Raymarching Toolbox Thread
category: code [glöplog]
oh really
STOP IT NOW. Behave. ;)
*puts away manger joke and walks away*
Really!
Sorry, couldn't resist. :))
Sorry, couldn't resist. :))
Fold and cut polyhedra:
Code:
float DE(vec3 z)
{
// Fold:
//This is a folding set with dodecahedral symmetry
z = abs(z);
z-=2.0 * max(0.0, dot(z, n1)) * n1;//Thanks AndyAlias for the optimization
z-=2.0 * max(0.0, dot(z, n2)) * n2;
z = abs(z);
z-=2.0 * max(0.0, dot(z, n1)) * n1;
z-=2.0 * max(0.0, dot(z, n2)) * n2;
z = abs(z);
z-=2.0 * max(0.0, dot(z, n1)) * n1;
z-=2.0 * max(0.0, dot(z, n2)) * n2;
//Cut:
//Distance to the plane going through vec3(Size,0.,0.) and which normal is plnormal (must be normalized)
//You can also use curved and/or multiple cuts
return dot(z-vec3(Size,0.,0.),plnormal);
} any screenshot ?
Thanks a lot for all your DE work knighty, i don't know how we could visualize all these fractals without you. (: And the first one in your link reminds surprisingly of a recent 4k intro too. ;)
I like the last examples , especially the subblue link.
Sorry for emphasize obvious facts.
Sorry for emphasize obvious facts.
nystep: ;o)
iq: About "cantor fracals", there is also this thread where I've given the method to get the distance estimate. I'm not surprised that it was already used in demoscene but it's a pitty such techniques are keps secret for so long time. That said I'm still wondering how the menger sponge was rendered in stargazer.
tigrou: It's cool you found the same technique as KIFS to render the Menger sponge. :)
You can of course experiment with other folding planes.
Maybe this is a little bit out of subject but It's related to it and I thought you may find this of some interrest: using foldings to generate triangle group tilings.
here is what I did so far:
evaldraw script
fragmentarium scripts
here is where that idea came from:
"kleinian drops" boxplorer scripts. There is an hyperbolic tesselation hiding in one of the examples. ;)
video by marius.
iq: About "cantor fracals", there is also this thread where I've given the method to get the distance estimate. I'm not surprised that it was already used in demoscene but it's a pitty such techniques are keps secret for so long time. That said I'm still wondering how the menger sponge was rendered in stargazer.
tigrou: It's cool you found the same technique as KIFS to render the Menger sponge. :)
You can of course experiment with other folding planes.
Maybe this is a little bit out of subject but It's related to it and I thought you may find this of some interrest: using foldings to generate triangle group tilings.
here is what I did so far:
evaldraw script
fragmentarium scripts
here is where that idea came from:
"kleinian drops" boxplorer scripts. There is an hyperbolic tesselation hiding in one of the examples. ;)
video by marius.
geometry:
Code:
Octahedron:
float octahedron(float3 p, float size)
{
p=abs(p);
return (p.x+p.y+p.z-size)/3;
}
Rounded Cubes:
float rcube(float3 p,float3 box,float rad )
{
return length(max(abs(p)-box+rad,0.0))-rad;
}
Hexagon:
float hexagonX(float3 p, float x,float y )
{
p=abs(p);
return max(p.x-y,max(p.y+p.z*.5,p.z)-x);
}
Rings:
float ring(float3 p,float r,float r2,float c)
{
return max(abs(length(p.xz)-r)-r2,abs(p.y)-c);
}
the hexagon is just on x-plane, also the dividing is not correct, but smaller! (for 4k-usage) ;)
Does anyone have a copy of zeno.pdf that they can upload? I can't seem to find it anywhere and the links posted earlier are both dead.
I also want to mention that the "Hexagonal Prism - unsigned" isn't really symmetric.
IQ's dist functions
should maybe be (coeffs are sin/cos of 60°)
IQ's dist functions
Code:
q.x+q.y*0.57735
should maybe be (coeffs are sin/cos of 60°)
Code:
q.x*0.866025 + q.y*0.5
I did some RM stuff a while back and stumbled onto the scale repeater thing, I'm sure u guys have done it but just in case:
rad=4*floor(mod(length(p),128)-64)*(mod(normalize(p),0.1));
dist=Box(rad,0.5);
I'm not sure how it works - since it was a pure accident for me - I was trying to do something else.
this was the start of the math fumbling
http://www.youtube.com/watch?v=JuVtAwqRWL0
and ended up with this
http://www.youtube.com/watch?v=f7ATokNn3RM
never had so much fun before :P
rad=4*floor(mod(length(p),128)-64)*(mod(normalize(p),0.1));
dist=Box(rad,0.5);
I'm not sure how it works - since it was a pure accident for me - I was trying to do something else.
this was the start of the math fumbling
http://www.youtube.com/watch?v=JuVtAwqRWL0
and ended up with this
http://www.youtube.com/watch?v=f7ATokNn3RM
never had so much fun before :P
las:
cos(60) / sin(60) = 0.57735
cos(60) / sin(60) = 0.57735
ooops, you are right, i missed one multiplication!!!!
correct code:
Code:
float udHexPrism( vec3 p, vec2 h )
{
float q = abs(p);
return max(q.z-h.y,max(q.x+q.y*0.57735,q.y*1.154700538)-h.x);
}
(q is a vec3, of course)
0.57735 = 1/sqrt(3), 1.1547=2/sqrt(3)
i'll fix my website
Code:
#ifdef GL_ES
precision highp float;
#endif
uniform vec2 resolution;
float udHexPrism( vec2 p, float h )
{
vec2 q = abs(p);
return max(q.x+q.y*0.57735,q.y*1.1547)-h;
}
void main(void)
{
vec2 p = -1.0 + 2.0 * gl_FragCoord.xy / resolution.xy;
p.x *= 1.3333;
float d1 = udHexPrism( p, 1.0 );
float d2 = length(p) - 1.0;
float f = 0.0; if( d1<0.0 ) f=1.0;
float g = 0.0; if( d2<0.0 ) g=1.0;
vec3 col = vec3(f,g,0.0);
gl_FragColor = vec4(col,1.0);
}
i'll fix my website
Here is a comparison of both versions:
"inner" hexagon (yours)
"outer" hexagon (what I proposed)
There are good reasons for both versions - I really prefer the "outer" version for e.g. you have a "circle packing" and use the outer version on it -> it will be space filling.
And the "outer" version has another plus - you can easily express it as dot product (well depends whether it's really a plus - but I like dot products):
btw. there are still some people expecting an e-mail ;D
"inner" hexagon (yours)
"outer" hexagon (what I proposed)
There are good reasons for both versions - I really prefer the "outer" version for e.g. you have a "circle packing" and use the outer version on it -> it will be space filling.
And the "outer" version has another plus - you can easily express it as dot product (well depends whether it's really a plus - but I like dot products):
Code:
float udHexPrism( vec2 p, float h ) {
vec2 q = abs(p);
//return max(q.x*0.866025 + q.y*0.5, q.y)-h;
// can also be expressed as dot prod.
return max(dot(q, vec2(0.866025, 0.5)), q.y)-h;
}
btw. there are still some people expecting an e-mail ;D
i like inners in general (hex, tris, quads, ...) cause then the meaning of the "radious" sort of makes sense, which comes helpful for me when i have to pack or combine geometry, as i know exactly the bounds of my objects and therefore i know when things will touch, etc. but i suppose is a matter or taste and depends no the application. also, i noticed that in fact this gives negative distances in the inside, although i didn't check if they give the correct distance (the images look promising). can anybody check???
as for the email, i stopped the project i wanted to work on. perhaps i'll resume it in 2013...!
as for the email, i stopped the project i wanted to work on. perhaps i'll resume it in 2013...!