This repository has been archived on 2023-10-09. You can view files and clone it, but cannot push or open issues or pull requests.
Files
blender-archive/source/blender/blenlib/intern/noise.c

1409 lines
60 KiB
C

/*
*
* $Id$
*
* ***** BEGIN GPL/BL DUAL LICENSE BLOCK *****
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version. The Blender
* Foundation also sells licenses for use in proprietary software under
* the Blender License. See http://www.blender.org/BL/ for information
* about this.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software Foundation,
* Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
*
* The Original Code is Copyright (C) 2001-2002 by NaN Holding BV.
* All rights reserved.
*
* The Original Code is: all of this file.
*
* Contributor(s): none yet.
*
* ***** END GPL/BL DUAL LICENSE BLOCK *****
*
*/
#ifdef _WIN32
#pragma warning (disable : 4244) // "conversion from double to float"
#pragma warning (disable : 4305) // "truncation from const double to float"
#endif
#include <math.h>
#include "BLI_blenlib.h"
#ifdef HAVE_CONFIG_H
#include <config.h>
#endif
/* local */
float noise3_perlin(float vec[3]);
float turbulence_perlin(float *point, float lofreq, float hifreq);
float turbulencep(float noisesize, float x, float y, float z, int nr);
#define HASHVEC(x,y,z) hashvectf+3*hash[ (hash[ (hash[(z) & 255]+(y)) & 255]+(x)) & 255]
/* needed for voronoi */
#define HASHPNT(x,y,z) hashpntf+3*hash[ (hash[ (hash[(z) & 255]+(y)) & 255]+(x)) & 255]
static float hashpntf[768] = {0.536902, 0.020915, 0.501445, 0.216316, 0.517036, 0.822466, 0.965315,
0.377313, 0.678764, 0.744545, 0.097731, 0.396357, 0.247202, 0.520897,
0.613396, 0.542124, 0.146813, 0.255489, 0.810868, 0.638641, 0.980742,
0.292316, 0.357948, 0.114382, 0.861377, 0.629634, 0.722530, 0.714103,
0.048549, 0.075668, 0.564920, 0.162026, 0.054466, 0.411738, 0.156897,
0.887657, 0.599368, 0.074249, 0.170277, 0.225799, 0.393154, 0.301348,
0.057434, 0.293849, 0.442745, 0.150002, 0.398732, 0.184582, 0.915200,
0.630984, 0.974040, 0.117228, 0.795520, 0.763238, 0.158982, 0.616211,
0.250825, 0.906539, 0.316874, 0.676205, 0.234720, 0.667673, 0.792225,
0.273671, 0.119363, 0.199131, 0.856716, 0.828554, 0.900718, 0.705960,
0.635923, 0.989433, 0.027261, 0.283507, 0.113426, 0.388115, 0.900176,
0.637741, 0.438802, 0.715490, 0.043692, 0.202640, 0.378325, 0.450325,
0.471832, 0.147803, 0.906899, 0.524178, 0.784981, 0.051483, 0.893369,
0.596895, 0.275635, 0.391483, 0.844673, 0.103061, 0.257322, 0.708390,
0.504091, 0.199517, 0.660339, 0.376071, 0.038880, 0.531293, 0.216116,
0.138672, 0.907737, 0.807994, 0.659582, 0.915264, 0.449075, 0.627128,
0.480173, 0.380942, 0.018843, 0.211808, 0.569701, 0.082294, 0.689488,
0.573060, 0.593859, 0.216080, 0.373159, 0.108117, 0.595539, 0.021768,
0.380297, 0.948125, 0.377833, 0.319699, 0.315249, 0.972805, 0.792270,
0.445396, 0.845323, 0.372186, 0.096147, 0.689405, 0.423958, 0.055675,
0.117940, 0.328456, 0.605808, 0.631768, 0.372170, 0.213723, 0.032700,
0.447257, 0.440661, 0.728488, 0.299853, 0.148599, 0.649212, 0.498381,
0.049921, 0.496112, 0.607142, 0.562595, 0.990246, 0.739659, 0.108633,
0.978156, 0.209814, 0.258436, 0.876021, 0.309260, 0.600673, 0.713597,
0.576967, 0.641402, 0.853930, 0.029173, 0.418111, 0.581593, 0.008394,
0.589904, 0.661574, 0.979326, 0.275724, 0.111109, 0.440472, 0.120839,
0.521602, 0.648308, 0.284575, 0.204501, 0.153286, 0.822444, 0.300786,
0.303906, 0.364717, 0.209038, 0.916831, 0.900245, 0.600685, 0.890002,
0.581660, 0.431154, 0.705569, 0.551250, 0.417075, 0.403749, 0.696652,
0.292652, 0.911372, 0.690922, 0.323718, 0.036773, 0.258976, 0.274265,
0.225076, 0.628965, 0.351644, 0.065158, 0.080340, 0.467271, 0.130643,
0.385914, 0.919315, 0.253821, 0.966163, 0.017439, 0.392610, 0.478792,
0.978185, 0.072691, 0.982009, 0.097987, 0.731533, 0.401233, 0.107570,
0.349587, 0.479122, 0.700598, 0.481751, 0.788429, 0.706864, 0.120086,
0.562691, 0.981797, 0.001223, 0.192120, 0.451543, 0.173092, 0.108960,
0.549594, 0.587892, 0.657534, 0.396365, 0.125153, 0.666420, 0.385823,
0.890916, 0.436729, 0.128114, 0.369598, 0.759096, 0.044677, 0.904752,
0.088052, 0.621148, 0.005047, 0.452331, 0.162032, 0.494238, 0.523349,
0.741829, 0.698450, 0.452316, 0.563487, 0.819776, 0.492160, 0.004210,
0.647158, 0.551475, 0.362995, 0.177937, 0.814722, 0.727729, 0.867126,
0.997157, 0.108149, 0.085726, 0.796024, 0.665075, 0.362462, 0.323124,
0.043718, 0.042357, 0.315030, 0.328954, 0.870845, 0.683186, 0.467922,
0.514894, 0.809971, 0.631979, 0.176571, 0.366320, 0.850621, 0.505555,
0.749551, 0.750830, 0.401714, 0.481216, 0.438393, 0.508832, 0.867971,
0.654581, 0.058204, 0.566454, 0.084124, 0.548539, 0.902690, 0.779571,
0.562058, 0.048082, 0.863109, 0.079290, 0.713559, 0.783496, 0.265266,
0.672089, 0.786939, 0.143048, 0.086196, 0.876129, 0.408708, 0.229312,
0.629995, 0.206665, 0.207308, 0.710079, 0.341704, 0.264921, 0.028748,
0.629222, 0.470173, 0.726228, 0.125243, 0.328249, 0.794187, 0.741340,
0.489895, 0.189396, 0.724654, 0.092841, 0.039809, 0.860126, 0.247701,
0.655331, 0.964121, 0.672536, 0.044522, 0.690567, 0.837238, 0.631520,
0.953734, 0.352484, 0.289026, 0.034152, 0.852575, 0.098454, 0.795529,
0.452181, 0.826159, 0.186993, 0.820725, 0.440328, 0.922137, 0.704592,
0.915437, 0.738183, 0.733461, 0.193798, 0.929213, 0.161390, 0.318547,
0.888751, 0.430968, 0.740837, 0.193544, 0.872253, 0.563074, 0.274598,
0.347805, 0.666176, 0.449831, 0.800991, 0.588727, 0.052296, 0.714761,
0.420620, 0.570325, 0.057550, 0.210888, 0.407312, 0.662848, 0.924382,
0.895958, 0.775198, 0.688605, 0.025721, 0.301913, 0.791408, 0.500602,
0.831984, 0.828509, 0.642093, 0.494174, 0.525880, 0.446365, 0.440063,
0.763114, 0.630358, 0.223943, 0.333806, 0.906033, 0.498306, 0.241278,
0.427640, 0.772683, 0.198082, 0.225379, 0.503894, 0.436599, 0.016503,
0.803725, 0.189878, 0.291095, 0.499114, 0.151573, 0.079031, 0.904618,
0.708535, 0.273900, 0.067419, 0.317124, 0.936499, 0.716511, 0.543845,
0.939909, 0.826574, 0.715090, 0.154864, 0.750150, 0.845808, 0.648108,
0.556564, 0.644757, 0.140873, 0.799167, 0.632989, 0.444245, 0.471978,
0.435910, 0.359793, 0.216241, 0.007633, 0.337236, 0.857863, 0.380247,
0.092517, 0.799973, 0.919000, 0.296798, 0.096989, 0.854831, 0.165369,
0.568475, 0.216855, 0.020457, 0.835511, 0.538039, 0.999742, 0.620226,
0.244053, 0.060399, 0.323007, 0.294874, 0.988899, 0.384919, 0.735655,
0.773428, 0.549776, 0.292882, 0.660611, 0.593507, 0.621118, 0.175269,
0.682119, 0.794493, 0.868197, 0.632150, 0.807823, 0.509656, 0.482035,
0.001780, 0.259126, 0.358002, 0.280263, 0.192985, 0.290367, 0.208111,
0.917633, 0.114422, 0.925491, 0.981110, 0.255570, 0.974862, 0.016629,
0.552599, 0.575741, 0.612978, 0.615965, 0.803615, 0.772334, 0.089745,
0.838812, 0.634542, 0.113709, 0.755832, 0.577589, 0.667489, 0.529834,
0.325660, 0.817597, 0.316557, 0.335093, 0.737363, 0.260951, 0.737073,
0.049540, 0.735541, 0.988891, 0.299116, 0.147695, 0.417271, 0.940811,
0.524160, 0.857968, 0.176403, 0.244835, 0.485759, 0.033353, 0.280319,
0.750688, 0.755809, 0.924208, 0.095956, 0.962504, 0.275584, 0.173715,
0.942716, 0.706721, 0.078464, 0.576716, 0.804667, 0.559249, 0.900611,
0.646904, 0.432111, 0.927885, 0.383277, 0.269973, 0.114244, 0.574867,
0.150703, 0.241855, 0.272871, 0.199950, 0.079719, 0.868566, 0.962833,
0.789122, 0.320025, 0.905554, 0.234876, 0.991356, 0.061913, 0.732911,
0.785960, 0.874074, 0.069035, 0.658632, 0.309901, 0.023676, 0.791603,
0.764661, 0.661278, 0.319583, 0.829650, 0.117091, 0.903124, 0.982098,
0.161631, 0.193576, 0.670428, 0.857390, 0.003760, 0.572578, 0.222162,
0.114551, 0.420118, 0.530404, 0.470682, 0.525527, 0.764281, 0.040596,
0.443275, 0.501124, 0.816161, 0.417467, 0.332172, 0.447565, 0.614591,
0.559246, 0.805295, 0.226342, 0.155065, 0.714630, 0.160925, 0.760001,
0.453456, 0.093869, 0.406092, 0.264801, 0.720370, 0.743388, 0.373269,
0.403098, 0.911923, 0.897249, 0.147038, 0.753037, 0.516093, 0.739257,
0.175018, 0.045768, 0.735857, 0.801330, 0.927708, 0.240977, 0.591870,
0.921831, 0.540733, 0.149100, 0.423152, 0.806876, 0.397081, 0.061100,
0.811630, 0.044899, 0.460915, 0.961202, 0.822098, 0.971524, 0.867608,
0.773604, 0.226616, 0.686286, 0.926972, 0.411613, 0.267873, 0.081937,
0.226124, 0.295664, 0.374594, 0.533240, 0.237876, 0.669629, 0.599083,
0.513081, 0.878719, 0.201577, 0.721296, 0.495038, 0.079760, 0.965959,
0.233090, 0.052496, 0.714748, 0.887844, 0.308724, 0.972885, 0.723337,
0.453089, 0.914474, 0.704063, 0.823198, 0.834769, 0.906561, 0.919600,
0.100601, 0.307564, 0.901977, 0.468879, 0.265376, 0.885188, 0.683875,
0.868623, 0.081032, 0.466835, 0.199087, 0.663437, 0.812241, 0.311337,
0.821361, 0.356628, 0.898054, 0.160781, 0.222539, 0.714889, 0.490287,
0.984915, 0.951755, 0.964097, 0.641795, 0.815472, 0.852732, 0.862074,
0.051108, 0.440139, 0.323207, 0.517171, 0.562984, 0.115295, 0.743103,
0.977914, 0.337596, 0.440694, 0.535879, 0.959427, 0.351427, 0.704361,
0.010826, 0.131162, 0.577080, 0.349572, 0.774892, 0.425796, 0.072697,
0.500001, 0.267322, 0.909654, 0.206176, 0.223987, 0.937698, 0.323423,
0.117501, 0.490308, 0.474372, 0.689943, 0.168671, 0.719417, 0.188928,
0.330464, 0.265273, 0.446271, 0.171933, 0.176133, 0.474616, 0.140182,
0.114246, 0.905043, 0.713870, 0.555261, 0.951333};
unsigned char hash[512]= {
0xA2,0xA0,0x19,0x3B,0xF8,0xEB,0xAA,0xEE,0xF3,0x1C,0x67,0x28,0x1D,0xED,0x0,0xDE,0x95,0x2E,0xDC,0x3F,0x3A,0x82,0x35,0x4D,0x6C,0xBA,0x36,0xD0,0xF6,0xC,0x79,0x32,0xD1,0x59,0xF4,0x8,0x8B,0x63,0x89,0x2F,0xB8,0xB4,0x97,0x83,0xF2,0x8F,0x18,0xC7,0x51,0x14,0x65,0x87,0x48,0x20,0x42,0xA8,0x80,0xB5,0x40,0x13,0xB2,0x22,0x7E,0x57,
0xBC,0x7F,0x6B,0x9D,0x86,0x4C,0xC8,0xDB,0x7C,0xD5,0x25,0x4E,0x5A,0x55,0x74,0x50,0xCD,0xB3,0x7A,0xBB,0xC3,0xCB,0xB6,0xE2,0xE4,0xEC,0xFD,0x98,0xB,0x96,0xD3,0x9E,0x5C,0xA1,0x64,0xF1,0x81,0x61,0xE1,0xC4,0x24,0x72,0x49,0x8C,0x90,0x4B,0x84,0x34,0x38,0xAB,0x78,0xCA,0x1F,0x1,0xD7,0x93,0x11,0xC1,0x58,0xA9,0x31,0xF9,0x44,0x6D,
0xBF,0x33,0x9C,0x5F,0x9,0x94,0xA3,0x85,0x6,0xC6,0x9A,0x1E,0x7B,0x46,0x15,0x30,0x27,0x2B,0x1B,0x71,0x3C,0x5B,0xD6,0x6F,0x62,0xAC,0x4F,0xC2,0xC0,0xE,0xB1,0x23,0xA7,0xDF,0x47,0xB0,0x77,0x69,0x5,0xE9,0xE6,0xE7,0x76,0x73,0xF,0xFE,0x6E,0x9B,0x56,0xEF,0x12,0xA5,0x37,0xFC,0xAE,0xD9,0x3,0x8E,0xDD,0x10,0xB9,0xCE,0xC9,0x8D,
0xDA,0x2A,0xBD,0x68,0x17,0x9F,0xBE,0xD4,0xA,0xCC,0xD2,0xE8,0x43,0x3D,0x70,0xB7,0x2,0x7D,0x99,0xD8,0xD,0x60,0x8A,0x4,0x2C,0x3E,0x92,0xE5,0xAF,0x53,0x7,0xE0,0x29,0xA6,0xC5,0xE3,0xF5,0xF7,0x4A,0x41,0x26,0x6A,0x16,0x5E,0x52,0x2D,0x21,0xAD,0xF0,0x91,0xFF,0xEA,0x54,0xFA,0x66,0x1A,0x45,0x39,0xCF,0x75,0xA4,0x88,0xFB,0x5D,
0xA2,0xA0,0x19,0x3B,0xF8,0xEB,0xAA,0xEE,0xF3,0x1C,0x67,0x28,0x1D,0xED,0x0,0xDE,0x95,0x2E,0xDC,0x3F,0x3A,0x82,0x35,0x4D,0x6C,0xBA,0x36,0xD0,0xF6,0xC,0x79,0x32,0xD1,0x59,0xF4,0x8,0x8B,0x63,0x89,0x2F,0xB8,0xB4,0x97,0x83,0xF2,0x8F,0x18,0xC7,0x51,0x14,0x65,0x87,0x48,0x20,0x42,0xA8,0x80,0xB5,0x40,0x13,0xB2,0x22,0x7E,0x57,
0xBC,0x7F,0x6B,0x9D,0x86,0x4C,0xC8,0xDB,0x7C,0xD5,0x25,0x4E,0x5A,0x55,0x74,0x50,0xCD,0xB3,0x7A,0xBB,0xC3,0xCB,0xB6,0xE2,0xE4,0xEC,0xFD,0x98,0xB,0x96,0xD3,0x9E,0x5C,0xA1,0x64,0xF1,0x81,0x61,0xE1,0xC4,0x24,0x72,0x49,0x8C,0x90,0x4B,0x84,0x34,0x38,0xAB,0x78,0xCA,0x1F,0x1,0xD7,0x93,0x11,0xC1,0x58,0xA9,0x31,0xF9,0x44,0x6D,
0xBF,0x33,0x9C,0x5F,0x9,0x94,0xA3,0x85,0x6,0xC6,0x9A,0x1E,0x7B,0x46,0x15,0x30,0x27,0x2B,0x1B,0x71,0x3C,0x5B,0xD6,0x6F,0x62,0xAC,0x4F,0xC2,0xC0,0xE,0xB1,0x23,0xA7,0xDF,0x47,0xB0,0x77,0x69,0x5,0xE9,0xE6,0xE7,0x76,0x73,0xF,0xFE,0x6E,0x9B,0x56,0xEF,0x12,0xA5,0x37,0xFC,0xAE,0xD9,0x3,0x8E,0xDD,0x10,0xB9,0xCE,0xC9,0x8D,
0xDA,0x2A,0xBD,0x68,0x17,0x9F,0xBE,0xD4,0xA,0xCC,0xD2,0xE8,0x43,0x3D,0x70,0xB7,0x2,0x7D,0x99,0xD8,0xD,0x60,0x8A,0x4,0x2C,0x3E,0x92,0xE5,0xAF,0x53,0x7,0xE0,0x29,0xA6,0xC5,0xE3,0xF5,0xF7,0x4A,0x41,0x26,0x6A,0x16,0x5E,0x52,0x2D,0x21,0xAD,0xF0,0x91,0xFF,0xEA,0x54,0xFA,0x66,0x1A,0x45,0x39,0xCF,0x75,0xA4,0x88,0xFB,0x5D,
};
float hashvectf[768]= {
0.33783,0.715698,-0.611206,-0.944031,-0.326599,-0.045624,-0.101074,-0.416443,-0.903503,0.799286,0.49411,-0.341949,-0.854645,0.518036,0.033936,0.42514,-0.437866,-0.792114,-0.358948,0.597046,0.717377,-0.985413,0.144714,0.089294,-0.601776,-0.33728,-0.723907,-0.449921,0.594513,0.666382,0.208313,-0.10791,
0.972076,0.575317,0.060425,0.815643,0.293365,-0.875702,-0.383453,0.293762,0.465759,0.834686,-0.846008,-0.233398,-0.47934,-0.115814,0.143036,-0.98291,0.204681,-0.949036,-0.239532,0.946716,-0.263947,0.184326,-0.235596,0.573822,0.784332,0.203705,-0.372253,-0.905487,0.756989,-0.651031,0.055298,0.497803,
0.814697,-0.297363,-0.16214,0.063995,-0.98468,-0.329254,0.834381,0.441925,0.703827,-0.527039,-0.476227,0.956421,0.266113,0.119781,0.480133,0.482849,0.7323,-0.18631,0.961212,-0.203125,-0.748474,-0.656921,-0.090393,-0.085052,-0.165253,0.982544,-0.76947,0.628174,-0.115234,0.383148,0.537659,0.751068,
0.616486,-0.668488,-0.415924,-0.259979,-0.630005,0.73175,0.570953,-0.087952,0.816223,-0.458008,0.023254,0.888611,-0.196167,0.976563,-0.088287,-0.263885,-0.69812,-0.665527,0.437134,-0.892273,-0.112793,-0.621674,-0.230438,0.748566,0.232422,0.900574,-0.367249,0.22229,-0.796143,0.562744,-0.665497,-0.73764,
0.11377,0.670135,0.704803,0.232605,0.895599,0.429749,-0.114655,-0.11557,-0.474243,0.872742,0.621826,0.604004,-0.498444,-0.832214,0.012756,0.55426,-0.702484,0.705994,-0.089661,-0.692017,0.649292,0.315399,-0.175995,-0.977997,0.111877,0.096954,-0.04953,0.994019,0.635284,-0.606689,-0.477783,-0.261261,
-0.607422,-0.750153,0.983276,0.165436,0.075958,-0.29837,0.404083,-0.864655,-0.638672,0.507721,0.578156,0.388214,0.412079,0.824249,0.556183,-0.208832,0.804352,0.778442,0.562012,0.27951,-0.616577,0.781921,-0.091522,0.196289,0.051056,0.979187,-0.121216,0.207153,-0.970734,-0.173401,-0.384735,0.906555,
0.161499,-0.723236,-0.671387,0.178497,-0.006226,-0.983887,-0.126038,0.15799,0.97934,0.830475,-0.024811,0.556458,-0.510132,-0.76944,0.384247,0.81424,0.200104,-0.544891,-0.112549,-0.393311,-0.912445,0.56189,0.152222,-0.813049,0.198914,-0.254517,-0.946381,-0.41217,0.690979,-0.593811,-0.407257,0.324524,
0.853668,-0.690186,0.366119,-0.624115,-0.428345,0.844147,-0.322296,-0.21228,-0.297546,-0.930756,-0.273071,0.516113,0.811798,0.928314,0.371643,0.007233,0.785828,-0.479218,-0.390778,-0.704895,0.058929,0.706818,0.173248,0.203583,0.963562,0.422211,-0.904297,-0.062469,-0.363312,-0.182465,0.913605,0.254028,
-0.552307,-0.793945,-0.28891,-0.765747,-0.574554,0.058319,0.291382,0.954803,0.946136,-0.303925,0.111267,-0.078156,0.443695,-0.892731,0.182098,0.89389,0.409515,-0.680298,-0.213318,0.701141,0.062469,0.848389,-0.525635,-0.72879,-0.641846,0.238342,-0.88089,0.427673,0.202637,-0.532501,-0.21405,0.818878,
0.948975,-0.305084,0.07962,0.925446,0.374664,0.055817,0.820923,0.565491,0.079102,0.25882,0.099792,-0.960724,-0.294617,0.910522,0.289978,0.137115,0.320038,-0.937408,-0.908386,0.345276,-0.235718,-0.936218,0.138763,0.322754,0.366577,0.925934,-0.090637,0.309296,-0.686829,-0.657684,0.66983,0.024445,
0.742065,-0.917999,-0.059113,-0.392059,0.365509,0.462158,-0.807922,0.083374,0.996399,-0.014801,0.593842,0.253143,-0.763672,0.974976,-0.165466,0.148285,0.918976,0.137299,0.369537,0.294952,0.694977,0.655731,0.943085,0.152618,-0.295319,0.58783,-0.598236,0.544495,0.203796,0.678223,0.705994,-0.478821,
-0.661011,0.577667,0.719055,-0.1698,-0.673828,-0.132172,-0.965332,0.225006,-0.981873,-0.14502,0.121979,0.763458,0.579742,0.284546,-0.893188,0.079681,0.442474,-0.795776,-0.523804,0.303802,0.734955,0.67804,-0.007446,0.15506,0.986267,-0.056183,0.258026,0.571503,-0.778931,-0.681549,-0.702087,-0.206116,
-0.96286,-0.177185,0.203613,-0.470978,-0.515106,0.716095,-0.740326,0.57135,0.354095,-0.56012,-0.824982,-0.074982,-0.507874,0.753204,0.417969,-0.503113,0.038147,0.863342,0.594025,0.673553,-0.439758,-0.119873,-0.005524,-0.992737,0.098267,-0.213776,0.971893,-0.615631,0.643951,0.454163,0.896851,-0.441071,
0.032166,-0.555023,0.750763,-0.358093,0.398773,0.304688,0.864929,-0.722961,0.303589,0.620544,-0.63559,-0.621948,-0.457306,-0.293243,0.072327,0.953278,-0.491638,0.661041,-0.566772,-0.304199,-0.572083,-0.761688,0.908081,-0.398956,0.127014,-0.523621,-0.549683,-0.650848,-0.932922,-0.19986,0.299408,0.099426,
0.140869,0.984985,-0.020325,-0.999756,-0.002319,0.952667,0.280853,-0.11615,-0.971893,0.082581,0.220337,0.65921,0.705292,-0.260651,0.733063,-0.175537,0.657043,-0.555206,0.429504,-0.712189,0.400421,-0.89859,0.179352,0.750885,-0.19696,0.630341,0.785675,-0.569336,0.241821,-0.058899,-0.464111,0.883789,
0.129608,-0.94519,0.299622,-0.357819,0.907654,0.219238,-0.842133,-0.439117,-0.312927,-0.313477,0.84433,0.434479,-0.241211,0.053253,0.968994,0.063873,0.823273,0.563965,0.476288,0.862152,-0.172516,0.620941,-0.298126,0.724915,0.25238,-0.749359,-0.612122,-0.577545,0.386566,0.718994,-0.406342,-0.737976,
0.538696,0.04718,0.556305,0.82959,-0.802856,0.587463,0.101166,-0.707733,-0.705963,0.026428,0.374908,0.68457,0.625092,0.472137,0.208405,-0.856506,-0.703064,-0.581085,-0.409821,-0.417206,-0.736328,0.532623,-0.447876,-0.20285,-0.870728,0.086945,-0.990417,0.107086,0.183685,0.018341,-0.982788,0.560638,
-0.428864,0.708282,0.296722,-0.952576,-0.0672,0.135773,0.990265,0.030243,-0.068787,0.654724,0.752686,0.762604,-0.551758,0.337585,-0.819611,-0.407684,0.402466,-0.727844,-0.55072,-0.408539,-0.855774,-0.480011,0.19281,0.693176,-0.079285,0.716339,0.226013,0.650116,-0.725433,0.246704,0.953369,-0.173553,
-0.970398,-0.239227,-0.03244,0.136383,-0.394318,0.908752,0.813232,0.558167,0.164368,0.40451,0.549042,-0.731323,-0.380249,-0.566711,0.730865,0.022156,0.932739,0.359741,0.00824,0.996552,-0.082306,0.956635,-0.065338,-0.283722,-0.743561,0.008209,0.668579,-0.859589,-0.509674,0.035767,-0.852234,0.363678,
-0.375977,-0.201965,-0.970795,-0.12915,0.313477,0.947327,0.06546,-0.254028,-0.528259,0.81015,0.628052,0.601105,0.49411,-0.494385,0.868378,0.037933,0.275635,-0.086426,0.957336,-0.197937,0.468903,-0.860748,0.895599,0.399384,0.195801,0.560791,0.825012,-0.069214,0.304199,-0.849487,0.43103,0.096375,
0.93576,0.339111,-0.051422,0.408966,-0.911072,0.330444,0.942841,-0.042389,-0.452362,-0.786407,0.420563,0.134308,-0.933472,-0.332489,0.80191,-0.566711,-0.188934,-0.987946,-0.105988,0.112518,-0.24408,0.892242,-0.379791,-0.920502,0.229095,-0.316376,0.7789,0.325958,0.535706,-0.912872,0.185211,-0.36377,
-0.184784,0.565369,-0.803833,-0.018463,0.119537,0.992615,-0.259247,-0.935608,0.239532,-0.82373,-0.449127,-0.345947,-0.433105,0.659515,0.614349,-0.822754,0.378845,-0.423676,0.687195,-0.674835,-0.26889,-0.246582,-0.800842,0.545715,-0.729187,-0.207794,0.651978,0.653534,-0.610443,-0.447388,0.492584,-0.023346,
0.869934,0.609039,0.009094,-0.79306,0.962494,-0.271088,-0.00885,0.2659,-0.004913,0.963959,0.651245,0.553619,-0.518951,0.280548,-0.84314,0.458618,-0.175293,-0.983215,0.049805,0.035339,-0.979919,0.196045,-0.982941,0.164307,-0.082245,0.233734,-0.97226,-0.005005,-0.747253,-0.611328,0.260437,0.645599,
0.592773,0.481384,0.117706,-0.949524,-0.29068,-0.535004,-0.791901,-0.294312,-0.627167,-0.214447,0.748718,-0.047974,-0.813477,-0.57959,-0.175537,0.477264,-0.860992,0.738556,-0.414246,-0.53183,0.562561,-0.704071,0.433289,-0.754944,0.64801,-0.100586,0.114716,0.044525,-0.992371,0.966003,0.244873,-0.082764,
};
/**************************/
/* IMPROVED PERLIN NOISE */
/**************************/
#define lerp(t, a, b) ((a)+(t)*((b)-(a)))
#define npfade(t) ((t)*(t)*(t)*((t)*((t)*6-15)+10))
static float grad(int hash, float x, float y, float z)
{
int h = hash & 15; // CONVERT LO 4 BITS OF HASH CODE
float u = h<8 ? x : y, // INTO 12 GRADIENT DIRECTIONS.
v = h<4 ? y : h==12||h==14 ? x : z;
return ((h&1) == 0 ? u : -u) + ((h&2) == 0 ? v : -v);
}
/* instead of adding another permutation array, just use hash table defined above */
static float newPerlin(float x, float y, float z)
{
int A, AA, AB, B, BA, BB;
float u=floor(x), v=floor(y), w=floor(z);
int X=((int)u) & 255, Y=((int)v) & 255, Z=((int)w) & 255; // FIND UNIT CUBE THAT CONTAINS POINT
x -= u; // FIND RELATIVE X,Y,Z
y -= v; // OF POINT IN CUBE.
z -= w;
u = npfade(x); // COMPUTE FADE CURVES
v = npfade(y); // FOR EACH OF X,Y,Z.
w = npfade(z);
A = hash[X ]+Y; AA = hash[A]+Z; AB = hash[A+1]+Z; // HASH COORDINATES OF
B = hash[X+1]+Y; BA = hash[B]+Z; BB = hash[B+1]+Z; // THE 8 CUBE CORNERS,
return lerp(w, lerp(v, lerp(u, grad(hash[AA ], x , y , z ), // AND ADD
grad(hash[BA ], x-1, y , z )), // BLENDED
lerp(u, grad(hash[AB ], x , y-1, z ), // RESULTS
grad(hash[BB ], x-1, y-1, z ))),// FROM 8
lerp(v, lerp(u, grad(hash[AA+1], x , y , z-1 ), // CORNERS
grad(hash[BA+1], x-1, y , z-1 )), // OF CUBE
lerp(u, grad(hash[AB+1], x , y-1, z-1 ),
grad(hash[BB+1], x-1, y-1, z-1 ))));
}
/* for use with BLI_gNoise()/BLI_gTurbulence(), returns unsigned improved perlin noise */
static float newPerlinU(float x, float y, float z)
{
return (0.5+0.5*newPerlin(x, y, z));
}
/**************************/
/* END OF IMPROVED PERLIN */
/**************************/
/* Was BLI_hnoise(), removed noisesize, so other functions can call it without scaling. */
static float orgBlenderNoise(float x, float y, float z)
{
register float cn1, cn2, cn3, cn4, cn5, cn6, i, *h;
float ox, oy, oz, jx, jy, jz;
float n= 0.5;
int ix, iy, iz, b00, b01, b10, b11, b20, b21;
ox= (x- (ix= (int)floor(x)) );
oy= (y- (iy= (int)floor(y)) );
oz= (z- (iz= (int)floor(z)) );
jx= ox-1;
jy= oy-1;
jz= oz-1;
cn1=ox*ox; cn2=oy*oy; cn3=oz*oz;
cn4=jx*jx; cn5=jy*jy; cn6=jz*jz;
cn1= 1.0-3.0*cn1+2.0*cn1*ox;
cn2= 1.0-3.0*cn2+2.0*cn2*oy;
cn3= 1.0-3.0*cn3+2.0*cn3*oz;
cn4= 1.0-3.0*cn4-2.0*cn4*jx;
cn5= 1.0-3.0*cn5-2.0*cn5*jy;
cn6= 1.0-3.0*cn6-2.0*cn6*jz;
b00= hash[ hash[ix & 255]+(iy & 255)];
b10= hash[ hash[(ix+1) & 255]+(iy & 255)];
b01= hash[ hash[ix & 255]+((iy+1) & 255)];
b11= hash[ hash[(ix+1) & 255]+((iy+1) & 255)];
b20=iz & 255; b21= (iz+1) & 255;
/* 0 */
i= (cn1*cn2*cn3);
h=hashvectf+ 3*hash[b20+b00];
n+= i*(h[0]*ox+h[1]*oy+h[2]*oz);
/* 1 */
i= (cn1*cn2*cn6);
h=hashvectf+ 3*hash[b21+b00];
n+= i*(h[0]*ox+h[1]*oy+h[2]*jz);
/* 2 */
i= (cn1*cn5*cn3);
h=hashvectf+ 3*hash[b20+b01];
n+= i*(h[0]*ox+h[1]*jy+h[2]*oz);
/* 3 */
i= (cn1*cn5*cn6);
h=hashvectf+ 3*hash[b21+b01];
n+= i*(h[0]*ox+h[1]*jy+h[2]*jz);
/* 4 */
i= cn4*cn2*cn3;
h=hashvectf+ 3*hash[b20+b10];
n+= i*(h[0]*jx+h[1]*oy+h[2]*oz);
/* 5 */
i= cn4*cn2*cn6;
h=hashvectf+ 3*hash[b21+b10];
n+= i*(h[0]*jx+h[1]*oy+h[2]*jz);
/* 6 */
i= cn4*cn5*cn3;
h=hashvectf+ 3*hash[b20+b11];
n+= i*(h[0]*jx+h[1]*jy+h[2]*oz);
/* 7 */
i= (cn4*cn5*cn6);
h=hashvectf+ 3*hash[b21+b11];
n+= i*(h[0]*jx+h[1]*jy+h[2]*jz);
if(n<0.0) n=0.0; else if(n>1.0) n=1.0;
return n;
}
/* as orgBlenderNoise(), returning signed noise */
static float orgBlenderNoiseS(float x, float y, float z)
{
return (2.0*orgBlenderNoise(x, y, z)-1.0);
}
/* separated from orgBlenderNoise above, with scaling */
float BLI_hnoise(float noisesize, float x, float y, float z)
{
if(noisesize==0.0) return 0.0;
x= (1.0+x)/noisesize;
y= (1.0+y)/noisesize;
z= (1.0+z)/noisesize;
return orgBlenderNoise(x, y, z);
}
/* original turbulence functions */
float BLI_turbulence(float noisesize, float x, float y, float z, int nr)
{
float s, d= 0.5, div=1.0;
s= BLI_hnoise(noisesize, x, y, z);
while(nr>0) {
s+= d*BLI_hnoise(noisesize*d, x, y, z);
div+= d;
d*= 0.5;
nr--;
}
return s/div;
}
float BLI_turbulence1(float noisesize, float x, float y, float z, int nr)
{
float s, d= 0.5, div=1.0;
s= fabs( (-1.0+2.0*BLI_hnoise(noisesize, x, y, z)));
while(nr>0) {
s+= fabs(d* (-1.0+2.0*BLI_hnoise(noisesize*d, x, y, z)));
div+= d;
d*= 0.5;
nr--;
}
return s/div;
}
/* ********************* FROM PERLIN HIMSELF: ******************** */
static char p[512+2]= {
0xA2,0xA0,0x19,0x3B,0xF8,0xEB,0xAA,0xEE,0xF3,0x1C,0x67,0x28,0x1D,0xED,0x0,0xDE,0x95,0x2E,0xDC,0x3F,0x3A,0x82,0x35,0x4D,0x6C,0xBA,0x36,0xD0,0xF6,0xC,0x79,0x32,0xD1,0x59,0xF4,0x8,0x8B,0x63,0x89,0x2F,0xB8,0xB4,0x97,0x83,0xF2,0x8F,0x18,0xC7,0x51,0x14,0x65,0x87,0x48,0x20,0x42,0xA8,0x80,0xB5,0x40,0x13,0xB2,0x22,0x7E,0x57,
0xBC,0x7F,0x6B,0x9D,0x86,0x4C,0xC8,0xDB,0x7C,0xD5,0x25,0x4E,0x5A,0x55,0x74,0x50,0xCD,0xB3,0x7A,0xBB,0xC3,0xCB,0xB6,0xE2,0xE4,0xEC,0xFD,0x98,0xB,0x96,0xD3,0x9E,0x5C,0xA1,0x64,0xF1,0x81,0x61,0xE1,0xC4,0x24,0x72,0x49,0x8C,0x90,0x4B,0x84,0x34,0x38,0xAB,0x78,0xCA,0x1F,0x1,0xD7,0x93,0x11,0xC1,0x58,0xA9,0x31,0xF9,0x44,0x6D,
0xBF,0x33,0x9C,0x5F,0x9,0x94,0xA3,0x85,0x6,0xC6,0x9A,0x1E,0x7B,0x46,0x15,0x30,0x27,0x2B,0x1B,0x71,0x3C,0x5B,0xD6,0x6F,0x62,0xAC,0x4F,0xC2,0xC0,0xE,0xB1,0x23,0xA7,0xDF,0x47,0xB0,0x77,0x69,0x5,0xE9,0xE6,0xE7,0x76,0x73,0xF,0xFE,0x6E,0x9B,0x56,0xEF,0x12,0xA5,0x37,0xFC,0xAE,0xD9,0x3,0x8E,0xDD,0x10,0xB9,0xCE,0xC9,0x8D,
0xDA,0x2A,0xBD,0x68,0x17,0x9F,0xBE,0xD4,0xA,0xCC,0xD2,0xE8,0x43,0x3D,0x70,0xB7,0x2,0x7D,0x99,0xD8,0xD,0x60,0x8A,0x4,0x2C,0x3E,0x92,0xE5,0xAF,0x53,0x7,0xE0,0x29,0xA6,0xC5,0xE3,0xF5,0xF7,0x4A,0x41,0x26,0x6A,0x16,0x5E,0x52,0x2D,0x21,0xAD,0xF0,0x91,0xFF,0xEA,0x54,0xFA,0x66,0x1A,0x45,0x39,0xCF,0x75,0xA4,0x88,0xFB,0x5D,
0xA2,0xA0,0x19,0x3B,0xF8,0xEB,0xAA,0xEE,0xF3,0x1C,0x67,0x28,0x1D,0xED,0x0,0xDE,0x95,0x2E,0xDC,0x3F,0x3A,0x82,0x35,0x4D,0x6C,0xBA,0x36,0xD0,0xF6,0xC,0x79,0x32,0xD1,0x59,0xF4,0x8,0x8B,0x63,0x89,0x2F,0xB8,0xB4,0x97,0x83,0xF2,0x8F,0x18,0xC7,0x51,0x14,0x65,0x87,0x48,0x20,0x42,0xA8,0x80,0xB5,0x40,0x13,0xB2,0x22,0x7E,0x57,
0xBC,0x7F,0x6B,0x9D,0x86,0x4C,0xC8,0xDB,0x7C,0xD5,0x25,0x4E,0x5A,0x55,0x74,0x50,0xCD,0xB3,0x7A,0xBB,0xC3,0xCB,0xB6,0xE2,0xE4,0xEC,0xFD,0x98,0xB,0x96,0xD3,0x9E,0x5C,0xA1,0x64,0xF1,0x81,0x61,0xE1,0xC4,0x24,0x72,0x49,0x8C,0x90,0x4B,0x84,0x34,0x38,0xAB,0x78,0xCA,0x1F,0x1,0xD7,0x93,0x11,0xC1,0x58,0xA9,0x31,0xF9,0x44,0x6D,
0xBF,0x33,0x9C,0x5F,0x9,0x94,0xA3,0x85,0x6,0xC6,0x9A,0x1E,0x7B,0x46,0x15,0x30,0x27,0x2B,0x1B,0x71,0x3C,0x5B,0xD6,0x6F,0x62,0xAC,0x4F,0xC2,0xC0,0xE,0xB1,0x23,0xA7,0xDF,0x47,0xB0,0x77,0x69,0x5,0xE9,0xE6,0xE7,0x76,0x73,0xF,0xFE,0x6E,0x9B,0x56,0xEF,0x12,0xA5,0x37,0xFC,0xAE,0xD9,0x3,0x8E,0xDD,0x10,0xB9,0xCE,0xC9,0x8D,
0xDA,0x2A,0xBD,0x68,0x17,0x9F,0xBE,0xD4,0xA,0xCC,0xD2,0xE8,0x43,0x3D,0x70,0xB7,0x2,0x7D,0x99,0xD8,0xD,0x60,0x8A,0x4,0x2C,0x3E,0x92,0xE5,0xAF,0x53,0x7,0xE0,0x29,0xA6,0xC5,0xE3,0xF5,0xF7,0x4A,0x41,0x26,0x6A,0x16,0x5E,0x52,0x2D,0x21,0xAD,0xF0,0x91,0xFF,0xEA,0x54,0xFA,0x66,0x1A,0x45,0x39,0xCF,0x75,0xA4,0x88,0xFB,0x5D,
0xA2,0xA0};
float g[512+2][3]= {
0.33783,0.715698,-0.611206,-0.944031,-0.326599,-0.045624,-0.101074,-0.416443,-0.903503,0.799286,0.49411,-0.341949,-0.854645,0.518036,0.033936,0.42514,-0.437866,-0.792114,-0.358948,0.597046,0.717377,-0.985413,0.144714,0.089294,-0.601776,-0.33728,-0.723907,-0.449921,0.594513,0.666382,0.208313,-0.10791,
0.972076,0.575317,0.060425,0.815643,0.293365,-0.875702,-0.383453,0.293762,0.465759,0.834686,-0.846008,-0.233398,-0.47934,-0.115814,0.143036,-0.98291,0.204681,-0.949036,-0.239532,0.946716,-0.263947,0.184326,-0.235596,0.573822,0.784332,0.203705,-0.372253,-0.905487,0.756989,-0.651031,0.055298,0.497803,
0.814697,-0.297363,-0.16214,0.063995,-0.98468,-0.329254,0.834381,0.441925,0.703827,-0.527039,-0.476227,0.956421,0.266113,0.119781,0.480133,0.482849,0.7323,-0.18631,0.961212,-0.203125,-0.748474,-0.656921,-0.090393,-0.085052,-0.165253,0.982544,-0.76947,0.628174,-0.115234,0.383148,0.537659,0.751068,
0.616486,-0.668488,-0.415924,-0.259979,-0.630005,0.73175,0.570953,-0.087952,0.816223,-0.458008,0.023254,0.888611,-0.196167,0.976563,-0.088287,-0.263885,-0.69812,-0.665527,0.437134,-0.892273,-0.112793,-0.621674,-0.230438,0.748566,0.232422,0.900574,-0.367249,0.22229,-0.796143,0.562744,-0.665497,-0.73764,
0.11377,0.670135,0.704803,0.232605,0.895599,0.429749,-0.114655,-0.11557,-0.474243,0.872742,0.621826,0.604004,-0.498444,-0.832214,0.012756,0.55426,-0.702484,0.705994,-0.089661,-0.692017,0.649292,0.315399,-0.175995,-0.977997,0.111877,0.096954,-0.04953,0.994019,0.635284,-0.606689,-0.477783,-0.261261,
-0.607422,-0.750153,0.983276,0.165436,0.075958,-0.29837,0.404083,-0.864655,-0.638672,0.507721,0.578156,0.388214,0.412079,0.824249,0.556183,-0.208832,0.804352,0.778442,0.562012,0.27951,-0.616577,0.781921,-0.091522,0.196289,0.051056,0.979187,-0.121216,0.207153,-0.970734,-0.173401,-0.384735,0.906555,
0.161499,-0.723236,-0.671387,0.178497,-0.006226,-0.983887,-0.126038,0.15799,0.97934,0.830475,-0.024811,0.556458,-0.510132,-0.76944,0.384247,0.81424,0.200104,-0.544891,-0.112549,-0.393311,-0.912445,0.56189,0.152222,-0.813049,0.198914,-0.254517,-0.946381,-0.41217,0.690979,-0.593811,-0.407257,0.324524,
0.853668,-0.690186,0.366119,-0.624115,-0.428345,0.844147,-0.322296,-0.21228,-0.297546,-0.930756,-0.273071,0.516113,0.811798,0.928314,0.371643,0.007233,0.785828,-0.479218,-0.390778,-0.704895,0.058929,0.706818,0.173248,0.203583,0.963562,0.422211,-0.904297,-0.062469,-0.363312,-0.182465,0.913605,0.254028,
-0.552307,-0.793945,-0.28891,-0.765747,-0.574554,0.058319,0.291382,0.954803,0.946136,-0.303925,0.111267,-0.078156,0.443695,-0.892731,0.182098,0.89389,0.409515,-0.680298,-0.213318,0.701141,0.062469,0.848389,-0.525635,-0.72879,-0.641846,0.238342,-0.88089,0.427673,0.202637,-0.532501,-0.21405,0.818878,
0.948975,-0.305084,0.07962,0.925446,0.374664,0.055817,0.820923,0.565491,0.079102,0.25882,0.099792,-0.960724,-0.294617,0.910522,0.289978,0.137115,0.320038,-0.937408,-0.908386,0.345276,-0.235718,-0.936218,0.138763,0.322754,0.366577,0.925934,-0.090637,0.309296,-0.686829,-0.657684,0.66983,0.024445,
0.742065,-0.917999,-0.059113,-0.392059,0.365509,0.462158,-0.807922,0.083374,0.996399,-0.014801,0.593842,0.253143,-0.763672,0.974976,-0.165466,0.148285,0.918976,0.137299,0.369537,0.294952,0.694977,0.655731,0.943085,0.152618,-0.295319,0.58783,-0.598236,0.544495,0.203796,0.678223,0.705994,-0.478821,
-0.661011,0.577667,0.719055,-0.1698,-0.673828,-0.132172,-0.965332,0.225006,-0.981873,-0.14502,0.121979,0.763458,0.579742,0.284546,-0.893188,0.079681,0.442474,-0.795776,-0.523804,0.303802,0.734955,0.67804,-0.007446,0.15506,0.986267,-0.056183,0.258026,0.571503,-0.778931,-0.681549,-0.702087,-0.206116,
-0.96286,-0.177185,0.203613,-0.470978,-0.515106,0.716095,-0.740326,0.57135,0.354095,-0.56012,-0.824982,-0.074982,-0.507874,0.753204,0.417969,-0.503113,0.038147,0.863342,0.594025,0.673553,-0.439758,-0.119873,-0.005524,-0.992737,0.098267,-0.213776,0.971893,-0.615631,0.643951,0.454163,0.896851,-0.441071,
0.032166,-0.555023,0.750763,-0.358093,0.398773,0.304688,0.864929,-0.722961,0.303589,0.620544,-0.63559,-0.621948,-0.457306,-0.293243,0.072327,0.953278,-0.491638,0.661041,-0.566772,-0.304199,-0.572083,-0.761688,0.908081,-0.398956,0.127014,-0.523621,-0.549683,-0.650848,-0.932922,-0.19986,0.299408,0.099426,
0.140869,0.984985,-0.020325,-0.999756,-0.002319,0.952667,0.280853,-0.11615,-0.971893,0.082581,0.220337,0.65921,0.705292,-0.260651,0.733063,-0.175537,0.657043,-0.555206,0.429504,-0.712189,0.400421,-0.89859,0.179352,0.750885,-0.19696,0.630341,0.785675,-0.569336,0.241821,-0.058899,-0.464111,0.883789,
0.129608,-0.94519,0.299622,-0.357819,0.907654,0.219238,-0.842133,-0.439117,-0.312927,-0.313477,0.84433,0.434479,-0.241211,0.053253,0.968994,0.063873,0.823273,0.563965,0.476288,0.862152,-0.172516,0.620941,-0.298126,0.724915,0.25238,-0.749359,-0.612122,-0.577545,0.386566,0.718994,-0.406342,-0.737976,
0.538696,0.04718,0.556305,0.82959,-0.802856,0.587463,0.101166,-0.707733,-0.705963,0.026428,0.374908,0.68457,0.625092,0.472137,0.208405,-0.856506,-0.703064,-0.581085,-0.409821,-0.417206,-0.736328,0.532623,-0.447876,-0.20285,-0.870728,0.086945,-0.990417,0.107086,0.183685,0.018341,-0.982788,0.560638,
-0.428864,0.708282,0.296722,-0.952576,-0.0672,0.135773,0.990265,0.030243,-0.068787,0.654724,0.752686,0.762604,-0.551758,0.337585,-0.819611,-0.407684,0.402466,-0.727844,-0.55072,-0.408539,-0.855774,-0.480011,0.19281,0.693176,-0.079285,0.716339,0.226013,0.650116,-0.725433,0.246704,0.953369,-0.173553,
-0.970398,-0.239227,-0.03244,0.136383,-0.394318,0.908752,0.813232,0.558167,0.164368,0.40451,0.549042,-0.731323,-0.380249,-0.566711,0.730865,0.022156,0.932739,0.359741,0.00824,0.996552,-0.082306,0.956635,-0.065338,-0.283722,-0.743561,0.008209,0.668579,-0.859589,-0.509674,0.035767,-0.852234,0.363678,
-0.375977,-0.201965,-0.970795,-0.12915,0.313477,0.947327,0.06546,-0.254028,-0.528259,0.81015,0.628052,0.601105,0.49411,-0.494385,0.868378,0.037933,0.275635,-0.086426,0.957336,-0.197937,0.468903,-0.860748,0.895599,0.399384,0.195801,0.560791,0.825012,-0.069214,0.304199,-0.849487,0.43103,0.096375,
0.93576,0.339111,-0.051422,0.408966,-0.911072,0.330444,0.942841,-0.042389,-0.452362,-0.786407,0.420563,0.134308,-0.933472,-0.332489,0.80191,-0.566711,-0.188934,-0.987946,-0.105988,0.112518,-0.24408,0.892242,-0.379791,-0.920502,0.229095,-0.316376,0.7789,0.325958,0.535706,-0.912872,0.185211,-0.36377,
-0.184784,0.565369,-0.803833,-0.018463,0.119537,0.992615,-0.259247,-0.935608,0.239532,-0.82373,-0.449127,-0.345947,-0.433105,0.659515,0.614349,-0.822754,0.378845,-0.423676,0.687195,-0.674835,-0.26889,-0.246582,-0.800842,0.545715,-0.729187,-0.207794,0.651978,0.653534,-0.610443,-0.447388,0.492584,-0.023346,
0.869934,0.609039,0.009094,-0.79306,0.962494,-0.271088,-0.00885,0.2659,-0.004913,0.963959,0.651245,0.553619,-0.518951,0.280548,-0.84314,0.458618,-0.175293,-0.983215,0.049805,0.035339,-0.979919,0.196045,-0.982941,0.164307,-0.082245,0.233734,-0.97226,-0.005005,-0.747253,-0.611328,0.260437,0.645599,
0.592773,0.481384,0.117706,-0.949524,-0.29068,-0.535004,-0.791901,-0.294312,-0.627167,-0.214447,0.748718,-0.047974,-0.813477,-0.57959,-0.175537,0.477264,-0.860992,0.738556,-0.414246,-0.53183,0.562561,-0.704071,0.433289,-0.754944,0.64801,-0.100586,0.114716,0.044525,-0.992371,0.966003,0.244873,-0.082764,
0.33783,0.715698,-0.611206,-0.944031,-0.326599,-0.045624,-0.101074,-0.416443,-0.903503,0.799286,0.49411,-0.341949,-0.854645,0.518036,0.033936,0.42514,-0.437866,-0.792114,-0.358948,0.597046,0.717377,-0.985413,0.144714,0.089294,-0.601776,-0.33728,-0.723907,-0.449921,0.594513,0.666382,0.208313,-0.10791,
0.972076,0.575317,0.060425,0.815643,0.293365,-0.875702,-0.383453,0.293762,0.465759,0.834686,-0.846008,-0.233398,-0.47934,-0.115814,0.143036,-0.98291,0.204681,-0.949036,-0.239532,0.946716,-0.263947,0.184326,-0.235596,0.573822,0.784332,0.203705,-0.372253,-0.905487,0.756989,-0.651031,0.055298,0.497803,
0.814697,-0.297363,-0.16214,0.063995,-0.98468,-0.329254,0.834381,0.441925,0.703827,-0.527039,-0.476227,0.956421,0.266113,0.119781,0.480133,0.482849,0.7323,-0.18631,0.961212,-0.203125,-0.748474,-0.656921,-0.090393,-0.085052,-0.165253,0.982544,-0.76947,0.628174,-0.115234,0.383148,0.537659,0.751068,
0.616486,-0.668488,-0.415924,-0.259979,-0.630005,0.73175,0.570953,-0.087952,0.816223,-0.458008,0.023254,0.888611,-0.196167,0.976563,-0.088287,-0.263885,-0.69812,-0.665527,0.437134,-0.892273,-0.112793,-0.621674,-0.230438,0.748566,0.232422,0.900574,-0.367249,0.22229,-0.796143,0.562744,-0.665497,-0.73764,
0.11377,0.670135,0.704803,0.232605,0.895599,0.429749,-0.114655,-0.11557,-0.474243,0.872742,0.621826,0.604004,-0.498444,-0.832214,0.012756,0.55426,-0.702484,0.705994,-0.089661,-0.692017,0.649292,0.315399,-0.175995,-0.977997,0.111877,0.096954,-0.04953,0.994019,0.635284,-0.606689,-0.477783,-0.261261,
-0.607422,-0.750153,0.983276,0.165436,0.075958,-0.29837,0.404083,-0.864655,-0.638672,0.507721,0.578156,0.388214,0.412079,0.824249,0.556183,-0.208832,0.804352,0.778442,0.562012,0.27951,-0.616577,0.781921,-0.091522,0.196289,0.051056,0.979187,-0.121216,0.207153,-0.970734,-0.173401,-0.384735,0.906555,
0.161499,-0.723236,-0.671387,0.178497,-0.006226,-0.983887,-0.126038,0.15799,0.97934,0.830475,-0.024811,0.556458,-0.510132,-0.76944,0.384247,0.81424,0.200104,-0.544891,-0.112549,-0.393311,-0.912445,0.56189,0.152222,-0.813049,0.198914,-0.254517,-0.946381,-0.41217,0.690979,-0.593811,-0.407257,0.324524,
0.853668,-0.690186,0.366119,-0.624115,-0.428345,0.844147,-0.322296,-0.21228,-0.297546,-0.930756,-0.273071,0.516113,0.811798,0.928314,0.371643,0.007233,0.785828,-0.479218,-0.390778,-0.704895,0.058929,0.706818,0.173248,0.203583,0.963562,0.422211,-0.904297,-0.062469,-0.363312,-0.182465,0.913605,0.254028,
-0.552307,-0.793945,-0.28891,-0.765747,-0.574554,0.058319,0.291382,0.954803,0.946136,-0.303925,0.111267,-0.078156,0.443695,-0.892731,0.182098,0.89389,0.409515,-0.680298,-0.213318,0.701141,0.062469,0.848389,-0.525635,-0.72879,-0.641846,0.238342,-0.88089,0.427673,0.202637,-0.532501,-0.21405,0.818878,
0.948975,-0.305084,0.07962,0.925446,0.374664,0.055817,0.820923,0.565491,0.079102,0.25882,0.099792,-0.960724,-0.294617,0.910522,0.289978,0.137115,0.320038,-0.937408,-0.908386,0.345276,-0.235718,-0.936218,0.138763,0.322754,0.366577,0.925934,-0.090637,0.309296,-0.686829,-0.657684,0.66983,0.024445,
0.742065,-0.917999,-0.059113,-0.392059,0.365509,0.462158,-0.807922,0.083374,0.996399,-0.014801,0.593842,0.253143,-0.763672,0.974976,-0.165466,0.148285,0.918976,0.137299,0.369537,0.294952,0.694977,0.655731,0.943085,0.152618,-0.295319,0.58783,-0.598236,0.544495,0.203796,0.678223,0.705994,-0.478821,
-0.661011,0.577667,0.719055,-0.1698,-0.673828,-0.132172,-0.965332,0.225006,-0.981873,-0.14502,0.121979,0.763458,0.579742,0.284546,-0.893188,0.079681,0.442474,-0.795776,-0.523804,0.303802,0.734955,0.67804,-0.007446,0.15506,0.986267,-0.056183,0.258026,0.571503,-0.778931,-0.681549,-0.702087,-0.206116,
-0.96286,-0.177185,0.203613,-0.470978,-0.515106,0.716095,-0.740326,0.57135,0.354095,-0.56012,-0.824982,-0.074982,-0.507874,0.753204,0.417969,-0.503113,0.038147,0.863342,0.594025,0.673553,-0.439758,-0.119873,-0.005524,-0.992737,0.098267,-0.213776,0.971893,-0.615631,0.643951,0.454163,0.896851,-0.441071,
0.032166,-0.555023,0.750763,-0.358093,0.398773,0.304688,0.864929,-0.722961,0.303589,0.620544,-0.63559,-0.621948,-0.457306,-0.293243,0.072327,0.953278,-0.491638,0.661041,-0.566772,-0.304199,-0.572083,-0.761688,0.908081,-0.398956,0.127014,-0.523621,-0.549683,-0.650848,-0.932922,-0.19986,0.299408,0.099426,
0.140869,0.984985,-0.020325,-0.999756,-0.002319,0.952667,0.280853,-0.11615,-0.971893,0.082581,0.220337,0.65921,0.705292,-0.260651,0.733063,-0.175537,0.657043,-0.555206,0.429504,-0.712189,0.400421,-0.89859,0.179352,0.750885,-0.19696,0.630341,0.785675,-0.569336,0.241821,-0.058899,-0.464111,0.883789,
0.129608,-0.94519,0.299622,-0.357819,0.907654,0.219238,-0.842133,-0.439117,-0.312927,-0.313477,0.84433,0.434479,-0.241211,0.053253,0.968994,0.063873,0.823273,0.563965,0.476288,0.862152,-0.172516,0.620941,-0.298126,0.724915,0.25238,-0.749359,-0.612122,-0.577545,0.386566,0.718994,-0.406342,-0.737976,
0.538696,0.04718,0.556305,0.82959,-0.802856,0.587463,0.101166,-0.707733,-0.705963,0.026428,0.374908,0.68457,0.625092,0.472137,0.208405,-0.856506,-0.703064,-0.581085,-0.409821,-0.417206,-0.736328,0.532623,-0.447876,-0.20285,-0.870728,0.086945,-0.990417,0.107086,0.183685,0.018341,-0.982788,0.560638,
-0.428864,0.708282,0.296722,-0.952576,-0.0672,0.135773,0.990265,0.030243,-0.068787,0.654724,0.752686,0.762604,-0.551758,0.337585,-0.819611,-0.407684,0.402466,-0.727844,-0.55072,-0.408539,-0.855774,-0.480011,0.19281,0.693176,-0.079285,0.716339,0.226013,0.650116,-0.725433,0.246704,0.953369,-0.173553,
-0.970398,-0.239227,-0.03244,0.136383,-0.394318,0.908752,0.813232,0.558167,0.164368,0.40451,0.549042,-0.731323,-0.380249,-0.566711,0.730865,0.022156,0.932739,0.359741,0.00824,0.996552,-0.082306,0.956635,-0.065338,-0.283722,-0.743561,0.008209,0.668579,-0.859589,-0.509674,0.035767,-0.852234,0.363678,
-0.375977,-0.201965,-0.970795,-0.12915,0.313477,0.947327,0.06546,-0.254028,-0.528259,0.81015,0.628052,0.601105,0.49411,-0.494385,0.868378,0.037933,0.275635,-0.086426,0.957336,-0.197937,0.468903,-0.860748,0.895599,0.399384,0.195801,0.560791,0.825012,-0.069214,0.304199,-0.849487,0.43103,0.096375,
0.93576,0.339111,-0.051422,0.408966,-0.911072,0.330444,0.942841,-0.042389,-0.452362,-0.786407,0.420563,0.134308,-0.933472,-0.332489,0.80191,-0.566711,-0.188934,-0.987946,-0.105988,0.112518,-0.24408,0.892242,-0.379791,-0.920502,0.229095,-0.316376,0.7789,0.325958,0.535706,-0.912872,0.185211,-0.36377,
-0.184784,0.565369,-0.803833,-0.018463,0.119537,0.992615,-0.259247,-0.935608,0.239532,-0.82373,-0.449127,-0.345947,-0.433105,0.659515,0.614349,-0.822754,0.378845,-0.423676,0.687195,-0.674835,-0.26889,-0.246582,-0.800842,0.545715,-0.729187,-0.207794,0.651978,0.653534,-0.610443,-0.447388,0.492584,-0.023346,
0.869934,0.609039,0.009094,-0.79306,0.962494,-0.271088,-0.00885,0.2659,-0.004913,0.963959,0.651245,0.553619,-0.518951,0.280548,-0.84314,0.458618,-0.175293,-0.983215,0.049805,0.035339,-0.979919,0.196045,-0.982941,0.164307,-0.082245,0.233734,-0.97226,-0.005005,-0.747253,-0.611328,0.260437,0.645599,
0.592773,0.481384,0.117706,-0.949524,-0.29068,-0.535004,-0.791901,-0.294312,-0.627167,-0.214447,0.748718,-0.047974,-0.813477,-0.57959,-0.175537,0.477264,-0.860992,0.738556,-0.414246,-0.53183,0.562561,-0.704071,0.433289,-0.754944,0.64801,-0.100586,0.114716,0.044525,-0.992371,0.966003,0.244873,-0.082764,
0.33783,0.715698,-0.611206,-0.944031,-0.326599,-0.045624};
#define DOT(a,b) (a[0] * b[0] + a[1] * b[1] + a[2] * b[2])
#define setup(i,b0,b1,r0,r1) \
t = vec[i] + 10000.; \
b0 = ((int)t) & 255; \
b1 = (b0+1) & 255; \
r0 = t - (int)t; \
r1 = r0 - 1.;
float noise3_perlin(float vec[3])
{
int bx0, bx1, by0, by1, bz0, bz1, b00, b10, b01, b11;
float rx0, rx1, ry0, ry1, rz0, rz1, *q, sx, sy, sz, a, b, c, d, t, u, v;
register int i, j;
setup(0, bx0,bx1, rx0,rx1);
setup(1, by0,by1, ry0,ry1);
setup(2, bz0,bz1, rz0,rz1);
i = p[ bx0 ];
j = p[ bx1 ];
b00 = p[ i + by0 ];
b10 = p[ j + by0 ];
b01 = p[ i + by1 ];
b11 = p[ j + by1 ];
#define at(rx,ry,rz) ( rx * q[0] + ry * q[1] + rz * q[2] )
#define surve(t) ( t * t * (3. - 2. * t) )
/* lerp moved to improved perlin above */
sx = surve(rx0);
sy = surve(ry0);
sz = surve(rz0);
q = g[ b00 + bz0 ] ;
u = at(rx0,ry0,rz0);
q = g[ b10 + bz0 ] ;
v = at(rx1,ry0,rz0);
a = lerp(sx, u, v);
q = g[ b01 + bz0 ] ;
u = at(rx0,ry1,rz0);
q = g[ b11 + bz0 ] ;
v = at(rx1,ry1,rz0);
b = lerp(sx, u, v);
c = lerp(sy, a, b); /* interpolate in y at lo x */
q = g[ b00 + bz1 ] ;
u = at(rx0,ry0,rz1);
q = g[ b10 + bz1 ] ;
v = at(rx1,ry0,rz1);
a = lerp(sx, u, v);
q = g[ b01 + bz1 ] ;
u = at(rx0,ry1,rz1);
q = g[ b11 + bz1 ] ;
v = at(rx1,ry1,rz1);
b = lerp(sx, u, v);
d = lerp(sy, a, b); /* interpolate in y at hi x */
return 1.5 * lerp(sz, c, d); /* interpolate in z */
}
float turbulence_perlin(float *point, float lofreq, float hifreq)
{
float freq, t, p[3];
p[0] = point[0] + 123.456;
p[1] = point[1];
p[2] = point[2];
t = 0;
for (freq = lofreq ; freq < hifreq ; freq *= 2.) {
t += fabs(noise3_perlin(p)) / freq;
p[0] *= 2.;
p[1] *= 2.;
p[2] *= 2.;
}
return t - 0.3; /* readjust to make mean value = 0.0 */
}
/* for use with BLI_gNoise/gTurbulence, returns signed noise */
static float orgPerlinNoise(float x, float y, float z)
{
float v[3];
v[0] = x;
v[1] = y;
v[2] = z;
return noise3_perlin(v);
}
/* for use with BLI_gNoise/gTurbulence, returns unsigned noise */
static float orgPerlinNoiseU(float x, float y, float z)
{
float v[3];
v[0] = x;
v[1] = y;
v[2] = z;
return (0.5+0.5*noise3_perlin(v));
}
/* *************** CALL AS: *************** */
float BLI_hnoisep(float noisesize, float x, float y, float z)
{
float vec[3];
vec[0]= x/noisesize;
vec[1]= y/noisesize;
vec[2]= z/noisesize;
return noise3_perlin(vec);
}
float turbulencep(float noisesize, float x, float y, float z, int nr)
{
float vec[3];
vec[0]= x/noisesize;
vec[1]= y/noisesize;
vec[2]= z/noisesize;
nr++;
return turbulence_perlin(vec, 1.0, (float)(1<<nr));
}
/******************/
/* VORONOI/WORLEY */
/******************/
/* distance metrics for voronoi, e parameter only used in Minkovsky */
/* Camberra omitted, didn't seem useful */
/* distance squared */
static float dist_Squared(float x, float y, float z, float e) { return (x*x + y*y + z*z); }
/* real distance */
static float dist_Real(float x, float y, float z, float e) { return sqrt(x*x + y*y + z*z); }
/* manhattan/taxicab/cityblock distance */
static float dist_Manhattan(float x, float y, float z, float e) { return (fabs(x) + fabs(y) + fabs(z)); }
/* Chebychev */
static float dist_Chebychev(float x, float y, float z, float e)
{
float t;
x = fabs(x);
y = fabs(y);
z = fabs(z);
t = (x>y)?x:y;
return ((z>t)?z:t);
}
/* minkovsky preset exponent 0.5 */
static float dist_MinkovskyH(float x, float y, float z, float e)
{
float d = sqrt(fabs(x)) + sqrt(fabs(y)) + sqrt(fabs(z));
return (d*d);
}
/* minkovsky preset exponent 4 */
static float dist_Minkovsky4(float x, float y, float z, float e)
{
x *= x;
y *= y;
z *= z;
return sqrt(sqrt(x*x + y*y + z*z));
}
/* Minkovsky, general case, slow, maybe too slow to be useful */
static float dist_Minkovsky(float x, float y, float z, float e)
{
return pow(pow(fabs(x), e) + pow(fabs(y), e) + pow(fabs(z), e), 1.0/e);
}
/* Not 'pure' Worley, but the results are virtually the same.
Returns distances in da and point coords in pa */
void voronoi(float x, float y, float z, float* da, float* pa, float me, int dtype)
{
int xx, yy, zz, xi, yi, zi;
float xd, yd, zd, d, *p;
float (*distfunc)(float, float, float, float);
switch (dtype) {
case 1:
distfunc = dist_Squared;
break;
case 2:
distfunc = dist_Manhattan;
break;
case 3:
distfunc = dist_Chebychev;
break;
case 4:
distfunc = dist_MinkovskyH;
break;
case 5:
distfunc = dist_Minkovsky4;
break;
case 6:
distfunc = dist_Minkovsky;
break;
case 0:
default:
distfunc = dist_Real;
}
xi = (int)(floor(x));
yi = (int)(floor(y));
zi = (int)(floor(z));
da[0] = da[1] = da[2] = da[3] = 1e10f;
for (xx=xi-1;xx<=xi+1;xx++) {
for (yy=yi-1;yy<=yi+1;yy++) {
for (zz=zi-1;zz<=zi+1;zz++) {
p = HASHPNT(xx, yy, zz);
xd = x - (p[0] + xx);
yd = y - (p[1] + yy);
zd = z - (p[2] + zz);
d = distfunc(xd, yd, zd, me);
if (d<da[0]) {
da[3]=da[2]; da[2]=da[1]; da[1]=da[0]; da[0]=d;
pa[9]=pa[6]; pa[10]=pa[7]; pa[11]=pa[8];
pa[6]=pa[3]; pa[7]=pa[4]; pa[8]=pa[5];
pa[3]=pa[0]; pa[4]=pa[1]; pa[5]=pa[2];
pa[0]=p[0]+xx; pa[1]=p[1]+yy; pa[2]=p[2]+zz;
}
else if (d<da[1]) {
da[3]=da[2]; da[2]=da[1]; da[1]=d;
pa[9]=pa[6]; pa[10]=pa[7]; pa[11]=pa[8];
pa[6]=pa[3]; pa[7]=pa[4]; pa[8]=pa[5];
pa[3]=p[0]+xx; pa[4]=p[1]+yy; pa[5]=p[2]+zz;
}
else if (d<da[2]) {
da[3]=da[2]; da[2]=d;
pa[9]=pa[6]; pa[10]=pa[7]; pa[11]=pa[8];
pa[6]=p[0]+xx; pa[7]=p[1]+yy; pa[8]=p[2]+zz;
}
else if (d<da[3]) {
da[3]=d;
pa[9]=p[0]+xx; pa[10]=p[1]+yy; pa[11]=p[2]+zz;
}
}
}
}
}
/* returns different feature points for use in BLI_gNoise() */
static float voronoi_F1(float x, float y, float z)
{
float da[4], pa[12];
voronoi(x, y, z, da, pa, 1, 0);
return da[0];
}
static float voronoi_F2(float x, float y, float z)
{
float da[4], pa[12];
voronoi(x, y, z, da, pa, 1, 0);
return da[1];
}
static float voronoi_F3(float x, float y, float z)
{
float da[4], pa[12];
voronoi(x, y, z, da, pa, 1, 0);
return da[2];
}
static float voronoi_F4(float x, float y, float z)
{
float da[4], pa[12];
voronoi(x, y, z, da, pa, 1, 0);
return da[3];
}
static float voronoi_F1F2(float x, float y, float z)
{
float da[4], pa[12];
voronoi(x, y, z, da, pa, 1, 0);
return (da[1]-da[0]);
}
/* Crackle type pattern, just a scale/clamp of F2-F1 */
static float voronoi_Cr(float x, float y, float z)
{
float t = 10*voronoi_F1F2(x, y, z);
if (t>1.f) return 1.f;
return t;
}
/* Signed version of all 6 of the above, just 2x-1, not really correct though (range is potentially (0, sqrt(6)).
Used in the musgrave functions */
static float voronoi_F1S(float x, float y, float z)
{
float da[4], pa[12];
voronoi(x, y, z, da, pa, 1, 0);
return (2.0*da[0]-1.0);
}
static float voronoi_F2S(float x, float y, float z)
{
float da[4], pa[12];
voronoi(x, y, z, da, pa, 1, 0);
return (2.0*da[1]-1.0);
}
static float voronoi_F3S(float x, float y, float z)
{
float da[4], pa[12];
voronoi(x, y, z, da, pa, 1, 0);
return (2.0*da[2]-1.0);
}
static float voronoi_F4S(float x, float y, float z)
{
float da[4], pa[12];
voronoi(x, y, z, da, pa, 1, 0);
return (2.0*da[3]-1.0);
}
static float voronoi_F1F2S(float x, float y, float z)
{
float da[4], pa[12];
voronoi(x, y, z, da, pa, 1, 0);
return (2.0*(da[1]-da[0])-1.0);
}
/* Crackle type pattern, just a scale/clamp of F2-F1 */
static float voronoi_CrS(float x, float y, float z)
{
float t = 10*voronoi_F1F2(x, y, z);
if (t>1.f) return 1.f;
return (2.0*t-1.0);
}
/***************/
/* voronoi end */
/***************/
/*************/
/* CELLNOISE */
/*************/
/* returns unsigned cellnoise */
static float cellNoiseU(float x, float y, float z)
{
int xi = (int)(floor(x));
int yi = (int)(floor(y));
int zi = (int)(floor(z));
unsigned int n = xi + yi*1301 + zi*314159;
n ^= (n<<13);
return ((float)(n*(n*n*15731 + 789221) + 1376312589) / 4294967296.0);
}
/* idem, signed */
float cellNoise(float x, float y, float z)
{
return (2.0*cellNoiseU(x, y, z)-1.0);
}
/* returns a vector/point/color in ca, using point hasharray directly */
void cellNoiseV(float x, float y, float z, float *ca)
{
int xi = (int)(floor(x));
int yi = (int)(floor(y));
int zi = (int)(floor(z));
float *p = HASHPNT(xi, yi, zi);
ca[0] = p[0];
ca[1] = p[1];
ca[2] = p[2];
}
/*****************/
/* end cellnoise */
/*****************/
/* newnoise: generic noise function for use with different noisebases */
float BLI_gNoise(float noisesize, float x, float y, float z, int hard, int noisebasis)
{
float (*noisefunc)(float, float, float);
switch (noisebasis) {
case 1:
noisefunc = orgPerlinNoiseU;
break;
case 2:
noisefunc = newPerlinU;
break;
case 3:
noisefunc = voronoi_F1;
break;
case 4:
noisefunc = voronoi_F2;
break;
case 5:
noisefunc = voronoi_F3;
break;
case 6:
noisefunc = voronoi_F4;
break;
case 7:
noisefunc = voronoi_F1F2;
break;
case 8:
noisefunc = voronoi_Cr;
break;
case 14:
noisefunc = cellNoiseU;
break;
case 0:
default: {
noisefunc = orgBlenderNoise;
/* add one to make return value same as BLI_hnoise */
x += 1;
y += 1;
z += 1;
}
}
if (noisesize!=0.0) {
noisesize = 1.0/noisesize;
x *= noisesize;
y *= noisesize;
z *= noisesize;
}
if (hard) return fabs(2.0*noisefunc(x, y, z)-1.0);
return noisefunc(x, y, z);
}
/* newnoise: generic turbulence function for use with different noisebasis */
float BLI_gTurbulence(float noisesize, float x, float y, float z, int oct, int hard, int noisebasis)
{
float (*noisefunc)(float, float, float);
float sum, t, amp=1, fscale=1;
int i;
switch (noisebasis) {
case 1:
noisefunc = orgPerlinNoiseU;
break;
case 2:
noisefunc = newPerlinU;
break;
case 3:
noisefunc = voronoi_F1;
break;
case 4:
noisefunc = voronoi_F2;
break;
case 5:
noisefunc = voronoi_F3;
break;
case 6:
noisefunc = voronoi_F4;
break;
case 7:
noisefunc = voronoi_F1F2;
break;
case 8:
noisefunc = voronoi_Cr;
break;
case 14:
noisefunc = cellNoiseU;
break;
case 0:
default:
noisefunc = orgBlenderNoise;
x += 1;
y += 1;
z += 1;
}
if (noisesize!=0.0) {
noisesize = 1.0/noisesize;
x *= noisesize;
y *= noisesize;
z *= noisesize;
}
sum = 0;
for (i=0;i<=oct;i++, amp*=0.5, fscale*=2) {
t = noisefunc(fscale*x, fscale*y, fscale*z);
if (hard) t = fabs(2.0*t-1.0);
sum += t * amp;
}
sum *= ((float)(1<<oct)/(float)((1<<(oct+1))-1));
return sum;
}
/*************************************/
/* NOISE FUNCTIONS BY KEN MUSGRAVE */
/* Copyright 1994 F. Kenton Musgrave */
/*************************************/
/* All of these are modified to be able to use them with different noisebasis.
In some cases the original code seemed to contain errors, so it is not exactly
the same now as the orginal code (from "Texturing and Modelling: A procedural approach") */
/*
* Procedural fBm evaluated at "point"; returns value stored in "value".
*
* Copyright 1994 F. Kenton Musgrave
*
* Parameters:
* ``H'' is the fractal increment parameter
* ``lacunarity'' is the gap between successive frequencies
* ``octaves'' is the number of frequencies in the fBm
*/
float mg_fBm(float x, float y, float z, float H, float lacunarity, float octaves, int noisebasis)
{
float rmd, value=0.0, pwr=1.0, pwHL=pow(lacunarity, -H);
int i;
float (*noisefunc)(float, float, float);
switch (noisebasis) {
case 1:
noisefunc = orgPerlinNoise;
break;
case 2:
noisefunc = newPerlin;
break;
case 3:
noisefunc = voronoi_F1S;
break;
case 4:
noisefunc = voronoi_F2S;
break;
case 5:
noisefunc = voronoi_F3S;
break;
case 6:
noisefunc = voronoi_F4S;
break;
case 7:
noisefunc = voronoi_F1F2S;
break;
case 8:
noisefunc = voronoi_CrS;
break;
case 14:
noisefunc = cellNoise;
break;
case 0:
default: {
noisefunc = orgBlenderNoiseS;
}
}
for (i=0; i<(int)octaves; i++) {
value += noisefunc(x, y, z) * pwr;
pwr *= pwHL;
x *= lacunarity;
y *= lacunarity;
z *= lacunarity;
}
rmd = octaves - floor(octaves);
if (rmd!=0.f) value += rmd * noisefunc(x, y, z) * pwr;
return value;
} /* fBm() */
/*
* Procedural multifractal evaluated at "point";
* returns value stored in "value".
*
* Copyright 1994 F. Kenton Musgrave
*
* Parameters:
* ``H'' determines the highest fractal dimension
* ``lacunarity'' is gap between successive frequencies
* ``octaves'' is the number of frequencies in the fBm
* ``offset'' is the zero offset, which determines multifractality (NOT USED??)
*/
/* this one is in fact rather confusing,
* there seem to be errors in the original source code (in all three versions of proc.text&mod),
* I modified it to something that made sense to me, so it might be wrong... */
float mg_MultiFractal(float x, float y, float z, float H, float lacunarity, float octaves, int noisebasis)
{
float rmd, value=1.0, pwr=1.0, pwHL=pow(lacunarity, -H);
int i;
float (*noisefunc)(float, float, float);
switch (noisebasis) {
case 1:
noisefunc = orgPerlinNoise;
break;
case 2:
noisefunc = newPerlin;
break;
case 3:
noisefunc = voronoi_F1S;
break;
case 4:
noisefunc = voronoi_F2S;
break;
case 5:
noisefunc = voronoi_F3S;
break;
case 6:
noisefunc = voronoi_F4S;
break;
case 7:
noisefunc = voronoi_F1F2S;
break;
case 8:
noisefunc = voronoi_CrS;
break;
case 14:
noisefunc = cellNoise;
break;
case 0:
default: {
noisefunc = orgBlenderNoiseS;
}
}
for (i=0; i<(int)octaves; i++) {
value *= (pwr * noisefunc(x, y, z) + 1.0);
pwr *= pwHL;
x *= lacunarity;
y *= lacunarity;
z *= lacunarity;
}
rmd = octaves - floor(octaves);
if (rmd!=0.0) value *= (rmd * noisefunc(x, y, z) * pwr + 1.0);
return value;
} /* multifractal() */
/*
* Heterogeneous procedural terrain function: stats by altitude method.
* Evaluated at "point"; returns value stored in "value".
*
* Copyright 1994 F. Kenton Musgrave
*
* Parameters:
* ``H'' determines the fractal dimension of the roughest areas
* ``lacunarity'' is the gap between successive frequencies
* ``octaves'' is the number of frequencies in the fBm
* ``offset'' raises the terrain from `sea level'
*/
float mg_HeteroTerrain(float x, float y, float z, float H, float lacunarity, float octaves, float offset, int noisebasis)
{
float value, increment, rmd;
int i;
float pwHL = pow(lacunarity, -H);
float pwr = pwHL; /* starts with i=1 instead of 0 */
float (*noisefunc)(float, float, float);
switch (noisebasis) {
case 1:
noisefunc = orgPerlinNoise;
break;
case 2:
noisefunc = newPerlin;
break;
case 3:
noisefunc = voronoi_F1S;
break;
case 4:
noisefunc = voronoi_F2S;
break;
case 5:
noisefunc = voronoi_F3S;
break;
case 6:
noisefunc = voronoi_F4S;
break;
case 7:
noisefunc = voronoi_F1F2S;
break;
case 8:
noisefunc = voronoi_CrS;
break;
case 14:
noisefunc = cellNoise;
break;
case 0:
default: {
noisefunc = orgBlenderNoiseS;
}
}
/* first unscaled octave of function; later octaves are scaled */
value = offset + noisefunc(x, y, z);
x *= lacunarity;
y *= lacunarity;
z *= lacunarity;
for (i=1; i<(int)octaves; i++) {
increment = (noisefunc(x, y, z) + offset) * pwr * value;
value += increment;
pwr *= pwHL;
x *= lacunarity;
y *= lacunarity;
z *= lacunarity;
}
rmd = octaves - floor(octaves);
if (rmd!=0.0) {
increment = (noisefunc(x, y, z) + offset) * pwr * value;
value += rmd * increment;
}
return value;
}
/* Hybrid additive/multiplicative multifractal terrain model.
*
* Copyright 1994 F. Kenton Musgrave
*
* Some good parameter values to start with:
*
* H: 0.25
* offset: 0.7
*/
float mg_HybridMultiFractal(float x, float y, float z, float H, float lacunarity, float octaves, float offset, float gain, int noisebasis)
{
float result, signal, weight, rmd;
int i;
float pwHL = pow(lacunarity, -H);
float pwr = pwHL; /* starts with i=1 instead of 0 */
float (*noisefunc)(float, float, float);
switch (noisebasis) {
case 1:
noisefunc = orgPerlinNoise;
break;
case 2:
noisefunc = newPerlin;
break;
case 3:
noisefunc = voronoi_F1S;
break;
case 4:
noisefunc = voronoi_F2S;
break;
case 5:
noisefunc = voronoi_F3S;
break;
case 6:
noisefunc = voronoi_F4S;
break;
case 7:
noisefunc = voronoi_F1F2S;
break;
case 8:
noisefunc = voronoi_CrS;
break;
case 14:
noisefunc = cellNoise;
break;
case 0:
default: {
noisefunc = orgBlenderNoiseS;
}
}
result = noisefunc(x, y, z) + offset;
weight = gain * result;
x *= lacunarity;
y *= lacunarity;
z *= lacunarity;
for (i=1; (weight>0.001) && (i<(int)octaves); i++) {
if (weight>1.0) weight=1.0;
signal = (noisefunc(x, y, z) + offset) * pwr;
pwr *= pwHL;
result += weight * signal;
weight *= gain * signal;
x *= lacunarity;
y *= lacunarity;
z *= lacunarity;
}
rmd = octaves - floor(octaves);
if (rmd!=0.f) result += rmd * ((noisefunc(x, y, z) + offset) * pwr);
return result;
} /* HybridMultifractal() */
/* Ridged multifractal terrain model.
*
* Copyright 1994 F. Kenton Musgrave
*
* Some good parameter values to start with:
*
* H: 1.0
* offset: 1.0
* gain: 2.0
*/
float mg_RidgedMultiFractal(float x, float y, float z, float H, float lacunarity, float octaves, float offset, float gain, int noisebasis)
{
float result, signal, weight;
int i;
float pwHL = pow(lacunarity, -H);
float pwr = pwHL; /* starts with i=1 instead of 0 */
float (*noisefunc)(float, float, float);
switch (noisebasis) {
case 1:
noisefunc = orgPerlinNoise;
break;
case 2:
noisefunc = newPerlin;
break;
case 3:
noisefunc = voronoi_F1S;
break;
case 4:
noisefunc = voronoi_F2S;
break;
case 5:
noisefunc = voronoi_F3S;
break;
case 6:
noisefunc = voronoi_F4S;
break;
case 7:
noisefunc = voronoi_F1F2S;
break;
case 8:
noisefunc = voronoi_CrS;
break;
case 14:
noisefunc = cellNoise;
break;
case 0:
default: {
noisefunc = orgBlenderNoiseS;
}
}
signal = offset - fabs(noisefunc(x, y, z));
signal *= signal;
result = signal;
weight = 1.f;
for( i=1; i<(int)octaves; i++ ) {
x *= lacunarity;
y *= lacunarity;
z *= lacunarity;
weight = signal * gain;
if (weight>1.0) weight=1.0; else if (weight<0.0) weight=0.0;
signal = offset - fabs(noisefunc(x, y, z));
signal *= signal;
signal *= weight;
result += signal * pwr;
pwr *= pwHL;
}
return result;
} /* RidgedMultifractal() */
/* "Variable Lacunarity Noise"
* A distorted variety of Perlin noise.
*
* Copyright 1994 F. Kenton Musgrave
*/
float mg_VLNoise(float x, float y, float z, float distortion, int nbas1, int nbas2)
{
float rv[3];
float (*noisefunc1)(float, float, float);
float (*noisefunc2)(float, float, float);
switch (nbas1) {
case 1:
noisefunc1 = orgPerlinNoise;
break;
case 2:
noisefunc1 = newPerlin;
break;
case 3:
noisefunc1 = voronoi_F1S;
break;
case 4:
noisefunc1 = voronoi_F2S;
break;
case 5:
noisefunc1 = voronoi_F3S;
break;
case 6:
noisefunc1 = voronoi_F4S;
break;
case 7:
noisefunc1 = voronoi_F1F2S;
break;
case 8:
noisefunc1 = voronoi_CrS;
break;
case 14:
noisefunc1 = cellNoise;
break;
case 0:
default: {
noisefunc1 = orgBlenderNoiseS;
}
}
switch (nbas2) {
case 1:
noisefunc2 = orgPerlinNoise;
break;
case 2:
noisefunc2 = newPerlin;
break;
case 3:
noisefunc2 = voronoi_F1S;
break;
case 4:
noisefunc2 = voronoi_F2S;
break;
case 5:
noisefunc2 = voronoi_F3S;
break;
case 6:
noisefunc2 = voronoi_F4S;
break;
case 7:
noisefunc2 = voronoi_F1F2S;
break;
case 8:
noisefunc2 = voronoi_CrS;
break;
case 14:
noisefunc2 = cellNoise;
break;
case 0:
default: {
noisefunc2 = orgBlenderNoiseS;
}
}
/* get a random vector and scale the randomization */
rv[0] = noisefunc1(x+13.5, y+13.5, z+13.5) * distortion;
rv[1] = noisefunc1(x, y, z) * distortion;
rv[2] = noisefunc1(x-13.5, y-13.5, z-13.5) * distortion;
return noisefunc2(x+rv[0], y+rv[1], z+rv[2]); /* distorted-domain noise */
}
/****************/
/* musgrave end */
/****************/