a shader each day

[Jason]

Another Day

I had the chance to speak as an opening act for "Oldschool Blog Live!" at Siggraph2008.

Well done and congratulations on your presentation! I'm sorry we didn't get a chance to meet up at SIGGRAPH..

This thread is very cool, by the way. Keep it coming:)

[rdg]

Day 10: Lorem Ipsum was my Cradle Song

Slowly occluded toNDC() bricks with a hint of rotate().

Surprisingly bricksVOP got it's own lighting model, which I had to bypass to make it working without lights.

Most of the luminance() calls can be disregarded by working with floats instead of colors - I was just to lazy to substitute the brickVOP with something more efficient.

loremIspumwasmyLullaby.hip

[rdg]

Day 11: XOR

hm ... not even sure if this is XOR but let's pretend it is ...

A shader that emulates basic/java/flash boolean diplay writing.

eetu commented:

"(12:25:28 AM) eetu: yea, tracing sounds a tad overkill "

Well, but how to cookie this?

I first felt that this is the time for gather() ... but ...

Reading some of Mario's posts I guess a gather needs a rayimport and both can be in the same shader ...

Anyway a trace() looked quite right once I transformed my direction vector into space:current.

But now:

The same struction from above.

Some raytracing limit? Just right as eetu suggested?

I guessed that I need to limit my rays to exactly one hit, but this wan't work probably as the shading works top down.

XOR works bottom up.

Well, this is 3D and I'm not limited to Y down cameras ...

This is a shaded XOR and DOF:

this is the base geometry:

My shaders still look better without lighting model ...

...

rdg_XOR.hip

[Andz]

Is RDG a poet or a philosofer?

Edited October 4, 2008 by Andz

[rdg]

Day 12: RDG vs Easing Functions

This not a shader.

It's my first canny steps into the HDK.

I 'implemented' some easing functions as described e.g. at actionscript.org[1]

Not yet the standart Robert Penner Easing Functions [2] - most likely caused by my lack of c/c++ knowledge.

Thanks to Chip Collier and Mark Elendt I was able to have a function that takes more than two arguments

I compiled these on openSuse 10.3 64bit - I guess for other plattforms you have recompile them.

There is also a exhelp file, but how this might be merged to other existent files is a mystery to me.

It works fine, as I don't have any other exhelp files in $HOME/houdini9.5/help

Adds

* rdgeasein()

* rdgeaseout()

* rdgeaseexp()

* rdgeasecirc()

* rdgeasesin()

see exhelp rdg for details

rdgEasingFunctions.tar.gz

[1] http://www.actionscript.org/resources/arti...ning/Page3.html

[2] http://www.robertpenner.com/easing/

Edited October 3, 2008 by rdg

[Mario Marengo]

Day 12: RDG vs Easing Functions

Hey rdg,

Looking good, but how are you implementing your ease function?

I haven't looked at your references, but AFAIK, the whole point of an ease() function (as an interpolation, not a spline) is that the slope at both ends is zero (a cubic spline with dydx=0 at the ends). I don't see that in your image (but maybe there aren't enough divs to show it?)... just curious.

[rdg]

Hey rdg,

Looking good, but how are you implementing your ease function?

I haven't looked at your references, but AFAIK, the whole point of an ease() function (as an interpolation, not a spline) is that the slope at both ends is zero (a cubic spline with dydx=0 at the ends). I don't see that in your image (but maybe there aren't enough divs to show it?)... just curious.

ohm ... ahm ... well

What?

I guess easing might be the wrong term then.

The functions (easein, easeout, easeexp, easecirc, easesin) interpolate between 0...1.

E.g. :

fn easecirc(v) = -sqrt(1-v*v)-1;

Can the slope in an easein be zero at the end?

Edited October 3, 2008 by rdg

[Mario Marengo]

ohm ... ahm ... well

What?

For those not on IRC...

I'm aware that there are an infinite number of ways to interpolate two values, and all these ways are valid. So there's nothing "wrong" with your function (I never said there was). However, it doesn't resemble Houdini's "ease" function, and I thought that was perhaps what you were striving to get.

Ease:

3 x² - 2 x³

And ease with "bias" or "power ease":

3 x^2*bias - 2 x^3*bias

And the easep() family (second function above) looks like this:

[rdg]

However, it doesn't resemble Houdini's "ease" function, and I thought that was perhaps what you were striving to get.

I did't want to resemble Houdini'S ease, which I never got to work in SOPs btw.

The functions I compiled are more like shortcuts to different interpolation methods.

[Mario Marengo]

fn easecirc(v) = -sqrt(1-v*v)-1;

Just doodling here...

Seeing as that function there is a 90 degree arc of a circle sqrt(1-(1-x)²), it occurred to me that there's a natural way to parameterize it...

Here's a "power" version of your "easecirc()" function:

easecircp(x) (always assuming x in [0,1], and bias>0):

pow(1-pow(1-x,bias), 1/bias)

This one is linear at bias=1, circle-arc (like yours) at bias=2, and inverse arc at bias=1/2.

One more for your collection

[rdg]

Day 13: Not implemented in Code

Yeah, it's Lorenz time again.

As the easing functions caused more confusion than good - apart from the missing discussion "expressions vs python: why the hassle anyway?" - I decided to take the next step.

The result is this LorenzAttractorSOP. Instead of mofifying incomming geometry like my pythonSOP [1] it generates the attractor on the fly.

10000000 iterations are easily generated - but they don't render ...

Even 999999 throw a DSO_ERROR every now and then.

But this is not a r3nd3rP0rn ... we want beautifikation.

[1] http://www.preset.de/2007/0711/lorenz/

[jacob clark]

Nice!

I was curious though. Have you had a look at the geo.createPoint() function?

You can actually build your entire geometry in the the Python SOP itself. I think that the poly build function is available now too, as to connect the points?

Keep it up!

[jacob clark]

Actually, I have a python SOP of similar nature that might give you some ideas to try... Heh, if nothing else, you can yell at me for my Python implementation!

It's a graph of some random Calculus function I found on Wolfram Mathmatics site.

-j

SOP_fs_RiemannFuntionA.otl

SOP_fs_RiemannFuntionA.otl

Edited October 4, 2008 by jacob clark

[rdg]

Day 14: Diffusion-Limited Aggregation

Thanks Jacob, if I ever knew about geo.createPoint() I forgot about it.

The tasks I tried to solve using the pythonSOP somehow shipwrecked most of the time on Houdini's SIMD/SMID/whatever approach. Creating new points instead of modifying existing ones is a good way to work around this limitation. But it often also requires a lot of dictionaries and structures to keep track of data and states.

I appreciate the pythonSOP for being a rapid prototype tool, but it's all about membership [1], isn't it?

Well, further experiments with a Clifford Attractor [2] and the HDK showed that it's currently been "Fortunes favours fools".

Not knowing enough about c++ arrays I had to strech this 2D structure into the third dimension: time evolves to space ... you get the point.

But his also allows me to work without Lyapunov exponents - the rate a system changes over time. No, not really, the exponent tells if a system is stable or chaotic.

Usually this value is also used to colourise the images. I accumulate the colour using an additive opacity shader.

The subtext is probably there is a lot more to explore in a vanilla Houdini before reaching out for the dark stars twinkling out there.

Ihab pointed me again to DLA - Diffusion-Limited Aggregation[3] - and we discussed if the HDK might be the way to get this running.

This being definitely out of reach for my c++ "skills", I got to resort in a POPnetwork.

Sure a pythonSOP might help here, but in my opinion this is close to creating the simulation in an external program and importing it into Houdini.

The image shows an aggregated structure around a star shaped core after 15000 steps of simulation. I didn't remove the diffusing particles.

Here is a render of every 25th frame of this process: DLA Seq 3

As pagefan already pointed out: Nature's quite picky when it comes to parameters, And indeed it's a trial and select process finding the values that allow the particles to react. The obviously most dictating parameter is the diffusion algorithm, which I fake currently with a random walk. The angle a particle can go astray makes the difference. Depending on the parameters there are to many missing particles. Or they just accumulate to a big blob without growing the typical tendrils.

To optimize the process the system needs know at which rate the structure grows. This allows the diffusion particles to be born in a zone around the structure.

I am quite ego I found a way to solve this within POPs - without any fancy SOPsolvers and/or pushing geometry in/out of POPs.

It's been definitely a geek-y weekend here, but it was raining, if this is an excuse.

[1] Justice - DVNO: http://www.youtube.com/watch?v=0qPtx819idI

[2] Bourke, Paul, Clifford Attractors - http://local.wasp.uwa.edu.au/~pbourke/fractals/clifford/

[3] Wirtz, Franz-Josef, Diffusion-Limited Aggregation and its Simulation: http://www.gut-wirtz.de/dla/

Edited October 8, 2008 by rdg

[edward]

Have you seen Andy Lomas' DLA pics? They're pretty cool ... from SIGGRAPH 2005:

http://www.andylomas.com/aggregationImages.html

(be sure to check out the larger 4K pic)

[rdg]

don't panic!

No, thanks for the link!

Very inspiring.

simulating the paths of millions of particles

This makes me wish again learning an auxiliary language. Not sure if an application like Houdini is even the right tool for such tasks.

But as Andy mentions Haeckel:

I cannot deny being a great fan of Haeckel myself.

This is the link to one of his books as on-line facsimile:

Ernst Haeckel:

Kunstformen der Natur

1899-1904

[stevenong]

Hey Georg,

I'm speechless but I'll tell you I'm using the Clifford Attractor image as my desktop image now.

Very very nice images.

Cheers!

steven

PS: May I have a 1600x1200 png of the Clifford Attractor render please?

[rdg]

[...] I'm using the Clifford Attractor image as my desktop image now. smile.gif [...]

PS: May I have a 1600x1200 png of the Clifford Attractor render please?

Thank you very much!

[rdg]

Day 15: SSS - SkypeSpriteShader

This one is a shader.

It uses the skype emoticon resource file and maps the anim of a random emoticon to a sprite.

So much for theory.

The file is a grid of 1500 animation frames and of course I lost track of the sequences sooner than later.

That's why the emoticons are popping in this sample animation

Here's another animation on vimeo: http://vimeo.com/2035099

Well, I used a VEX array and it's not a VOP VEX Surface SHOP!

rdg_SSSskypeSpriteShader.tar.gz

Of course the emoticons are intellectual property of Skype.com and I respect this fact.

Edited October 22, 2008 by rdg

[aracid]

wow, this is cool

thanks for sharing