* 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
* OnQueueEmpty is still called async, but will not be called for a given category if the previous callback for that category is still running. This is the most balanced behavior I could find, and seems to work well
* Added support for the old [ClientStack.LindenUDP] settings (including setting the receive buffer size) and added the new token bucket and global throttle settings
* Added the AssetLoaderEnabled config variable to optionally disable loading assets from XML every startup. This gives a dramatic improvement in startup times for those who don't need the functionality every startup
* Changed the way the QueueEmpty callback is fired. It will be fired asynchronously as soon as an empty queue is detected (this can happen immediately following a dequeue), and will not be fired again until at least one packet is dequeued from that queue. This will give callbacks advanced notice of an empty queue and prevent callbacks from stacking up while the queue is empty
* Added LLUDPClient.IsConnected checks in several places to prevent unwanted network activity after a client disconnects
* Prevent LLClientView.Close() from being called twice every disconnect
* Removed the packet resend limit and improved the client timeout check
* Changed the ClientManager interface to reduce potential errors with duplicate or mismatched keys
* Added IClientAPI.RemoteEndPoint, which can (hopefully) eventually replace IClientAPI.CircuitCode
* Changed the order of operations during client shutdown
* Moved all priority queue access to helper functions to improve reliability and readability
* New version of CSJ2K.dll (from libomv source tree) that fixes a JPEG2000 comment decoding bug
* Fixed the incorrect "out of memory?" exception messages and replaced them with useful information
* Cleaned up J2KImage and LLImageManager to conform to coding conventions
This change moves texture send processing out of the main
packet processing loop and moves it to a timer based
processing cycle.
Texture packets are sent to the client consistently over
time. The timer is discontinued whenever there are no
textures to transmit.
The behavior of the texture sending mechanism is controlled
by three variables in the LLCLient section of the config
file:
[1] TextureRequestRate (mS) determines how many times per second
texture send processing will occur. The default is 100mS.
[2] TextureSendLimit determines how many different textures
will be considered on each cycle. Textures are selected
by priority. The old mechanism specified a value of 10 for
this parameter and this is the default
[3] TextureDataLimit determines how many packets will be sent for
each of the selected textures. The old mechanism specified a
value of 5, so this is the default.
So the net effect is that TextureSendLimit*TextureDataLimit
packets will be sent every TextureRequestRate mS.
Once we have gotten a reasonable feeling for how these parameters
affect overall processing, it would be nice to autonmically manage
these values using information about the current status of the
region and network.
Note that this also resolves the pathologcal problem that
previously existed which was that a seated avatar generated very
few in-bound packets (theoretically) and would therefore be the
least able to retrieve the images being displayed by a
projector script.
This may break a lot of things, but it needs to go in. It was tested in standalone and the UCI grid, but it needs a lot more testing.
Known problems:
* HG asset transfers are borked for now
* missing texture is missing
* 3 unit tests commented out for now
Following feedback from 0003440, i've made some changes to the new texture pipeline to optimise
performance. The changes are:
- Fixed a math issue where a small percentage of images with a certain size (on the packet boundary) would not have their final data delivered. This issue has been present since pre- 0003440
- It was suggested that a discardlevel of -1 and a prioriy of 0 meant to abandon the transfer, this is incorrect and caused some textures to clog.
- The texture throttle blocking queue is now only filled in relation to the actual throttle amount.. i.e, on a connection throttled to 300k, only twenty packets will be placed in the queue at a time, on a larger connection it will be much more. This is to balance responsiveness to requests and speed, and to minimise wasted packets.
- The engine now keeps track of the number of pending textures, and the stack will not be walked if there's no textures pending, saving CPU. Textures are only considered "pending" when they've already been decoded.
- As part of the above, some textures may receive twice as much data per cycle if the number of pending textures is below the cycle threshold, this should prevent loading from slowing down when there are fewer textures in the queue.
* WebStatsModule doesn't crash on restart. GodsModule doesn't crash when there is no Dialog Module. LLUDPServer doesn't crash when the Operation was Aborted.
* ODEPlugin does 'Almost NaN' sanity checks.
* ODEPlugin sacrifices NaN avatars to the NaN black hole to appease it and keep it from sucking the rest of the world in.
* Revamps the server side texture pipeline
* Textures should load faster, get clogged less, and be less blurry
* Minor tweak to ensure the outgoing texture throttle stays private.
* Fixes mantis 3440
* Added log4net dependency to physxplugin in prebuild.xml.
* Added missing m_log fields to classes.
* Replaced Console.WriteLine with appropriate m_log.Xxxx
* Tested that nant test target runs succesfully.
* Tested that local opensim sandbox starts up without errors.
These changes replace all direct references to the AssetCache with
IAssetCache. There is no change to functionality. Everything works as
before.
This is laying the groundwork for making it possible to register
alternative asset caching mechanisms without disrupting other parts of
OpenSim or their dependencies upon AssetCache functionality.
OpenSim.Region.Environment into a "framework" part and a modules only
part. This first changeset refactors OpenSim.Region.Environment.Scenes,
OpenSim.Region.Environment.Interfaces, and OpenSim.Region.Interfaces
into OpenSim.Region.Framework.{Interfaces,Scenes} leaving only region
modules in OpenSim.Region.Environment.
The next step will be to move region modules up from
OpenSim.Region.Environment.Modules to OpenSim.Region.CoreModules and
then sort out which modules are really core modules and which should
move out to forge.
I've been very careful to NOT BREAK anything. i hope i've
succeeded. as this is the work of a whole week i hope i managed to
keep track with the applied patches of the last week --- could any of
you that did check in stuff have a look at whether it survived? thx!
* Increases the number of ImageDataPackets we send per PriorityQueue pop and tweak it so that the number of packets is ( (2 * decode level) + 1 ) * 2, and (((2 * (5-decode level)) + 1) * 2). The first one sends more data for low quality textures, the second one sends more data for high quality textures.
