Actually, I usta have a bet on this one, to whit - betcha can't put on this vr helmet and not rip it off in two minutes or less. Always won.
Why - system read compass and attitude sensor from headset, used this to control logic simulating about Mach 1.5 over a 90 meter resolution terrain (ok, so we dropped all the linear/near linear pts to get the pig to run, it was a few years ago). Whenever the machine detected a head wobble, it'd pull an inverted normal change to camera direction (turns opposite from tilt 90 degrees). And yup, when your eyes tell you you're screaming along over the ground at some ridiculous speed and doing 90 degree turns, and those turns are basically out of phase with your own head wobbles ___and your ears say "ain't nothing happening", you turn green. Every time.
So this is a long way of trying to point out that immersion ain't just the immediate see and hear bits, it's also what you feel, and what "subsystems" of your own report on the physical environment. Tell the right lie and a 90 foot fall is really about 3/4 inch (Disney) or even apply a light voltage and fool the balance organs (Stanford ????) but if you don't cover every base, you may soon find motion sickness acts like a system panic message indicating someone is lying about something.
Mutatis mutandi is latin for (roughly) "all things considered".
If you work through the math, it's the cost of (8) lights in the scene.
Software rendering (assuming efficient) is often cumulatively more expensive than hardware, yet is not necessarily a direct multiple. HUH ? In other words, in software, you got memory, it's kinda evident you can cache a bunch of values during computation, i.e. when you calculate the surface normal (direction surface is "pointed") you can cache this and use (instead of recompute) later. In hardware, this may be too expensive, it really may be cheaper to recompute the whole shebang every time. OK, lets try this again. Esp since my first post was deemed so bad that the question in response to it scored higher. Don't they teach Shannon in school anymore ???-D If you're doing it in software, you by definition have scratch memory. Your fundamental algorithm is slower, but it has scratch memory to store temporary results, which improves it's efficiency when solving roughly the same problem multiple times, i.e. multiple lights On the other hand, a hardware solution may literally be a set of transistors wired to an addressable register (variable) which results in a solution that is damn fast, but incapable of remembering temporary results (register only holds the source values, temporaries recalculated each time) so time per iteration is significantly faster, but overall time is simple linear product, not an exponentially decreasing product. Oh damn. Lets try again. If you solve a problem slowly, but can remember and use interim value to accellerate the solution of related problems, then if you have ANY machine that can solve the problem faster for a fixed number of solutions, there is a point at which the slow solution is always the best choice, because over N calculations, the calculation savings for the cached values exceed the time of efficiently recalculating the solution from scratch. Urrk! I've had a busy day, and my brain is seizing. If in pure software it takes me X to compute a single light and 1/2X as much time to compute each subsequent light, or I have hardware that computes any one light in 1/2X time, then obviously if I have five lights, software is much better. One light, and it's the other way around. That's the main reason grafix systems can get so damn persnicety about the number of lights in a scene. Back to the original post and your first assertion, yup - if the number of lights you use exceeds the hardwares max, you're going to see one hell of a falloff in scene rendering speed. If the solution is hardware based (likely) and there are a finite nunber of "potential" solutions, when the solution requirement exceeds the hardware limit, there's going to be a nasty falloff in rendering rate, cause a) software has to do it, and b) nobody remembers a damn thing about earlier results, Wow! How cathartic, for reasons I can't even explain. Yo, MSFT, methinks a little detail is the fact you're claiming a straight line path based on # lights, i.e. you don't cache a damn thing!
OK, this is enough. Everything that involves light falling on it (photons from lite, photons to your eye, everything) means you need to work thru geometry. Whole damn thing is what do you compute, what do you recall, how do those costs blend to form best algorithm. Damn nasty question, JC is bright damn fellow, but he ruled out the big outdoors to get his answer. Yes, that's a deliberately cryptic SD troll:-D~
Umm, 3D labs is making a good run for worst opengl implementation on Win2K, supposedly the future of winders...
1) Oxygen VX1 driver for Win2K is pretty poor Problems with texture management, loosing current bindings Regular snow squalls
2) Permedia2 reference driver nonexistant for W2K
and then.... The damn driver that ships with W2K for permedia2 (actually one from msft and one from 3dlabs) accellerates D3D !but not openGL!, or if this is nt this is an opengl only accelerator if this is w2k this is a D3D only accelerator
I agree tho, I like nVidia. I mean like I like nVidia even when we're talking about spending some money on the card. More for less, and it works.
I dunno if I'd go that far, but I can see some of the apps we do (VR for teaching dr's how to do needle sticks) being greate canidates for the x box. Right platform for other things than just games.
Actually it always seems to work out the same way, the time you save by caching is the time you loose in managing the data, turns out Knuth's been on the money all the time. Just hard to understand. Anyway, easiest way to visualize the problem is that the contribution a light makes to the scene in primitive lighting systems is based on what that light can see, i.e. the calculation of lighting from a directional light is fundamentally the same as calculating the view, mutatis mutandis. I was actually hoping to see some kinda hint as to "snazzy new heuristic" but it smells like the solution is hardware based. I wonder what happens if you use more than n lights, where n is some chip level register limit ?
A switch in any network is a box that makes and breaks connections between the wires that flow in and out of it. These days, big busy switches use optical fiber cause of the sheer volume of data (noise) at their level. So the real deal is we can move from the bad old way of a) convert light to electrons, b) use electrons to make path between the incoming light fiber and the outgoing light fiber, and c) convert back to light to the brand new way of directly guiding the photon where we want it. In terms of development, it beats the hell out of vibrating mirrors.
Yeah, I see the RISC issues fine, you can look at "code morphing" as nothing more than microcoding, if they did ok on risc architecture the standard benefits accrue. But what really intrigues me is the variable nature of the speed control. If that's a completely on the fly control where the os can alter processor execution speed whenever it wants, I'd bet that this might indeed translate to pretty incredible gains in power usage depending on _how_ you use it. Most of the time, a modern desktop os is blocking on the slowest peripheral:-) And if they really can change speed on the fly, think of the benefits for Java/JNI space processors. A constantly variable processor clock is a cool idea, as long as you can still keep absolute time
Hmmmm. I don't think the secret police example can really fly. It's basically saying the keys got comprimised, and it's the same if I take out your eyeball, snag your retinal scan for an authentication key, and use that to grab your private pgp key. Any crypto that interfaces with the outside world through some form of ident/authent is gonna be weak at that point. Show Clinton a pretty girl and I bet somebody sweats security on that little black bag. In terms of using these random data blocks as representational strategy, i.e. my message can be defined as Pn = Pm XOR Pt, where Pm is a true random block and Pt is the plaintext, Pn isn't actually random,it's a regular product of a function applied to Pa and the plaintext message. Do this a bunch of times and it's definitely obfuscated beyond belief, but it ain't crypto. However, for grins, we could say, imagine a repository of pads, zillions and zillions on them. You go to this repository, and you say, here's a plaintext message, and I want you to run it through X pad permutations. Assume further this repository, or a napster like distributed version, can associate a unique index with each pad. We take your plaintext, smush it against some randomly selected pad, and generate a new pad from that. That new pad both goes into the soup for other conversions, and also forms part of your converted message, i.e. myNewPad == Pad_sub_random XOR Pad_sub_plaintext
myNewPad,Pad_sub_random,Pad_sub_plaintext are all expressable as unique scalars corresponding to an index to the pad in question.
OK, this means that a plaintext message is equivalent to an xoring of two pads selected by index. OK, say I smash these two values together. Theoretically, 2^256 should do fine, so I make a number MyPlainText == (Pad_sub_random 256) + Pad_sub_xorProduct;
Hmmmm. This suggests godel numbers to me. I can take an arbitrary collection of plaintext, convert it to a number, and loose forever that plaintext, keeping only a weird kinda table of coefficients in the form of these random and random xor plaintext blocks. Bottom line is I can use this wonking big number and the system knows exactly wot I got, even tho it don't directly got it, it gotta derive it.
Gee, ahh, duhhhh. I had this bad thought too late, went and checked, and yes, if anybodies seen my sense of humor please return it to me. I have stenciled the meaning of sarcasm on my right palm and do understand that the initial poster is not recommending we all become porcine programmers wallowing in wasted cycles.
Yup. I think this stuff is pretty interesting, would enable a nice interactive 3d demo of a working watch. Brute force ain't a solution in the new worlds of immersive 3D systems, the interfaces that are evolving to consume all these processing cycles. In the old days, it used to be fun to replace the unix idle code with an approximator for PI, i.e. when your machine wasn't doing anything else, it found digits for PI. Nowadays, when it's idle, it uses PI a gazillion times to draw 3D things all over your screen. In theory, the "standard user experience" is fundamentally driven by systems where the full capacity of the system is exploited to make the users life easier, more productive, whatever. So brute force is a cheap temporary system in those brief periods where user demands don't exceed machine specification. Most of the time, software is in the position of scrambling to keep up. In my world, knowing how the mechanisms of gears works and successfully communicating it to a machine means that the machine can now do it efficiently for many cases. Brute force means I can't maintain frame rate. When window based systems became popular, awesome DOS machines turned into absolute dogs under windows, cause the demands on them went up an order of magnitude. The same thing will happen again as 3D moves from games into the general environment, once more it's gonna be shaving cycles. Given this, the only justification I've ever seen for brute force is as a temporary solution while you make the real one, or because it's an npc problem and you're just gambling anyway. Over time the effective lifetime of software, I'd argue that the lifetime of a knowledge based solution always exceeds the lifetime of a brute force solution. I guess I'll end my rant now, it's just that I think anyone who buys that is by definition a dinosaur, i.e. they have only the present moment, no future at all.
If you mean fly through the wireframe boxes representing the filesytem,yup that was an sgi demo for quite a while. Another version of text in 3D, seems neat before you see it, tedious and pointless afterwards
I've always wondered if some of the issues in not releasing star wars and jurassic park series on dvd was due to the high resolution and easily available format of some of lucas's fundamental product, i.e. computer graphics thingies that can be accurately recaptured off of dvd. videotape is too damn noisy save for amateurs, but dvd data is rich enough to swipe, redigitize, and then alter just enough to be legal. The fan market for merchandise would be cannibleized, and the actual properties themselves could be used by competitors to cut down on their workloads. Wait six years as 3d,computer,tv,blahblah converge, and between amortization and changes in fundamental cost structures, the point is moot. But right now, put out a dvd, and quite a few companies will have cute digital dinos for one thing or another, just far enough off you can't sue em. If there's any mba's out there, what's the net present value of dvd sales against the cost of cannibleizing the fan market and potentially aiding competitors ?
OK, I've been very curious about this myself, and I know the key wouldn't be called nsakey for the simple reason that those types +have+ to come up with a name for the project, like ohh, Silent Storm. However the mechanism still bothers me. I am rusty on the mechanism here, but as I understand it, if you have the correct private components of either of these two keys, you have the ability to remotely change crypto behavior on an NT system ? I guess in essence, I am quite curious about who holds each of the private keys that go with the public key information, and what rights can they extert through that key ?
Oh you bet, the few faltering first steps on that path are the core of our business model. Tad Williams Otherland series is more likely a "better vision" than the Matrix tho, in terms of the nature of the beast. The matrix plot assumes some kind of giant hierarchy, in loose terms, whereas real functioning networks, are, well, they're loose. In short, not even machines could muster the level of agreement to manifest a single complex thing such as the matrix. We must build partitioned worlds before we can build a seamless whole, and hell, some still think VRML is a good idea. So we do have a ways to go. At a certain level of system evolution, my enthusiasm does begin to drift towards trepidation. My personal experiance with VR systems is that they can be quite "reality bending", especially when they are deliberately matched to the real world. That is to say, they perform a deliberate assault on your ability to detect the difference between synthesized virtual information and real physical information. Admittedly our first steps are still pretty crude, on the other hand, we can generate really spectacular misfires. My personal favorite are immersive VR goggles. Fly someone right off the deck at 900Kph, and make a few 90 degree left and right turns at that speed. With their brain seeing a view that tells them one thing, their innear ear saying something completely different, most participants turn a delightful shade of green about thirty seconds into the process. This is a brutal example of a usually more subtly occuring phenomenon, the fact that users already delightfully muddle up the differences between reality and the infospace whenever the frame rate is high enough. That's the entire reason for all those pretty pictures from supercomputers, as well as the basis of companies like id. Of course I want people to get into VR, I have spent a long time trying to help make that possible. I don't think we can actually "build" a virtual space like we built the moon rockets, because the nature of virtual environments and the systems that form them is fundamentally much messier, as it's a dynamically evolving system. I've also learned to have a healthy respect for the interaction of another dimension in my user interface and my visual cortex. People who have forgotten this have already lost billions on wall street, by basing their actions on ephemera in virtual worlds that they thought they saw. The field seems to exert a siren song over us all, and I think the interaction of the technology and us deserves healthy respect. There is something to be said for the aquisition of powers to reshape reality, virtual or otherwise.
Umm, having worked in the guts of VR systems for too many years now I just have to cough up 2c. I am a geek, I liked the matrix, but my nongeek friends all like the matrix too. Can't a movie just be phun?! Now, onward into deconstruction.
There is no question any complex simulation has certain assumptions hardwired into the fabric of the system. In fact, it is a process that made Kurt Godel famous, basically the formalization of the fact that if you create any formal expression system (the matrix) the very system contains elements which cannot be expressed in it. Hence phone booths are ok, cause this may be the way the fundamental system is wired. The system itself is incapable of adjusting or altering them because it cannot "express" them.
Agents and stuff are simple, one can make the argument that since they don't exist at all, this is simply the representation "you" are seeing. In this world, 10 people can look at the "same thing" and it really is 10 different nonthings.
OTOH, beating such a system _cannot_ be an issue of "I've got more firepower than you in the shared concensus", i.e. the gunfight in the lobby. One interesting thing about a lot of VR simulations is that they have to be slowed down for the human observer. In short, in many cases you beat the computer in twitch games cause the developers deliberately waste enough time to give you a chance. We've got a demo for our system that flies you through a nasty corkscrew under the computers control. Under human control, nobody has yet made it through the second turn, and there are 32 such turns at speed. Hence, the way you beat such a system is by co-opting it against itself, by reexpressing the paradigm in such a way that you have the advantage. You've got one to start with, the machine is stuck there, it can't leave. The flomo bullet sequences are a good example. I do believe that this is the true meaning of the verb "to hack". But the first floor shootout was just stupid, imho. What would I have done is to say, OK if you know it's a phantom, shape the phantom. The effect I'd pick would be the Xaos bubble derez sequence as seen in lawnmower man 1, I'd do it at an order of magnitude higher resolution, and I'd just derez the first floor. 10 to 1 the core logic fabric then drops the second floor right into place. All you have to do is not be crushed:-)
I can't buy the extrapolation about the growth of code for blocking and suppressing things on the net is a free speech issue. The fundamental argument/premise is that spam and any other form of wanton broadcast is at it's heart the complete inverse of the dataflow of the net. Specifically, the net is about the ability of an individual to reach out for what they want, TV is for throwing things at them. I would argue that spam list killers and the like are not infringments on free speech because of the fact that they are all voluntary optins. If your provider opted you in and you don't like it, pick another provider! On the net, an infringment of free speech comes from the readers perspective, not the publishers. China infringes the free speech of their citizens (stop laughing) by denying them the ability to reach out to the information they want. They are prevented from accessing information published and waiting for them. Summarizing, the net is a pull medium, it functions by responding dynamically to the demands of those using it. Antispam issues are basically issues of a public nuisiance nature, as it is so easy for any net user to decide for themselves what they do and don't want. Access to sites is an issue of free speech. The ability to make a site can even be called an issue of free speech (scary), but filtering unsolicited data has nothing to do with it based on the fundamental nature of net communications
You can get through the real world by being oblivious to Euclid, trig, and calculus. But that doesn't mean you don't need or use them all the time. Think of this the next time you drive through a turn on the freeway in your car. A minor nit, but I just couldn't help myself. Math or programming, it's all really the same thing. Hence, you can make programs out of math and math is the foundation of all programming. It's all just formalized symbolism to us chickens.
We certainly aren't fans of Microsoft's stated intentions to provide only DX8/Fahrenheight as a supported 3D library. Already WGL is seeming an orphan child. It's weird, but right now John Carmack is worth a lot to my business, cause he's holding to openGL. IMHO this is a monumentally stupid blunder on Microsofts part. First, if they are aiming this at the real-time games market and the speed over quality requirements of that market and can't even get Carmack to bite,......HELLO Secondly, they are in effect forcing companies like mine aiming for the future when high end interactive visualization goes onto desktops to make plans for ports to other systems, for the simple reason they've now tarted their platform up too much for the game developers. Finally, this effort of theirs is helping the mass market for 3D games software, but it hurts everything else. Already, companies like Dell, Micron, and others have high markup workstation lines. Wintel workstations are usually treated as a high margin item. Gee, guys --- wouldn't it be better if we drove these systems into the mass market at the mass market price points. Exactly who gets to say, OK thats enough power, you'll have to pay premium beyond this point ? I understand supply and demand, I argue that across the board demand for intellegently applied processor power is still a big sea. So maybe we shouldn't make a peep, just be quiet and smile. When Linux and its friends are the first to provide workstation level features and performance and mass market interfaces (GUIS,SUIS) at a mass market price point....Game over. Microsoft looses. Because then the development equation becomes one of "well, it would cost us X to make this on Linux and 10x to do it on the Windows Game Playplatform" So lets do it on Linux first. Cool for Linux, pathetically stupid on Microsofts part.
Very good. I agree wholeheartedly. I don't think that DX is there yet, as I agree with the fundamentals of John Carmacks arguments. I violently object to Microsofts plan to try and make DX8/Fahrenheight the service for all users. The best world is separate apis for games (speed counts uber alles) and openGL (speed counts, but not at the expense of trashing the visuals). Probably is a common ground between the two at the hardware interface/bit blasting level, but after that they serve different masters
The standard wisdom in the CG community is that feature movie work is pretty rigid and demanding, and there is little room for creative flexibility on the CG side of the house. The result of this is a steady exodus of CG people from the feature movie side into advertising, where there is more creativity and ingenuity required of them. One comment in particular I remember was from an ILM artist who left after the completion of JP because he had just spent the last 18 months of his life making a dinousaurs butt wiggle realistically and wanted something with more freedom of expression. That said, the cast of technical persona at ILM contains some of the finest CG talent, tech and art. The only real gaffe I can't forget was in the nasty bluescreen bleed/scaling errors that was in the shot oof the plane dropping off the blimp in the last of the IJ movies.
Oh yeah, I forgot. Mea culpa slashdot. We are interested in other platforms than Windows. Gee, if you use that light streaming through the window trick, whatever will you do on linux/mesa, mac/opengl, sgi/(open)gl, etc ?
Hmmm. Our system is being used by a game company to try out a new idea, but it's also already the basis of a commercial app that does surgical training simulation for Drs. One of the main reasons we use openGL is that we can do things "correctly", i.e. the patient model lighting we provide is basically a crude approximation of what the Doc will see when they stick a needle in a real patient, albeit real patients aren't see through. John Carmack at id uses openGL to do things we wouldn't dare, such as the lightmaps, as his needs are different than ours, and nobody is going to put dead demons or players on the operating room table. But openGL satisfies us both. DX on the other hand has many of these "clever tricks" to do things like streaming light through windows, but places much less thought and emphasis on the general issues of rendering realistic images for a wide variety of tasks. Do you want to be operated on by a Doc who got any of his training on something optimized to do gore and explosions ? And even in the submarket they court, there are some knowledgable people, e.g. Carmack, who still tell them to stuff it. What amazes me about the DX product are two specific issues. 1) The mantra at Microsoft is framerate,framerate,framerate. Good mantra. Unfortunately, they feel this justifies them in adding internal logic to deliberately degrade the geometries they have been given in order to maintain framerate. Thats the deal in their new focus on curves, to maintain framerate, __not to increase visual quality__. Lessee, we have apps for Docs, soldiers, and others who really really really don't need this. And I'm sure id isn't thrilled about having many ducats invested in their artwork, and loosing control over the tradeoffs on responsiveness against visual quality. Hey, we're supposed to know what we are doing. If msft wants to sell tutorial aids, thats fine, but some of us carry high development expenses specifically because we wish to make those decisions ourselves. Hence we tend to like openGL. 2) MSFT went out and hired Jim Blinn (making obligatory obiesiance now to the great Blinn), and put him in their advanced R&D area (I think). Last summer he talks about the 10 things still undone in CG, and delivers a pretty specific complaint about clever but narrow and unstable tricks used to improve visual realism in a narrow range of applications. He even gently points the finger at 3D games. Does Microsoft seem to pay any attention to him. No! I personally find that disgusting. Summarizing, I guess I'm saying that as long as msft puts their efforts into stupid optimizations like this, to paraphrase the cryptogeeks, you can have my openGL libs when you pry them from my cold dead mind. It does seem some daze that msft is willing to go to that length, WGL is apparently doomed now.
Re:ahh, the wonders of reverse engineering...
on
Creative ports Glide
·
· Score: 1
DX6 to openGL translation is certainly not "impossible", I think scitex has a product out/coming called DirectGL. There are a number of openGL to DX6 translators out there, but this seems to be mostly a msft attempt to bring certain strategic apps into the DX fold, even if they really don't want to be there. My own take is that msft has put too many screwball optimizations (used advisedly) into DX6 to satisfy gamers that make the conversion potentially difficult, and certainly inefficient. Why invest all the time and energy into an app to get performance and visual quality, and then pass all your work through a translator? I think this is the central reason that there are such battles over graphics libraries, because efficient use of the library really requires an exclusive commitment to that library. And even in the game development community there is a lot of dissention, our fave is John Carmack, who has given the entire 3D efforts at msft some very good critiques from both technical and business perspectives. Bottom line is that computer graphics will remain a crude approximation for years to come, as reality is unfortunately infinitely complex and computers aren't. This kind of squelches translators for anyone trying to push the envelope.
Ummm, do you mean OpenGL OR Direct3D ? Or are you implying there is no difference? The first product out with our VR system is a training tool for surgeons. It is meant to run on pcs, but we certainly aren't interested in a tech (D3D) where the only input msft takes is from established game developers. The Dr who developed the app on top of our engine is horrified by the prospect of having to switch to D3D/DX8 if msft succeeds in eradicating openGL support. Basically, D3D is for game developers who are not of John Carmacks level. openGL is for anyone who is really serious about the images they put on the screen (and yes, I do mean realtime images)
Being in the VR biz, it's a subject that gets thought of from time to time. A few observations
1) OK, the math geeks have weighed in, and I only solve big polys, I'm not a math geek. Comments on transfinite numbers are an effective technical containment for this conversation, so I accept those as givens.
2) The real interesting bit is the historical meaning of real estate (this is the web, go to altavista or wherever). A fundamental truth is that it is completely based on a finite nonrenewable (yes recirculating) supply. Web space is not infinite, but at the same time it is not finite the way physical real estate is finite. Hence when the web has pricing metaphors for space that mimic the real estate model things get interesting, theoretically, and very lucrative, commercially. It's the conflict between extensibility (the web) and non-extensibility (real estate). I think that the real real estate (hmm, real^2 estate?) will be real interesting when we have real companies like bechtel making new real estate, e.g. undersea living. Yes I agree this should be done in an intellegent fashion, I do have a son and therefore do care about issues that extend beyond my lifetime.
3) As far as infinite real estate for cool new sites, don't I wish. It would theoretically make anyones chance of site success much bigger. I think the deal is that as individuals or as cultures, we have only so many balls we can keep in the air at once. It's that old saw that people can keep seven things in their head at once with a.333 error bar on that number. So there is a finite and limited set of "major approved web destinations" for each user, group of users, culture of users, pool of all users. Hence as this becomes saturated, to enter that zone is to need to bump someone else out. Happily, the net is very dynamic, so this does happen on a pretty regular basis, but I would call the sets of web sites held in memory from individuals to the collective web community to be the most limited real estate on the web. I have found slashdot. I like slashdot. Someone I used to like is going to have to get bumped down to "the things I have references to but never directly remember"
Slashdot Mirror
Actually, I usta have a bet on this one, to whit - betcha can't put on this vr helmet and not rip it off in two minutes or less. Always won.
Why - system read compass and attitude sensor from headset, used this to control logic simulating about Mach 1.5 over a 90 meter resolution terrain (ok, so we dropped all the linear/near linear pts to get the pig to run, it was a few years ago). Whenever the machine detected a head wobble, it'd pull an inverted normal change to camera direction (turns opposite from tilt 90 degrees). And yup, when your eyes tell you you're screaming along over the ground at some ridiculous speed and doing 90 degree turns, and those turns are basically out of phase with your own head wobbles ___and your ears say "ain't nothing happening", you turn green. Every time.
So this is a long way of trying to point out that immersion ain't just the immediate see and hear bits, it's also what you feel, and what "subsystems" of your own report on the physical environment. Tell the right lie and a 90 foot fall is really about 3/4 inch (Disney) or even apply a light voltage and fool the balance organs (Stanford ????) but if you don't cover every base, you may soon find motion sickness acts like a system panic message indicating someone is lying about something.
Mutatis mutandi is latin for (roughly) "all things considered".
:-D~
If you work through the math, it's the cost of (8) lights in the scene.
Software rendering (assuming efficient) is often cumulatively more expensive than hardware, yet is not necessarily a direct multiple. HUH ? In other words, in software, you got memory, it's kinda evident you can cache a bunch of values during computation, i.e. when you calculate the surface normal (direction surface is "pointed") you can cache this and use (instead of recompute) later.
In hardware, this may be too expensive, it really may be cheaper to recompute the whole shebang every time.
OK, lets try this again. Esp since my first post was deemed so bad that the question in response to it scored higher. Don't they teach Shannon in school anymore ???-D
If you're doing it in software, you by definition have scratch memory. Your fundamental algorithm is slower, but it has scratch memory to store temporary results, which improves it's efficiency when solving roughly the same problem multiple times, i.e. multiple lights On the other hand, a hardware solution may literally be a set of transistors wired to an addressable register (variable) which results in a solution that is damn fast, but incapable of remembering temporary results (register only holds the source values, temporaries recalculated each time) so time per iteration is significantly faster, but overall time is simple linear product, not an exponentially decreasing product.
Oh damn. Lets try again. If you solve a problem slowly, but can remember and use interim value to accellerate the solution of related problems, then if you have ANY machine that can solve the problem faster for a fixed number of solutions, there is a point at which the slow solution is always the best choice, because over N calculations, the calculation savings for the cached values exceed the time of efficiently recalculating the solution from scratch.
Urrk! I've had a busy day, and my brain is seizing. If in pure software it takes me X to compute a single light and 1/2X as much time to compute each subsequent light, or I have hardware that computes any one light in 1/2X time, then obviously if I have five lights, software is much better. One light, and it's the other way around. That's the main reason grafix systems can get so damn persnicety about the number of lights in a scene.
Back to the original post and your first assertion, yup - if the number of lights you use exceeds the hardwares max, you're going to see one hell of a falloff in scene rendering speed. If the solution is hardware based (likely) and there are a finite nunber of "potential" solutions, when the solution requirement exceeds the hardware limit, there's going to be a nasty falloff in rendering rate, cause a) software has to do it, and b) nobody remembers a damn thing about earlier results,
Wow! How cathartic, for reasons I can't even explain. Yo, MSFT, methinks a little detail is the fact you're claiming a straight line path based on # lights, i.e. you don't cache a damn thing!
OK, this is enough. Everything that involves light falling on it (photons from lite, photons to your eye, everything) means you need to work thru geometry. Whole damn thing is what do you compute, what do you recall, how do those costs blend to form best algorithm. Damn nasty question, JC is bright damn fellow, but he ruled out the big outdoors to get his answer. Yes, that's a deliberately cryptic SD troll
Umm, 3D labs is making a good run for worst opengl implementation on Win2K, supposedly the future of winders...
1) Oxygen VX1 driver for Win2K is pretty poor
Problems with texture management, loosing current bindings
Regular snow squalls
2) Permedia2 reference driver nonexistant for W2K
and then....
The damn driver that ships with W2K for permedia2 (actually one from msft and one from 3dlabs) accellerates D3D !but not openGL!, or
if this is nt this is an opengl only accelerator
if this is w2k this is a D3D only accelerator
I agree tho, I like nVidia. I mean like I like nVidia even when we're talking about spending some money on the card. More for less, and it works.
I dunno if I'd go that far, but I can see some of the apps we do (VR for teaching dr's how to do needle sticks) being greate canidates for the x box. Right platform for other things than just games.
Actually it always seems to work out the same way, the time you save by caching is the time you loose in managing the data, turns out Knuth's been on the money all the time. Just hard to understand.
Anyway, easiest way to visualize the problem is that the contribution a light makes to the scene in primitive lighting systems is based on what that light can see, i.e. the calculation of lighting from a directional light is fundamentally the same as calculating the view, mutatis mutandis. I was actually hoping to see some kinda hint as to "snazzy new heuristic" but it smells like the solution is hardware based. I wonder what happens if you use more than n lights, where n is some chip level register limit ?
A switch in any network is a box that makes and breaks connections between the wires that flow in and out of it. These days, big busy switches use optical fiber cause of the sheer volume of data (noise) at their level. So the real deal is we can move from the bad old way of a) convert light to electrons, b) use electrons to make path between the incoming light fiber and the outgoing light fiber, and c) convert back to light to the brand new way of directly guiding the photon where we want it.
In terms of development, it beats the hell out of vibrating mirrors.
Yeah, I see the RISC issues fine, you can look at "code morphing" as nothing more than microcoding, if they did ok on risc architecture the standard benefits accrue. :-)
But what really intrigues me is the variable nature of the speed control. If that's a completely on the fly control where the os can alter processor execution speed whenever it wants, I'd bet that this might indeed translate to pretty incredible gains in power usage depending on _how_ you use it. Most of the time, a modern desktop os is blocking on the slowest peripheral
And if they really can change speed on the fly, think of the benefits for Java/JNI space processors. A constantly variable processor clock is a cool idea, as long as you can still keep absolute time
Hmmmm. I don't think the secret police example can really fly. It's basically saying the keys got comprimised, and it's the same if I take out your eyeball, snag your retinal scan for an authentication key, and use that to grab your private pgp key. Any crypto that interfaces with the outside world through some form of ident/authent is gonna be weak at that point. Show Clinton a pretty girl and I bet somebody sweats security on that little black bag.
In terms of using these random data blocks as representational strategy, i.e. my message can be defined as Pn = Pm XOR Pt, where Pm is a true random block and Pt is the plaintext, Pn isn't actually random,it's a regular product of a function applied to Pa and the plaintext message. Do this a bunch of times and it's definitely obfuscated beyond belief, but it ain't crypto.
However, for grins, we could say, imagine a repository of pads, zillions and zillions on them. You go to this repository, and you say, here's a plaintext message, and I want you to run it through X pad permutations. Assume further this repository, or a napster like distributed version, can associate a unique index with each pad. We take your plaintext, smush it against some randomly selected pad, and generate a new pad from that. That new pad both goes into the soup for other conversions, and also forms part of your converted message, i.e.
myNewPad == Pad_sub_random XOR Pad_sub_plaintext
myNewPad,Pad_sub_random,Pad_sub_plaintext are all expressable as unique scalars corresponding to an index to the pad in question.
OK, this means that a plaintext message is equivalent to an xoring of two pads selected by index.
OK, say I smash these two values together. Theoretically, 2^256 should do fine, so I make a number MyPlainText == (Pad_sub_random 256) + Pad_sub_xorProduct;
Hmmmm. This suggests godel numbers to me. I can take an arbitrary collection of plaintext, convert it to a number, and loose forever that plaintext, keeping only a weird kinda table of coefficients in the form of these random and random xor plaintext blocks. Bottom line is I can use this wonking big number and the system knows exactly wot I got, even tho it don't directly got it, it gotta derive it.
Gee, ahh, duhhhh. I had this bad thought too late, went and checked, and yes, if anybodies seen my sense of humor please return it to me. I have stenciled the meaning of sarcasm on my right palm and do understand that the initial poster is not recommending we all become porcine programmers wallowing in wasted cycles.
Yup. I think this stuff is pretty interesting, would enable a nice interactive 3d demo of a working watch. Brute force ain't a solution in the new worlds of immersive 3D systems, the interfaces that are evolving to consume all these processing cycles. In the old days, it used to be fun to replace the unix idle code with an approximator for PI, i.e. when your machine wasn't doing anything else, it found digits for PI. Nowadays, when it's idle, it uses PI a gazillion times to draw 3D things all over your screen. In theory, the "standard user experience" is fundamentally driven by systems where the full capacity of the system is exploited to make the users life easier, more productive, whatever.
So brute force is a cheap temporary system in those brief periods where user demands don't exceed machine specification. Most of the time, software is in the position of scrambling to keep up. In my world, knowing how the mechanisms of gears works and successfully communicating it to a machine means that the machine can now do it efficiently for many cases. Brute force means I can't maintain frame rate.
When window based systems became popular, awesome DOS machines turned into absolute dogs under windows, cause the demands on them went up an order of magnitude. The same thing will happen again as 3D moves from games into the general environment, once more it's gonna be shaving cycles.
Given this, the only justification I've ever seen for brute force is as a temporary solution while you make the real one, or because it's an npc problem and you're just gambling anyway. Over time the effective lifetime of software, I'd argue that the lifetime of a knowledge based solution always exceeds the lifetime of a brute force solution.
I guess I'll end my rant now, it's just that I think anyone who buys that is by definition a dinosaur, i.e. they have only the present moment, no future at all.
If you mean fly through the wireframe boxes representing the filesytem,yup that was an sgi demo for quite a while. Another version of text in 3D, seems neat before you see it, tedious and pointless afterwards
I've always wondered if some of the issues in not releasing star wars and jurassic park series on dvd was due to the high resolution and easily available format of some of lucas's fundamental product, i.e. computer graphics thingies that can be accurately recaptured off of dvd. videotape is too damn noisy save for amateurs, but dvd data is rich enough to swipe, redigitize, and then alter just enough to be legal. The fan market for merchandise would be cannibleized, and the actual properties themselves could be used by competitors to cut down on their workloads.
Wait six years as 3d,computer,tv,blahblah converge, and between amortization and changes in fundamental cost structures, the point is moot. But right now, put out a dvd, and quite a few companies will have cute digital dinos for one thing or another, just far enough off you can't sue em.
If there's any mba's out there, what's the net present value of dvd sales against the cost of cannibleizing the fan market and potentially aiding competitors ?
OK, I've been very curious about this myself, and I know the key wouldn't be called nsakey for the simple reason that those types +have+ to come up with a name for the project, like ohh, Silent Storm. However the mechanism still bothers me. I am rusty on the mechanism here, but as I understand it, if you have the correct private components of either of these two keys, you have the ability to remotely change crypto behavior on an NT system ?
I guess in essence, I am quite curious about who holds each of the private keys that go with the public key information, and what rights can they extert through that key ?
Oh you bet, the few faltering first steps on that path are the core of our business model. Tad Williams Otherland series is more likely a "better vision" than the Matrix tho, in terms of the nature of the beast. The matrix plot assumes some kind of giant hierarchy, in loose terms, whereas real functioning networks, are, well, they're loose. In short, not even machines could muster the level of agreement to manifest a single complex thing such as the matrix. We must build partitioned worlds before we can build a seamless whole, and hell, some still think VRML is a good idea. So we do have a ways to go.
At a certain level of system evolution, my enthusiasm does begin to drift towards trepidation. My personal experiance with VR systems is that they can be quite "reality bending", especially when they are deliberately matched to the real world. That is to say, they perform a deliberate assault on your ability to detect the difference between synthesized virtual information and real physical information. Admittedly our first steps are still pretty crude, on the other hand, we can generate really spectacular misfires. My personal favorite are immersive VR goggles. Fly someone right off the deck at 900Kph, and make a few 90 degree left and right turns at that speed. With their brain seeing a view that tells them one thing, their innear ear saying something completely different, most participants turn a delightful shade of green about thirty seconds into the process. This is a brutal example of a usually more subtly occuring phenomenon, the fact that users already delightfully muddle up the differences between reality and the infospace whenever the frame rate is high enough. That's the entire reason for all those pretty pictures from supercomputers, as well as the basis of companies like id.
Of course I want people to get into VR, I have spent a long time trying to help make that possible. I don't think we can actually "build" a virtual space like we built the moon rockets, because the nature of virtual environments and the systems that form them is fundamentally much messier, as it's a dynamically evolving system. I've also learned to have a healthy respect for the interaction of another dimension in my user interface and my visual cortex. People who have forgotten this have already lost billions on wall street, by basing their actions on ephemera in virtual worlds that they thought they saw. The field seems to exert a siren song over us all, and I think the interaction of the technology and us deserves healthy respect. There is something to be said for the aquisition of powers to reshape reality, virtual or otherwise.
Umm, having worked in the guts of VR systems for too many years now I just have to cough up 2c. I am a geek, I liked the matrix, but my nongeek friends all like the matrix too. Can't a movie just be phun?! Now, onward into deconstruction.
:-)
There is no question any complex simulation has certain assumptions hardwired into the fabric of the system. In fact, it is a process that made Kurt Godel famous, basically the formalization of the fact that if you create any formal expression system (the matrix) the very system contains elements which cannot be expressed in it. Hence phone booths are ok, cause this may be the way the fundamental system is wired. The system itself is incapable of adjusting or altering them because it cannot "express" them.
Agents and stuff are simple, one can make the argument that since they don't exist at all, this is simply the representation "you" are seeing. In this world, 10 people can look at the "same thing" and it really is 10 different nonthings.
OTOH, beating such a system _cannot_ be an issue of "I've got more firepower than you in the shared concensus", i.e. the gunfight in the lobby. One interesting thing about a lot of VR simulations is that they have to be slowed down for the human observer. In short, in many cases you beat the computer in twitch games cause the developers deliberately waste enough time to give you a chance. We've got a demo for our system that flies you through a nasty corkscrew under the computers control. Under human control, nobody has yet made it through the second turn, and there are 32 such turns at speed.
Hence, the way you beat such a system is by co-opting it against itself, by reexpressing the paradigm in such a way that you have the advantage. You've got one to start with, the machine is stuck there, it can't leave. The flomo bullet sequences are a good example. I do believe that this is the true meaning of the verb "to hack".
But the first floor shootout was just stupid, imho. What would I have done is to say, OK if you know it's a phantom, shape the phantom. The effect I'd pick would be the Xaos bubble derez sequence as seen in lawnmower man 1, I'd do it at an order of magnitude higher resolution, and I'd just derez the first floor. 10 to 1 the core logic fabric then drops the second floor right into place. All you have to do is not be crushed
I can't buy the extrapolation about the growth of code for blocking and suppressing things on the net is a free speech issue. The fundamental argument/premise is that spam and any other form of wanton broadcast is at it's heart the complete inverse of the dataflow of the net. Specifically, the net is about the ability of an individual to reach out for what they want, TV is for throwing things at them.
I would argue that spam list killers and the like are not infringments on free speech because of the fact that they are all voluntary optins. If your provider opted you in and you don't like it, pick another provider!
On the net, an infringment of free speech comes from the readers perspective, not the publishers. China infringes the free speech of their citizens (stop laughing) by denying them the ability to reach out to the information they want. They are prevented from accessing information published and waiting for them.
Summarizing, the net is a pull medium, it functions by responding dynamically to the demands of those using it. Antispam issues are basically issues of a public nuisiance nature, as it is so easy for any net user to decide for themselves what they do and don't want. Access to sites is an issue of free speech. The ability to make a site can even be called an issue of free speech (scary), but filtering unsolicited data has nothing to do with it based on the fundamental nature of net communications
You can get through the real world by being oblivious to Euclid, trig, and calculus. But that doesn't mean you don't need or use them all the time. Think of this the next time you drive through a turn on the freeway in your car. A minor nit, but I just couldn't help myself.
Math or programming, it's all really the same thing. Hence, you can make programs out of math and math is the foundation of all programming. It's all just formalized symbolism to us chickens.
We certainly aren't fans of Microsoft's stated intentions to provide only DX8/Fahrenheight as a supported 3D library. Already WGL is seeming an orphan child. It's weird, but right now John Carmack is worth a lot to my business, cause he's holding to openGL.
IMHO this is a monumentally stupid blunder on Microsofts part. First, if they are aiming this at the real-time games market and the speed over quality requirements of that market and can't even get Carmack to bite,......HELLO
Secondly, they are in effect forcing companies like mine aiming for the future when high end interactive visualization goes onto desktops to make plans for ports to other systems, for the simple reason they've now tarted their platform up too much for the game developers.
Finally, this effort of theirs is helping the mass market for 3D games software, but it hurts everything else. Already, companies like Dell, Micron, and others have high markup workstation lines. Wintel workstations are usually treated as a high margin item. Gee, guys --- wouldn't it be better if we drove these systems into the mass market at the mass market price points. Exactly who gets to say, OK thats enough power, you'll have to pay premium beyond this point ? I understand supply and demand, I argue that across the board demand for intellegently applied processor power is still a big sea.
So maybe we shouldn't make a peep, just be quiet and smile. When Linux and its friends are the first to provide workstation level features and performance and mass market interfaces (GUIS,SUIS) at a mass market price point....Game over. Microsoft looses. Because then the development equation becomes one of "well, it would cost us X to make this on Linux and 10x to do it on the Windows Game Playplatform" So lets do it on Linux first. Cool for Linux, pathetically stupid on Microsofts part.
Very good. I agree wholeheartedly. I don't think that DX is there yet, as I agree with the fundamentals of John Carmacks arguments. I violently object to Microsofts plan to try and make DX8/Fahrenheight the service for all users. The best world is separate apis for games (speed counts uber alles) and openGL (speed counts, but not at the expense of trashing the visuals). Probably is a common ground between the two at the hardware interface/bit blasting level, but after that they serve different masters
The standard wisdom in the CG community is that feature movie work is pretty rigid and demanding, and there is little room for creative flexibility on the CG side of the house. The result of this is a steady exodus of CG people from the feature movie side into advertising, where there is more creativity and ingenuity required of them. One comment in particular I remember was from an ILM artist who left after the completion of JP because he had just spent the last 18 months of his life making a dinousaurs butt wiggle realistically and wanted something with more freedom of expression. That said, the cast of technical persona at ILM contains some of the finest CG talent, tech and art. The only real gaffe I can't forget was in the nasty bluescreen bleed/scaling errors that was in the shot oof the plane dropping off the blimp in the last of the IJ movies.
Oh yeah, I forgot. Mea culpa slashdot. We are interested in other platforms than Windows. Gee, if you use that light streaming through the window trick, whatever will you do on linux/mesa, mac/opengl, sgi/(open)gl, etc ?
Hmmm. Our system is being used by a game company to try out a new idea, but it's also already the basis of a commercial app that does surgical training simulation for Drs. One of the main reasons we use openGL is that we can do things "correctly", i.e. the patient model lighting we provide is basically a crude approximation of what the Doc will see when they stick a needle in a real patient, albeit real patients aren't see through.
John Carmack at id uses openGL to do things we wouldn't dare, such as the lightmaps, as his needs are different than ours, and nobody is going to put dead demons or players on the operating room table. But openGL satisfies us both.
DX on the other hand has many of these "clever tricks" to do things like streaming light through windows, but places much less thought and emphasis on the general issues of rendering realistic images for a wide variety of tasks. Do you want to be operated on by a Doc who got any of his training on something optimized to do gore and explosions ? And even in the submarket they court, there are some knowledgable people, e.g. Carmack, who still tell them to stuff it.
What amazes me about the DX product are two specific issues.
1) The mantra at Microsoft is framerate,framerate,framerate. Good mantra. Unfortunately, they feel this justifies them in adding internal logic to deliberately degrade the geometries they have been given in order to maintain framerate. Thats the deal in their new focus on curves, to maintain framerate, __not to increase visual quality__. Lessee, we have apps for Docs, soldiers, and others who really really really don't need this. And I'm sure id isn't thrilled about having many ducats invested in their artwork, and loosing control over the tradeoffs on responsiveness against visual quality. Hey, we're supposed to know what we are doing. If msft wants to sell tutorial aids, thats fine, but some of us carry high development expenses specifically because we wish to make those decisions ourselves. Hence we tend to like openGL.
2) MSFT went out and hired Jim Blinn (making obligatory obiesiance now to the great Blinn), and put him in their advanced R&D area (I think). Last summer he talks about the 10 things still undone in CG, and delivers a pretty specific complaint about clever but narrow and unstable tricks used to improve visual realism in a narrow range of applications. He even gently points the finger at 3D games. Does Microsoft seem to pay any attention to him. No! I personally find that disgusting.
Summarizing, I guess I'm saying that as long as msft puts their efforts into stupid optimizations like this, to paraphrase the cryptogeeks, you can have my openGL libs when you pry them from my cold dead mind. It does seem some daze that msft is willing to go to that length, WGL is apparently doomed now.
DX6 to openGL translation is certainly not "impossible", I think scitex has a product out/coming called DirectGL. There are a number of openGL to DX6 translators out there, but this seems to be mostly a msft attempt to bring certain strategic apps into the DX fold, even if they really don't want to be there.
My own take is that msft has put too many screwball optimizations (used advisedly) into DX6 to satisfy gamers that make the conversion potentially difficult, and certainly inefficient. Why invest all the time and energy into an app to get performance and visual quality, and then pass all your work through a translator? I think this is the central reason that there are such battles over graphics libraries, because efficient use of the library really requires an exclusive commitment to that library. And even in the game development community there is a lot of dissention, our fave is John Carmack, who has given the entire 3D efforts at msft some very good critiques from both technical and business perspectives.
Bottom line is that computer graphics will remain a crude approximation for years to come, as reality is unfortunately infinitely complex and computers aren't. This kind of squelches translators for anyone trying to push the envelope.
Ummm, do you mean OpenGL OR Direct3D ? Or are you implying there is no difference? The first product out with our VR system is a training tool for surgeons. It is meant to run on pcs, but we certainly aren't interested in a tech (D3D) where the only input msft takes is from established game developers. The Dr who developed the app on top of our engine is horrified by the prospect of having to switch to D3D/DX8 if msft succeeds in eradicating openGL support.
Basically, D3D is for game developers who are not of John Carmacks level. openGL is for anyone who is really serious about the images they put on the screen (and yes, I do mean realtime images)
Being in the VR biz, it's a subject that gets thought of from time to time. A few observations
.333 error bar on that number. So there is a finite and limited set of "major approved web destinations" for each user, group of users, culture of users, pool of all users. Hence as this becomes saturated, to enter that zone is to need to bump someone else out. Happily, the net is very dynamic, so this does happen on a pretty regular basis, but I would call the sets of web sites held in memory from individuals to the collective web community to be the most limited real estate on the web. I have found slashdot. I like slashdot. Someone I used to like is going to have to get bumped down to "the things I have references to but never directly remember"
1) OK, the math geeks have weighed in, and I only solve big polys, I'm not a math geek. Comments on transfinite numbers are an effective technical containment for this conversation, so I accept those as givens.
2) The real interesting bit is the historical meaning of real estate (this is the web, go to altavista or wherever). A fundamental truth is that it is completely based on a finite nonrenewable (yes recirculating) supply. Web space is not infinite, but at the same time it is not finite the way physical real estate is finite. Hence when the web has pricing metaphors for space that mimic the real estate model things get interesting, theoretically, and very lucrative, commercially. It's the conflict between extensibility (the web) and non-extensibility (real estate). I think that the real real estate (hmm, real^2 estate?) will be real interesting when we have real companies like bechtel making new real estate, e.g. undersea living. Yes I agree this should be done in an intellegent fashion, I do have a son and therefore do care about issues that extend beyond my lifetime.
3) As far as infinite real estate for cool new sites, don't I wish. It would theoretically make anyones chance of site success much bigger. I think the deal is that as individuals or as cultures, we have only so many balls we can keep in the air at once. It's that old saw that people can keep seven things in their head at once with a