2a75ad3 Committing the second part of Jim Greensky @ Intel Lab's patch, re-prioritizing updates 2009-10-17 18:01:22 -07:00
John Hurliman 9c21d672d4 Merge branch 'master' of ssh://opensimulator.org/var/git/opensim into prioritization 2009-10-16 18:32:12 -07:00
Teravus Ovares (Dan Olivares) 01051daaab * One more tweak to inform the user that they may not be able to move until relogging. 2009-10-16 21:24:08 -04:00
John Hurliman b813058635 Merge branch 'master' of ssh://opensimulator.org/var/git/opensim into prioritization 2009-10-16 13:29:12 -07:00
Teravus Ovares (Dan Olivares) f5964347bd * fix previous commit 2009-10-16 16:25:48 -04:00