Wow. I totally forget that I threw together a first pass of simplex noise in 3D for your preview pleasure. This is “totally” incomplete, but any interested party can muck around with it as is. It is intended to be a reference version that isn’t insanely slow.
I never got around to posting it because I’m not really happy with it as is. If this generates any interest maybe it’ll motivate me to get it to a more complete state.
public class SimplexNoise3D
{
/**
* If <i>true</i> the valid input set to noise is restricted
* to positive numbers. This will produce serious defects
* if any of the input coordinates are negative.
*/
private static final boolean restrictToPositive = false;
/**
* Compile time selection between Perlin's simplex vectors and
* the older vector set from Improved (gradient) noise.
*/
private static final boolean perlinVectors = false;
/**
* Compile time selection: If <i>true</i> vectors are expanded
* to a switch table instead of conditional moves.
*/
private static final boolean expandVectors = true;
/**
* Final scaling factor. Depends on vector set.
*/
private static final float finalScale;
static {
if (perlinVectors)
finalScale = 8;
else
finalScale = 32;
}
/**
*
*/
private static int pseudoFloor(float x)
{
return x >= 0 ? (int)x : (int)x-1;
}
/**
* Completes coordinate hashing and performs the dot product with the selected vector.
*/
private static float dotRandVector(int h, float x, float y, float z)
{
// complete hash generation (see eval)
h = posthash(h) >>> 24;
if (perlinVectors) {
float S;
if (false) {
int b5 = h >> 5 & 1;
int b4 = h >> 4 & 1;
int b3 = h >> 3 & 1;
int b2 = h >> 2 & 1;
int b = h & 3;
float P,Q,R;
switch (b) {
case 1: P=x; Q=y; R=z; break;
case 2: P=y; Q=z; R=x; break;
default: P=z; Q=x; R=y; break;
}
if (true) {
if (b5 == b3) P = -P;
if (b5 == b4) Q = -Q;
if (b5 != (b4^b3)) R = -R;
} else {
P = b5 == b3 ? -P : P;
Q = b5 == b4 ? -Q: Q;
R = b5 != (b4^b3) ? -R : R;
}
S = (b == 0 ? Q + R : b2 == 0 ? Q : R);
return (P + S);
}
else {
S = 0;//tempHack(h,x,y,z);
return S;
}
}
else {
// Use the "older" vector set from Improved Gradient Noise.
if (expandVectors) {
switch(h & 0xF) {
case 0x0: return x+y;
case 0x1: return -x+y;
case 0x2: return x-y;
case 0x3: return -x-y;
case 0x4: return x+z;
case 0x5: return -x+z;
case 0x6: return x-z;
case 0x7: return -x-z;
case 0x8: return y+z;
case 0x9: return -y+z;
case 0xA: return y-z;
case 0xB: return -y-z;
case 0xC: return y+x;
case 0xD: return -y+z;
case 0xE: return y-x;
case 0xF: return -y-z;
default: return 0; // never happens
}
}
else {
h &= 0xF;
float u = (h < 8) ? x : y;
float v = (h < 4) ? y : ((h == 12 || h == 14) ? x : z);
return ((h & 1) == 0 ? u : -u) + ((h & 2) == 0 ? v : -v);
}
}
}
/** Properly rounded: 1/3 */
private static final float K0 = 0x1.555556p-2f;
/** Properly rounded: 1/6 */
private static final float K1 = 0x1.555556p-3f;
private static final boolean simpleHash = true;
private static final int M = 0x5bd1e995;
private static final int prehash(int i, int j, int k)
{
if (simpleHash)
return (i<<16)^(j<<8)^(k);
else {
int h;
i *= M; i ^= i >>> 24; i *= M;
j *= M; j ^= j >>> 24; j *= M;
k *= M; k ^= k >>> 24; k *= M;
h = i; h *= M;
h ^= j; h *= M;
h ^= k;
return h;
}
}
private static final int posthash(int h)
{
if (simpleHash)
h *= 0xB5262C85;
else {
h ^= h >>> 13; h *= M; h ^= h >>> 15;
}
return h;
}
/**
*
*/
public static float eval(float x, float y, float z)
{
// project the input coordinate (via skew)
float s = (x + y + z) * K0;
int i, j, k;
if (restrictToPositive) {
// This is possible if input is restricted to positive values.
// The center of the image in the example is (0,0,0), so will
// show the (serious) defects if used and the inputs are negative.
//
// The major upside is that the code will run significantly
// faster if this input limitation is required.
i = (int)(x+s);
j = (int)(y+s);
k = (int)(z+s);
}
else {
i = pseudoFloor(x+s);
j = pseudoFloor(y+s);
k = pseudoFloor(z+s);
}
// unproject the coordinate back to Euclidean space.
float t = (i + j + k) * K1;
float x0 = x - (i-t);
float y0 = y - (j-t);
float z0 = z - (k-t);
// locate the simplex of the input coordinate.
int i1, j1, k1;
int i2, j2, k2;
if (x0 >= y0) {
i2 = 1;
j1 = 0;
if (y0 >= z0) {
i1 = j2 = 1;
k1 = k2 = 0;
}
else {
j2 = 0;
k2 = 1;
if (x0 >= z0)
i1 = 1;
else
i1 = 0;
k1 = i1 ^ 1;
}
}
else {
i1 = 0;
j2 = 1;
if (y0 < z0) {
j1 = i2 = 0;
k1 = k2 = 1;
}
else {
j1 = 1;
k1 = 0;
if (x0 < z0) {
i2 = 0;
k2 = 1;
}
else {
i2 = 1;
k2 = 0;
}
}
}
// calculate the four coordinates
float x1 = x0 - i1 + K1;
float y1 = y0 - j1 + K1;
float z1 = z0 - k1 + K1;
float x2 = x0 - i2 + 2*K1;
float y2 = y0 - j2 + 2*K1;
float z2 = z0 - k2 + 2*K1;
float x3 = x0 - 1 + 3*K1;
float y3 = y0 - 1 + 3*K1;
float z3 = z0 - 1 + 3*K1;
// Step 1 of hash generation, if needed the hashing
// is completed in 'doRandVector'.
// This could be improved
int h0 = prehash(i, j, k );
int h1 = prehash(i+i1, j+j1, k+k1);
int h2 = prehash(i+i2, j+j2, k+k2);
int h3 = prehash(i+ 1, j+ 1, k+ 1);
float n = 0;
// Calculate and sum the results of the four coordinates.
t = 0.6f - x0*x0 - y0*y0 - z0*z0;
if (t > 0){
t *= t;
n += t*t * dotRandVector(h0, x0, y0, z0);
}
t = 0.6f - x1*x1 - y1*y1 - z1*z1;
if (t > 0) {
t *= t;
n += t*t * dotRandVector(h1, x1, y1, z1);
}
t = 0.6f - x2*x2 - y2*y2 - z2*z2;
if (t > 0) {
t *= t;
n += t*t * dotRandVector(h2, x2, y2, z2);
}
t = 0.6f - x3*x3 - y3*y3 - z3*z3;
if (t > 0) {
t *= t;
n += t*t * dotRandVector(h3, x3, y3, z3);
}
return finalScale * n;
}
}
(EDIT 20120830 - added configurable hash)
