|
pixelshow 2007 invitro by The Lab [web]
[nfo]
|
||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
|
|
|||||||||
|
popularity : 60% |
|||||||||
alltime top: #11696 |
|
|||||||||
added on the 2007-05-02 00:57:41 by Nuclear  |
||||||||||
popularity helper
comments
They brought ducks!
We have a winner! collect your nut-kick at the pixelshow partyplace, or buy out my mercy with a sixpack of beer :)
The answer is simple: cause it's more fun (and much easier too) to do it that way, and there isn't any reason to avoid using shaders anymore as everyone has a ps2.0 capable gfxcard. And those who don't, aren't really expecting to be able to watch any demos.
The answer is simple: cause it's more fun (and much easier too) to do it that way, and there isn't any reason to avoid using shaders anymore as everyone has a ps2.0 capable gfxcard. And those who don't, aren't really expecting to be able to watch any demos.
Is there a way to watch the linux version on a mac?
Well, i found the colors abit weird...
But on the other hand, the music is very nice and funky, its got duck.3ds, it ran effortlessly on my x86_64 debian and it's GPL :-)
See you all at the partyplace :-)
But on the other hand, the music is very nice and funky, its got duck.3ds, it ran effortlessly on my x86_64 debian and it's GPL :-)
See you all at the partyplace :-)
sybex: assuming that you mean on MacOSX (cause you can have linux on a mac too, although I wouldn't bet on the availability of linux-ppc gfx drivers) it *should* be able to compile on OSX easily, with few modifications in the makefile. Talk to me through email if you would like to help produce an OSX binary.
Duck!!
It could work with a better font ;D
why?
ok-ish
Nice. Headbanging duckz!!! =)
nice ducks :)
Does the trick, a few comments:
The first image is probably the most wtf fugly thing I've seen in years. Just count: 4 fonts (2 of them being pseudo-3D), fake 'shadowing', layers with different amount of 'antialiasing', ms-paint horrible spray effect with FF00FF color...
Then the logo of pixelshow: Somebody took the original, minimalist B&W logo designed by amoivikos and ass-raped it by adding splashes of 'cool' FF00000 gritty textures and then bilinear-filtering enlarge it to make it fit. I bet you didn't even ask for his permission or approval for this crime of style.
Anyway, back to that demo: The ducks dancing are nice albeit very slow (like sub 10fps on a 9550) due to pixel shader nightmares involving several normalize, pow etc.etc. (cell-shading ofcourse can be done with plain-vanilla opengl or simple shader operations which would give you 100fps+ for a simple scene like this).
There is a bug with the text going up - there is some sort of step and instances where you see some horizontal lines on top of letters.
Anyway, thumbs up carrying the message across.
The first image is probably the most wtf fugly thing I've seen in years. Just count: 4 fonts (2 of them being pseudo-3D), fake 'shadowing', layers with different amount of 'antialiasing', ms-paint horrible spray effect with FF00FF color...
Then the logo of pixelshow: Somebody took the original, minimalist B&W logo designed by amoivikos and ass-raped it by adding splashes of 'cool' FF00000 gritty textures and then bilinear-filtering enlarge it to make it fit. I bet you didn't even ask for his permission or approval for this crime of style.
Anyway, back to that demo: The ducks dancing are nice albeit very slow (like sub 10fps on a 9550) due to pixel shader nightmares involving several normalize, pow etc.etc. (cell-shading ofcourse can be done with plain-vanilla opengl or simple shader operations which would give you 100fps+ for a simple scene like this).
There is a bug with the text going up - there is some sort of step and instances where you see some horizontal lines on top of letters.
Anyway, thumbs up carrying the message across.
Thank you all for your comments.
Navis:
About the loading screen, I totally agree.
About the pixelshow logo, two things.
A) I didn't do it (see previous point about not being an artist), just used the one from the pixelshow web site.
B) I think it's very nice. I would choose different colors definitely but even with those colors I think it looks cool. Minimalism is good too, but it's not the only valid design paradigm by a long shot.
Back to the demo: the ducks are VERY fast, try adding like 100 of them with the not-so-secret -d switch from the command line. you won't notice much slowdown. Of course everything in this invitro (almost) can be done without shaders, but no, the shaders are not affecting the performance. There is some sort of driver bug related with the glitch you mention on some ATI cards that slows everything down. I used an X1900 for ATI debugging and it looks/performs just fine, try updating the drivers.
Finally about the pixel shaders... what the *FUCK* are you talking about?
There are 4 normalizations there, they are all absolutely necessary. The light direction vector (light - pos) *must* be unit length, the normal *must* be renormalized after interpolation, the view vector (-vpos) *must* be unit length, and the half-angle vector *must* be normalized as well. If you run the thing through fixed-function with glLightModeli(GL_LIGHT_MODEL_LOCAL_VIEWER, 1), and a point light it has to do the exact same operations. Also you *know* that the dot product is raised to the specular exponent that's how the blinn reflectance model works. Are you being provocative just for the hell of it? If so fuck off.
Navis:
About the loading screen, I totally agree.
About the pixelshow logo, two things.
A) I didn't do it (see previous point about not being an artist), just used the one from the pixelshow web site.
B) I think it's very nice. I would choose different colors definitely but even with those colors I think it looks cool. Minimalism is good too, but it's not the only valid design paradigm by a long shot.
Back to the demo: the ducks are VERY fast, try adding like 100 of them with the not-so-secret -d switch from the command line. you won't notice much slowdown. Of course everything in this invitro (almost) can be done without shaders, but no, the shaders are not affecting the performance. There is some sort of driver bug related with the glitch you mention on some ATI cards that slows everything down. I used an X1900 for ATI debugging and it looks/performs just fine, try updating the drivers.
Finally about the pixel shaders... what the *FUCK* are you talking about?
There are 4 normalizations there, they are all absolutely necessary. The light direction vector (light - pos) *must* be unit length, the normal *must* be renormalized after interpolation, the view vector (-vpos) *must* be unit length, and the half-angle vector *must* be normalized as well. If you run the thing through fixed-function with glLightModeli(GL_LIGHT_MODEL_LOCAL_VIEWER, 1), and a point light it has to do the exact same operations. Also you *know* that the dot product is raised to the specular exponent that's how the blinn reflectance model works. Are you being provocative just for the hell of it? If so fuck off.
Nice prod. I don't like that it uses pixelshaders when the same stuff could be done without them, but then again i know there were some time limitation so it's ok. At the end, the result is what matters.
@Navis:
I think that you have something wrong with your card (didn't you used to have a GeForce btw?). We ran the demo with even 1000 ducks without any performance penalty. The only part where having 1000 ducks had a penalty was in the wireframe thing.
However note that Radeon 9550 is an old and slow card (R300 generation - actually RV350), which is just a little faster than a GeForceFX 5200 and the slower card from the whole R300 generation (according to information found at Wikipedia at least - i don't own these cards). The invitation is fully PS2.0 (the animating gradients and the text is all calculated in a pixel shader i think). So the slow speed is possibly because of that.
In any case, in any card i tried, including my own GeForce FX 5950 Ultra, it runs smoothly and flawlessly (even with 100 ducks - my FX can't cope with 1000 :-D).
@Navis:
I think that you have something wrong with your card (didn't you used to have a GeForce btw?). We ran the demo with even 1000 ducks without any performance penalty. The only part where having 1000 ducks had a penalty was in the wireframe thing.
However note that Radeon 9550 is an old and slow card (R300 generation - actually RV350), which is just a little faster than a GeForceFX 5200 and the slower card from the whole R300 generation (according to information found at Wikipedia at least - i don't own these cards). The invitation is fully PS2.0 (the animating gradients and the text is all calculated in a pixel shader i think). So the slow speed is possibly because of that.
In any case, in any card i tried, including my own GeForce FX 5950 Ultra, it runs smoothly and flawlessly (even with 100 ducks - my FX can't cope with 1000 :-D).
Party on.
/me likes
geez, you can't make a comment these days.. anyway,
I don't know how or why but while 'The little prince' by kolor runs at solid framerate (a 2001 demo mind), this invitation runs at like 10fps (both demos feature cell shading, also little prince has a few more polygons per scene :-)). If you can explain that...
My fair bet is again the shaders, namely obj.p.glsl which has all these operations and still manages to look like something you could do with a 1D texture look-up or some sort of clever environmental shading using opengl 1.0. I mean you even normalize a vec3 coming directly from the vertex shader - wouldn't it be better to normalize before passing it through ?? Then you have effectively 1D texture look-ups using 2D textures "texture2D(tex, vec2(d, 0.55))". Wastefull. Just a though :-)
I don't know how or why but while 'The little prince' by kolor runs at solid framerate (a 2001 demo mind), this invitation runs at like 10fps (both demos feature cell shading, also little prince has a few more polygons per scene :-)). If you can explain that...
My fair bet is again the shaders, namely obj.p.glsl which has all these operations and still manages to look like something you could do with a 1D texture look-up or some sort of clever environmental shading using opengl 1.0. I mean you even normalize a vec3 coming directly from the vertex shader - wouldn't it be better to normalize before passing it through ?? Then you have effectively 1D texture look-ups using 2D textures "texture2D(tex, vec2(d, 0.55))". Wastefull. Just a though :-)
No you can't normalize them before passing them through. You don't just "pass through" stuff to the pixel shader, it gets "interpolated". And when you interpolate between two unit vectors, the result is *not* a unit vector.
The 2D texture contains multiple "step" scanlines, by adjusting the vertical coodinate I can achieve different degrees of quantization, ergo it's much easier to tinker with than a fixed 1D texture with one quantization level.
The 2D texture contains multiple "step" scanlines, by adjusting the vertical coodinate I can achieve different degrees of quantization, ergo it's much easier to tinker with than a fixed 1D texture with one quantization level.
pixelshader sucks... it's don't running on my laptop :(
Today i think it's more "your laptop sucks, it doesn't have pixelshaders" :-P
for a well tesselated object the normal should be very close to being a unit, even with linear interpolation, giving you almost the same visual results.
All that doesn't change the fact that you have a 10 line shader to do something that pre 2002 demos did much faster in plain opengl.
Well, this demo doesn't feature a "well tesselated" object, but a low-res duck.3ds object. I tried to remove the normalize call from the shader, and the highlights looked like crap (well, they were flying all around :-P). I prefer the highlighted version.
In any case, if you so want to see it without pixelshaders and you believe that it can be done so fast without hassles, the source code is there, get it, modify it and post it there. I'm sure that Nuclear will include your modified source in his product and you'll make every 9550 owner a proud demoscener. If they care, that is.
In any case, if you so want to see it without pixelshaders and you believe that it can be done so fast without hassles, the source code is there, get it, modify it and post it there. I'm sure that Nuclear will include your modified source in his product and you'll make every 9550 owner a proud demoscener. If they care, that is.
We're in 2007, who cares if you use some performance-wise inefficient techniques to increase productivity?
talk about primadonas who cant take a critic... geez
Critique which is based on fact and reason is very welcome. Critique which is content-free, unfounded and spurious, and speaking out of ones ass just to make a fuss will not be accepted.
yeah it must be an unfound conspiracy that a 1996 effect runs like 50 times slower than it should be...
Naaah...
oh noes broken link :(
screenshot vote
Quote:
Critique which is based on fact and reason is very welcome. Critique which is content-free, unfounded and spurious, and speaking out of ones ass just to make a fuss will not be accepted.
ahahaha
Quote:
talk about primadonas who cant take a critic... geez
suddenly makes a lot of sense :D
Holy shit I forgot how ludicrous the comments on this were. What a bunch of retards... lol @ "screenshot vote", made my day. Get a clue you miserable idiots :)
Btw this rules. Die horribly.
Btw this rules. Die horribly.
.
submit changes
if this prod is a fake, some info is false or the download link is broken,
do not post about it in the comments, it will get lost.
instead, click here !
Well for the "why?" question from the nfo... I really would like to know anyway :)