Window Dressing

Am I insane to consider hanging flexible curtains on every one of the fifty-odd windows in my house? Is there a danger my soft furnishings might sodomise system performance? Will my full-perm nets knock the guests' graphics cards for six? Should I see about a cheap and cheerful alternative from Walmart?

Seeing how I have no takers on that frivolously framed conundrum, let me phrase this another way: just as Berkeley worried about objects of sense, I'm concerned about whether or not a preponderance of flexi prims in my house would matter if I were standing in a room where only a limited number of the things are visible.

Is there a diligent chip somewhere in the network running thousands of lag-worthy calculations on behalf of the otherwise invisible curtains? Perhaps more importantly, would it matter to a neighbour on the far side of the estate?

Pardon me if the question might seem a little obvious but we did make the 'puter in our own image and, for me at least, one or two conundrums of perception insist on persisting.

There are people who are going to argue both ways, personally, I'd give it a shot and see what it does to your lag and framerate. Flexi does increase lag, but honestly -- half of your female guests will show up in skirts using just as many flexi prims or hair.

If the flexi curtains make you happy, rock on. If it turns out that they make your user experience suck, then delete them.

I think this is something you need to experiment with, to see if flexiprims keep on flexing when the prims themselves have been culled by object-object occlusion.

How could the viewer know if a flexi prim was occluded (or, for that matter, occluding) before it had calculated its newly flexed shape?

The longest edge of the bounding envelope of all possible shapes for a flexi prim is no more than twice the length of the longest edge of the bounding envelope of the un-flexed prim. If that bounding envelope is occluded, the flexiprim is occluded.

Since curtains are alpha they can't occlude anything, so in this case it makes no difference if they're flexi or not.

So here is a casual experiment I did on site. The flexi prims are working fairly well in all four views.

This first floor interior view of three flexi prims curtaining three windows is roughly 30FPS:



This ground floor interior view with six flexi prim curtains drawn open on three windows is roughly 27 FPS:



Interestingly, the framerate drops from 27 to 15 FPS when I camera outside to view the same three windows:



There is no noticeable difference in the framerate if I camera up to this external view of the first floor:



The images are a little dark but please note that for the time being the walls at my place are mostly blank and the rooms are largely empty. However the outside walls are bumpmapped with bricks, which can be a little heavy on resources, I believe. Also, the windows are sculpted so both exterior views feature a total of nine sculpties.

Recalling Drongle's experiment with ktris in another thread about the effect of sculpties on performance, it seems that the sculpted windows hog more resources than the flexi prim curtains.

Am I right to draw that conclusion?

Your not just viewing three windows from the outside there are 5 windows in that image.

Aren't the same number of frames and curtains being rendered in each picture (still visible through the curtains/windows) ... in which case I don't think you can tell. Looks to me that the bump mapping is the only thing you can blame here. If you remember, I had to use hundreds of sculpties in the field of view to get any effect on frame rate. Maybe I should do it again with transparent flexies!

Dammit, you're right - and another thing I failed to take into account is that the exterior views are looking into the estate, whereas the interior views are looking out on linden water and a faux horizon of mountains beyond the sim boundary. I guess the big drop in FPS on the exterior views is due to the client rendering everything on the estate that is within drawing distance of my camera?

Still, if I rotate the camera slightly on the exterior view to take the two extra windows out, the framerate improves by only 1 FPS. Also, if you look at the next image showing the exterior upstairs, where the curtains are closed, there are also another two windows obscured by the curtain on the right but the FPS readings don't vary from the previous view - and the FPS is consistent even when I rotate the camera. Neither do I see a noticeable variation in framerates on similar views of other rooms without the curtains.

At any rate (literally) the flexi prims don't appear to have much of an effect.

Well the ground floor views have twice as many flexis as the views of the upper floor because they are drawn open but, as I noted above, the difference to FPS appears to be negligible.

Is bumpmapping seriously bad? I haven't gotten around to texturing the walls yet but I plan to make a brick texture to match the bumpmap and I have used a few other bumpmaps here and there around the house. I'd hate to think it is rogering resources.

And when you mention 'field of view' does that mean everything within drawing distance in front of the camera, in the radius of the camera or just what may be viewed at any given moment on screen? In which case, would the view of my house in relation to the rest of the estate as described above make more of a difference to FPS?

OK. I did an experiment.
2000m, no sky, no ground (nothing in view but blackness).
500 identical objects in view in rectangular array (no occlusion):

objects....texture....flexi.....fps
---------------------------------------
boxes......default....no........75
boxes......default....yes.......45*
boxes......alpha......no.........56
boxes......alpha......yes.......37*
sculpty.....default....no........65
sculpty.....alpha......no........56.
---------------------------------------

*flexing was very slow, about one update per two seconds, with clear waves of updates across the field of objects.

Sculpty was the default map

Conclusions - lag effects are flexi>alpha>sculpty.

Just wait till we have alpha flexi sculpties ... the expected future of plants?

Funnily enough, I was just thinking how good it would be to have flexi sculpty alpha curtains.

I think I need to get my head examined.

For flexy sculpty alpha hair?

No fly-away fringes for me, boyo. It took me a whole day and full tin of virtual Brylcreem to get this right:



Should do me for years to come - or at least until teflon-coated haircuts come back into fashion.

Whoa, neanderthal brow ridges and cro-magnon jaw, you're the missing link!

Yep. Just a tad further with the sliders and I'd have to go live in a tree.

A flexi alpha sculptie tree of course.

I don't know. I'll let someone else do those experiments.Again, I don't know. It gets very complicated with LOD effects as well. That's why I do my experiments against a black background and with a flat array of objects. Even then, everything is relative and the rates fluctuate.

I think a few would be excellent. I have to use hundreds of objects to see effects on frame rates. It's whole landscapes full of alpha-sculpty-flexi plants that worry me a bit....especially with lots of people wearing alpha-sculpty-flexi hair coming to look at them....maybe on an open space sim....after all, they are especially for landscape!

I think I should use llGetPrimitiveParams to add flexi to them but only when the windows are open (the sashes slide open on touch). I could also vary the amount of flexi depending which part of the window is open and perhaps add a temporary bit of swing when the curtains are drawn together or pulled apart.

It would be a shame not to use the effect because it is pleasant to watch the curtain texture catch the light as they move with the wind.

Funny thing is I can't stand curtains or blinds in RL.