Second Life News 2013-10

The main channel of server software got a new package from the RC channels. This package ran on all three RC channels last week. It is a crash fix and not much more.

New

The Blue Steel and Le TigreRC channels are getting Baker Linden’s object rez performance improvement. This is a multi-threaded object rez addition. Simon Linden says, “There was a laggy bit of work the server used to do when rezzing objects that has been moved into a side thread.” Paraphrasing Simon, people will notice it most with complex objects and link sets with lots of scripts.

Apparently we may not see it speed up the rez… but, it will have less impact on the simulator as the sim won’t stop to process an object. I take this to mean any single objects will not necessarily rez faster but more objects should rez in less time. This can be thought of as changing from rezzing one after the other to parallel rezing, rezing several at once.

There is some hinting that the actual rez time of simple objects may see a slight increase in rez time. Combat sims are likely to see the most change of any region. Depending on the overall rez speed of simple and complex objects.

Guns may see a slight slowing of bullet rez. Since combats sims limit fire rates to reduce lag it may not be visible in overall fire rates. A limit of 2 to 15 BPS (Bullets per second) is not likely to see any change in weapon performance. Also, since bullets tend to be small and as simple as possible the rez time change should be minimal. We’ll have to see if anyone notices.

The Magnum RC channel is getting a package that is said to fix some crash modes.

BUG-1612Unable to TP back home after tping out of my region. This fixes adds the ability for estate managers to teleport into any place in their region after setting teleport controls for visitors.

SVC-8019roll restart normalization time issues : cannot cross region borders for a while , very slow restart at roll restart , no fetching on server channel- version. This fixes neighboring region visibility delays after rolling restarts.

VWR-786IM does not respect “allow user to see my online status” flag in friends list. This too is supposed to be fixed.

New Stuff

Simon tells us, “I’ve been working on some combined viewer and simulator code to try and prevent viewers from falling apart when working in crowded areas such as clubs.” Paraphrasing the rest of his comments…

The effort is experimental. It is about changing the way the viewer renders avatars. Now, the viewer basically tries to draw everyone, and if that’s too much, backs off to using imposters. That first part, trying to draw everyone, is a problem and a lot of people don’t get past that as they crash or they get horrible performance.

Even with imposters, the viewer has to render the AV, getting shape and texture data.

This idea … again, only experimental … basically says you’ll probably want to first see those nearest, your friends and maybe anyone you’re in an IM chat with. If performance is OK, draw more. It needs a bunch of work to see how it feels and what the issues will be.

If you’re on a cheap laptop that can only handle 10 avatars, you should still be able to go to a club with 50 people around.

The sim will be sharing some render cost info … currently the viewer doesn’t really know how expensive it is to draw someone without getting all their data and trying to draw them. But, to get the render cost data something has to draw the avatar, which is currently all viewer side now.

There are other situations … for example, I was in a club a while back, and turned on the indicator for render costs. People seemed to be 75-100k on average … that’s just a rough number by eye-balling the crowd. That’s pretty high, but someone had over 1,000k. So that one person was 10x most others. If your viewer knew that without having to download all that expensive AVs shapes, attachments and textures, it can decide to skip that one.

The interesting part of this is what kind of smarts to put in the viewer. Imagine you drop into a crowded area, and before you draw anything you could find out how expensive each AV is to draw. The most basic idea is just having an upper limit … if it’s too much, don’t draw it.

However, if you’re into photography, working with a model on an empty region, you’ll want to draw everyone at high res.

Having a hard limit … say “you can’t come to the club, you’re too complex” is ugly because most people don’t know where to start to fix it.

All this has to tie into your own settings and computer abilities. If you have a hot machine, you want to draw everything with all the graphics extras. We’ve considered having more “cost” barriers … like scripts, but the experience is horrible. You try to go to popular place, you get rejected, and most people don’t know or can’t figure out their new hair has a hundred re-sizer scripts in it.

Anyway, at this point it’s experimental and I don’t know how far it will get. I do hope something will work out and it will be better for people to hit crowded areas.

This experiment may never make it to QA. Some don’t. If it makes a significant performance difference for the effort required to implement it, then we may see it. Since Simon is talking about it, I think changes are good we will see it in QA some day.

Magic Boxes

April 15, 2013 Magic Boxes will be pretty much phased out. You need to be switching over now.

See the announcement: Magic Box Retirement Starts April 2, 2013.

5 thoughts on “Second Life News 2013-10

  1. Thanks for the info – Simon’s work sounds very interesting. We still have the facility to derender avatars above a certain surface area or with attachments that have a heavy rendering costs, but their settings are sadly not exposed to the user in the standard viewer (seems that LL went cold on the idea after developing it). I’ve used them occasionally, but an automatic way to ‘imposterise’ avatars would be better.

  2. “Imagine you drop into a crowded area, and before you draw anything you could find out how expensive each AV is to draw. The most basic idea is just having an upper limit … if it’s too much, don’t draw it.”

    To me this sounds like Visual Auto Mute, which was introduced a year ago in the Viewers. In the official Viewer with Starlight skin it is called “Derender Avatars” and looks like this.

    What is the difference in this new project?

    • The difference is in the tech. All the render cut-offs based on avatar cost require the viewer to collect the attachments and textures and evaluate them. This new process could evaluate the avatar server side when the bake is done and cached. The Lindens may cache the render cost with the cached baked texture. Then the server could send the ‘evaluated’ information to the viewer.

      The viewer could decide to render the avatar or not based on render cost. If it decides not to, then the textures and attachments do not have to be downloaded. Thus allowing both the viewer and server to save work.

Leave a Reply

Your email address will not be published. Required fields are marked *