Some interesting things are happening. The Server Beta User Group met Thursday in Morris, ADITI. This was Oskar Linden’s weekly meeting. Oskar is gone, it is fact. Maestro Linden is taking over the meeting… or may be Coyot Linden. I’m not sure it has been decided. But, it is the intention to keep the meetings going. Also the Deploy threads in the forum will continue.
I think it looks like Maestro Linden will run the meetings. Maestro hopes to get more developers from the server project attending the meeting. There are a number of server enhancements and features moving forward. We hear very little about them. Getting more developers involved may give us more information.
In News 1 I covered the roll out packages, so I won’t repeat that information.
They have found a new crasher bug in the code running in Magnum. Maestro says they have already found a fix for that problem.
Script Time in the Viewer Statistics panel now correctly shows 0ms when scripts are disabled in the region.
Also, script error messages now include information about the object’s root prim, when certain operations fail due to the object’s Pathfinding setting. Previously we would see notices that referenced prims in the line set, which are often named ‘object’ and for practical purposes useless. The name and position of the root prim will now be provided, which should be way more useful.
Large Group Edit
The code for the Large Group Editing fix is currently running in all three release candidate channels. So, with any luck it will make it to the main channel and all regions on the main grid next week (46).
The viewer side code is now in the current Beta Viewer. With any luck that viewer will roll out as the main stable release in the very near future.
If both of these things happen, we will see this Large Group problem resolved. I do expect to see a few sets of revisions and minor changes to the viewer side of the fix. Things like a progress bar or something to indicate time needed for large group data to download.
If you are reading the release notes for the server code on Magnum you will see the notice to use the Project Viewer. My announcements have been telling you that too. Well… don’t. Use the Beta Viewer.
Interest List
We have been hearing about this for some time. Andrew Linden has been working to improve how the Interest List works, which should improve rez speed and server efficiency. We can expect a ‘pile-on’ test in ADITI sometime next week.
Andrew has apparently fixed most of the problems and is now working to get some performance improvement.
Region Crossing
Caleb Linden mentioned the multi-threaded region crossing code running in ADITI. There are still problems but they are getting resolved.
Materials System
See the old announcement in the SL Blog. Maestro says the server side of the feature is running in ADITI. Not the entire grid there, but in some servers. Look for DRTSIM-169. Chooqu is one region with it.
For the materials system to work additional information has to be stored server side. Those changes are in testing on ADITI.
Obviously, there are going to be viewer side changes too. We know Geenz Spad at Exodus Viewer is developing the project and undoubtedly has a development viewer. I am pretty sure it is not out where user can get a copy. Things just are not ready for open beta testing.
This gives me hope that the project is further along than I might have thought. But, I still doubt it will complete this year.
Linux Boost
NVidia has announced a big thing for Linux users. See the announcement: NVIDIA Delivers Massive Performance Boost To Linux Gaming.
The announcement says that all systems back to the nVidia 8800’s will benefit from the new drivers.
The current stable driver is 306.97, which I’ve been running on. It has had some problems but works well with Second Life. The beta version out is 310.33. You can get it in nVidia Beta & Legacy Drivers. I have yet to download and try it with my 560. But, I have downloaded it.
Linux users should see a big change.
But, remember. This is a beta release.
The “performance boos”t is supposed to be obtained by enabling threaded rendering in the driver. This is done by adding a line in the viewer wrapper script:
export __GL_THREADED_OPTIMIZATIONS=1
I tested it and frankly, it brings almost no FPS boost to the viewer, while it does use one more core at 100% (so, basically, it simply makes your CPU heat more and consume more electric power for nothing !)…
The tests were done with the drivers v310.14 (there’s no v310.33 available for Linux on NVIDIA’S site… Where did you get that number from ?).
So, in fact, the viewer doesn’t benefit from any boost under Linux (but it’s not really a surprise to me: it is already faster under Linux than it is under Windows and the Windows drivers already had the thread optimization).
Depending on how the render pipeline and shaders are written, an OpenGL program may or may not benefit from threaded graphics drivers: obviously, the viewer as it is currently coded can’t benefit from it.
Thanks for the information.
The 310.33 driver is the version I found for Windows.
Ah, ok… Well, since you were speaking about “Linux boost” in your post, I was assuming you were speaking about the Linux driver version…
Note also that the threaded optimization is nothing new under Windows (it was already in the non-beta drivers), thus why NVIDIA refers to “boost to Linux”.
thankfully with this rolling the temporary SIM-crossing problem, which were last week introduced is now healed.
Also we should now have a cool new havok-version.