illusion

dust's picture

so this is part of one of my gl flower spline pieces, the interesting thing with this is as soon as i added some tone burst feedback macro style to flower the walls in my house started moving. so its one of those illusions you stare at for 30 seconds without blinking then look at the wall. well this one is so powerful it doesn't matter if you blink or not. the longer you stare the better. im not sure if frame rate matters but this fluxing between 60-30fps on my system it should still work at 12fps im sure. if you want to try this illusion please be warned that it has a long recovery time.

PreviewAttachmentSize
illusion-flower.qtz50.47 KB
illusion.qtz39.69 KB
illusion-flower.zip803.09 KB

Lango's picture
Re: illusion

Hey dust

I really like this composition. Didn't get the walls moving and only around 20 fps though.

Also, I got an error trying to go to your http://numedia-sci.info/ site. And http://dvjdust.info/ seems to be timing out. Just though i'll give ya a heads up.

Cheers

Lango

leegrosbauer's picture
Re: illusion

I'll be damned .. it's true! The wall has the distinct appearance of rushing at me after twenty seconds of staring at illusion-flower!?! Reproduce-able with each observation I make. huh!

Anybody else? I'm about two feet from a textured beige wall in low-medium light conditions. 30 fps here.

The archived app is sluggish on a large monitor. Might be just fine on a smaller screen, however. (Control-Q to quit, btw.)

psonice's picture
Re: illusion

It's a pretty common illusion. When you see something that seems to be constantly moving away from you / shrinking for a long period, your brain seems to adapt to thinking it's staying still. Then when you look at at something that IS still, it seems to be expanding (as your brain is still trying to compensate for the shrinking..)

For another cool effect, try looking at your hand afterwards. It seems to grow fat...

leegrosbauer's picture
Re: illusion

virtual viagra! ahaha

psonice's picture
Re: illusion

o0

I won't ask why that was in your hand... ;D

gtoledo3's picture
Re: illusion

"Baby, uhm, look at this screen for a few minutes, I'll meet you in the bedroom".

I'm so sorry. Can't...help...myself....

It looks like it probably looks awesome. I get about .5fps out of it though (the illusion.qtz). When I run the app, it is actually is hosing up before it totally loads. (I even got a bit of color buffer glitch weirdness and cam image!) I'm getting the rainbow wheel of death...

On another note, I opened one of the qtz's on another thread, blowfish, and it looks so delicate and cool. I think it's going total software renderer for me though.

Are people around here getting really good fps with the spline? I notice cybero use it alot, and I see it pop up in this cool stuff from Dust. The patterns looks so delicate and intricate, but they perform really laggy on my system... is it a GPU thing (Intel in this case...)?

I'll have to try firing up on an nVidia. In any case, I just want to send a head's up that GL spline stuff really crawls on Intel GPU's (X3100), because it sounds like it's going pretty fast on other systems.

I'm fascinated with some of the spline patterns.

psonice's picture
Re: illusion

It ran 30-60fps here (on an ati 2600).

The intel chips are lacking a LOT of features, and the features they don't lack are seriously lacking speed. So if stuff doesn't run well, chances are that it's either just too slow, or it's using something that it "supports" (intel have a habit of "supporting" GPU features that don't actually exist on the chip.. i.e. the whole thing ends up running in software).

cybero's picture
Re: illusion

Nice Iterations, I often am left wondering just how to bring down the processing overhead without costing the quality of the generated result.

Your managing to get a good 28 - 30 fps at 1280 * 764 and 20 fps at Viewer full screen dimension - 1920 * 986.

Optical Illusions, I love 'em, got a deck of cards full of optical illusions - two decks actually [ no illusion].

PreviewAttachmentSize
imsogiddy_myheadisspinning.qtz50.99 KB

leegrosbauer's picture
Re: illusion

Regarding Dustin's Spline series; not sure if a causal relationship exists, but the two firefeed comps stalled out for me initially. Shortly after that, I updated GL Tools to GL Tools, v20090707 (1.3). Now, neither firefeed composition stalls out.

Blowfish didn't exhibit any anomalies at any time. For me, the spline comps under consideration generally vacillate between thirty and sixty fps with a few that run between twenty to thirty fps.

ATI Radeon HD2600

cwright's picture
Re: illusion

Isn't that "support" word fantastic? ;)

From my experience, Intel GPUs (at least the GMA950 and probably the X3100 -- not sure about others) don't do line drawing in hardware. To test this, try drawing a teapot in a Polygon Mode patch, and then change it to lines instead of textured -- the framerate drops even though it's drawing dramatically fewer pixels. Some of it could be attributed to line antialiasing (which has to read, modify and write vram), but even with that off the render speed isn't better or even tied. It's a shame, because fast line drawing in software isn't particularly difficult either (Bresenham line drawing is decades old, and wu line drawing (for AA) is almost 20 years old itself).

And to top it off, my super-ancient CL-GD43xx (from ~1989-1990 -- that's a Cirrus Logic VLB graphics card) had exactly 2 hardware accelerated features: PutPixel (way good when Vesa Bankswitching was the alternative) and HLine/VLine -- so most line drawing algorithms could find their horizontal or vertical spans, feed the ends to the gpu, and get screaming fast lines that are probably faster than intel's software rasterizer today that's ~500x faster in every way (2x 500x-faster cpus, bus is 50x faster, 4x wider, etc etc etc...)

(for me: textured teapot = ~143fps, wireframe teapot = ~14fps (yes, an order-of-magnitude difference in The Wrong Direction))

cybero's picture
Re: illusion

Quote:

Isn't that "support" word fantastic? ;)

LOL

Covers up a multitude of sins [ of omission ] :-) , more often than not.

All part of the sinister non existent conspiracy regarding the widespread proliferation of non - existent entities.~

Apple & their choices of graphics hardware - now I am given to understand that nVidia might be getting the push, as they failed to be sufficiently timeous in delivering chips.

Sustainable Development - a tricky problem.

cybero's picture
Re: illusion

I wonder if this is simply a cause and effect of what plugin series is deployed on the respective author's workbench ?

E.g

1.1 will upgrade to 1.2 , which will, in its turn, upgrade OK to 1.3, but going back down the line, 1.3 to 1.2 and so on causes these problems of nil render, nil full performance, etc.

Just noticed ATI, might it be plugin to chipset to plugin version related ?

cwright's picture
Re: illusion

This shouldn't be so much of a problem recently, but there was a break back in the old days of GLTools (we had to rename stuff internally to not risk breaking future stuff) -- so current GLTools patches (and even the +deprecated ones) will not get loaded by older plugins. I think that's all pre-1.1 era stuff, but I haven't checked.

Otherwise, everything should be normal between all of them, execpt for a few tweaks to the spline patch (which Dust makes use of in this composition I think) -- that might be the cause of this problem?

cybero's picture
Re: illusion

Tweaks to the spline patch that result in weirdly out of place poitional numbers? - If I want a spline's X1, X2, etc to stick to a single position nowadays I set a random patches to a single specific X1, X2, etc number range and pipe that into the spline's X1, X2, etc, cures most ills, unless I get that number wrong :-)

I call it tethering. Like tying down a kite or a sail.

mattgolsen's picture
Re: illusion

Would it be possible to setup KinemeCore to check for updates to installed Kineme plugins? An option to include beta versions would be great too.

leegrosbauer's picture
Re: illusion

Well, Dust's spline comps were published on July fifth, two days before GL Tools v20090707 (1.3) was released yesterday, if all the publication dates are accurately reflected in the forums. I was up to date with the previous version of GL Tools so there's little chance that he published a newer version than what I already had. Since the two firefeed comps are working well now when they previously were not, my only guess would be that the Spline patch perhaps got tweaked for the better in yesterday's release.

edit: Ah, I see that Chris has posted a similar consideration regarding the Spline patch while I was composing. Anyway .. all is working just fine now for whatever the reason, so that's good.

cwright's picture
Re: illusion

This bug should have been fixed a few betas ago -- are you sure it's still a problem?

gtoledo3's picture
Re: illusion

That holds true, but one interesting detail is that if I do wireframe on both sides if the speed picks up dramatically. That's ALWAYS the case for me... well, that I can think of. If I use the polygon mode, go to wireframe on one side, I get just what you describe. If I enable wireframe on both sides (or have one side of whatever it is culled at the start), then I get the same framerate as no wireframe more or less... maybe 2fps less on average.

...and GLSL stuff is all over the place :) I have a few things that will run on the intel that people seem to have problems with on the nVidia stuff, and then vice versa.

cwright's picture
Re: illusion

It's possible -- in fact, a large swath of KinemeCore's code is hand-holding Sparkle into working with bundles correctly (it's using a 2+-year old version of sparkle, probably fixed now?) for that purpose.

Part of the problem lies in getting this website's software (drupal) to suck less so that it's all automatic. it certainly wasn't the case when the experiment started, but it might be possible now (QuartzCrystal does auto-updating, after all, and has for a while). Is there any additional interest in this? It'd certainly help reduce a lot of frustration with trying to find latest versions.

Along with this, it also needs a control panel that clearly displays loaded plugins and their version data. This is also slowly in the works...

gtoledo3's picture
Re: illusion

For me, it's the app that goes crazy.

Now, all of this makes sense! It was really eluding me how/why everyone was wanting to make all these visualizers that were just doing a few fps :)

cybero's picture
Re: illusion

I think you're right, I could well be remembering from my very recent pre reinstall and just before building back my workbench experience. Can't reproduce it. Not on 1.3 GL Tools.

gtoledo3's picture
Re: illusion

Do notifications always go out with new plugin posts?

I think it's a really good idea if, as you are saying, you have a little read out that shows current against proposed update.

For instance, I remember a few times where you had betas of Kineme3D with a couple more functions or proposed fixes(? I think I'm remembering this correctly), and then you would post an actual release version that had less features, on the release page. Or maybe I'm wrong on that... but I guess I'm saying there would probably need to be some consideration of using release versions vs. beta? Hopefully I'm making sense here.

I have to say... I think you are totally right on about that thing of moving everything possible (plugins/patches) to the user folders. The only thing I don't put in user folders are my image units and free frame files. I really encourage that to everyone... I think that "environmental protection" idea was a really good one, even if I was reticent about it to start with. I seem to have less voodoo problems.

leegrosbauer's picture
Re: illusion

arrrghh, next thing you know we'll be asking for a Kineme benchmark service to rate our comps by chipsets. pffft. lol.

I got it! One of those little windows can pop up for each composition that will tell you that it has to check your system for compatibility before installing! ahahaha

dust's picture
Re: illusion

that is so funny........ viagra..lol

here is the a quartz crystal render. it helps if you stare at the middle.

dust's picture
Re: illusion

i did install the new gl tools and noticed a slight drop in frame rate yesterday with the splines not sure if i was just paying attention to frame-rate and not noticing the day before i have no idea or because i was adding feedback or all the above.

leegrosbauer's picture
Re: illusion

Sure! Auto update check sounds great. As it stands, I stay up to date by being obsessive-compulsive with clicking on the Beta Releases each day, but during those periods when I'm acting like a sane person I sometimes forget to check for lengths of time that are disadvantageous.

dust's picture
Re: illusion

they where authored with 1.2 but this one was modified with 1.3

dust's picture
Re: illusion

http://d0cut0uch.no-ip.info, i should have the other pointed soon i nuked my server a few days ago and started over.

cwright's picture
Re: illusion

For me, culling and matching-sides modes (wireframe on both front and back) has no effect -- always 10x slower. Maybe that was a fix on the X3100? ;)

One of the (non-public) projects I'm just wrapping up is something that runs at 90-330fps on my GMA950, but stuggles to top 60fps on an NVidia. Not even doing context switching or anything!

cwright's picture
Re: illusion

the way I found to reproduce it was to create a new composition, don't have the visualizer open, drop in a spline patch, and then save (so the spline never actually got rendered) -- the bug you're talking about would save garbage for the input X, Y, Z values. This doesn't happen anymore.

(Internally, the code was using some cached data when it saved, but the cached data was only created/kept up to date when the composition was running -- when you stopped, the cache would get reused for other things, causing a subsequent save to store garbage data.)

dust's picture
Re: illusion

i have some of those illusion slip mats on my turntables that i never use was thinking of scanning them in but yeah im pretty sure its the line art combined with the tb feedback zoom out. this patch wouldn't have happened with out your help though psonice. the rotation is coming from your continuous rotation example. the one where you feed numbers into a que and say if a < b continue on 360 degrees on the output.

leegrosbauer's picture
Re: illusion

Right, but there are three GL Tools v 1.3s, yesterday's release and the release of July 4th didn't get point upgrades.

... and, I just wanted to say thanks again for the lovely spline compositions, Dustin. They are truly beautiful.

cwright's picture
Re: illusion

What I've found is that some processes (Finder, Dock, and Spotlight, in particular) will fire up QC for something mundane (generate a thumbnail, get some composition metadata, whatever), and then Never unload the plugins. Usually this is ok, but sometimes attempting to replace a plugin with a new one won't go as smoothly, because something's still holding on the old version. For me, GLTools is one particularly clingy plugin (Finder keeps it around forever for some reason -- need to figure that out someday...)

I don't think I've ever regressed releases over betas -- a 1.x public release will always be as featureful as a 1.x beta. The only oddity with that in Kineme3D was that there were some weird smoothing bugs in one beta, so I reverted to the old stable slow smoothing algorithm (I needed to get 1.1 available to users because 1.0 was exploding in VDMX, causing no end to problems everywhere) -- normally that's not the case though.

cwright's picture
Re: illusion

the 20090707 beta contains a lot more debugging info than previous builds (there have been lots of nebulous error reports that I'm trying to hunt down, so adding debugging info helps me track down issues more quickly) -- that said, it shouldn't have a sever impact on performance. Load time will be longer (because there's more actual data), and execution time will be slightly longer (previously we omitted frame pointers, but now we're not so we can get accurate and meaningful backtraces) -- framepointer omission is only 4-6 instructions per function, so it's not a large performance hit most of the time (there are sometimes weird exceptions, but they're usually not too noticeable).

Splines also got some optimizations (thanks to my recent R&D with Kineme3D, documented on perpendiculo.us), so they should be about the same or slightly faster, performance-wise. If you can find a composition that clearly demonstrates the performance problem between 20090707 and a prior version, please let me know so I can run some regressions on it.

dust's picture
Re: illusion

you can do chipset testing with a logic patch and a gl info patch then grep out the string for the supported chipsets and bang the enable of the chipset is in the list. or use a javascript might be a little less noodles.

leegrosbauer's picture
Re: illusion

dust wrote:
... the rotation is coming from your (psonice) continuous rotation example. the one where you feed numbers into a que and say if a < b continue on 360 degrees on the output ...

A link to that continuous rotation example, perhaps? errr nevermind, I believe I located it in illusion-flower. Thanks to you both.

gtoledo3's picture
Re: illusion

No doubt. There's still the "finding out" process though :/

gtoledo3's picture
Re: illusion

Woah. It's amazing to see that going at a fast clip. That is gnarly.

cybero's picture
Re: illusion

swift

leegrosbauer's picture
Re: illusion

dust wrote:
you can do chipset testing with a logic patch and a gl info patch then grep out the string for the supported chipsets and bang the enable of the chipset is in the list. or use a javascript might be a little less noodles.
Well, most in here might be able to do those things, but not me. heh.

I was just kidding about the benchmarking, btw. Although ... regardless as to my own lay-person limitations, it is indeed looking like some really savvy folks amongst us are nonetheless not aware, in a timely fashion, whether or not a given composition is performing as well on their particular system as it might be performing on other systems. Guess there's no doing anything about easing that dilemma, however.

usefuldesign.au's picture
Re: illusion

Seeing as you guys seem to share my love for optical illusions check this out. It's not a great programming feat. I chanced upon it when exploring the GL Quad diagonal tear on my Nvidia FX GeForce 5200 card.

Very importantly VIEW FIRST this comp BEFORE you examine in the editor.

Ask yourself "how is this programmed"? Make a decision then check the answer.

Just looked at the comp myself, very overcomplicated for what it is showing, must have been exploring another idea at same time.

The illusion is the point: ie. anthropic colour recognition has numerous psychological components that mean we do not 'see' like a camera.

PreviewAttachmentSize
optical illustration 1.qtz13.46 KB

cybero's picture
Re: illusion

proving booleans have a persistence of vision with both maths and javascript

usefuldesign.au's picture
Re: illusion

hmm I think it proves I was in a weird mood when i made that.

Here the bit that matters. If I ever have to do a colour management lecture this is going in at the start.

Works best to go full screen and just stare at the central square. What is happening to the RGB values of the centre square?

PreviewAttachmentSize
optical illustration.qtz8.53 KB

usefuldesign.au's picture
Re: illusion

Now I get it… that's smoooooth