src/third_party/qcms/src/chain.c

   1 /* vim: set ts=8 sw=8 noexpandtab: */
   2 //  qcms
   3 //  Copyright (C) 2009 Mozilla Corporation
   4 //  Copyright (C) 1998-2007 Marti Maria
   5 //
   6 // Permission is hereby granted, free of charge, to any person obtaining
   7 // a copy of this software and associated documentation files (the "Software"),
   8 // to deal in the Software without restriction, including without limitation
   9 // the rights to use, copy, modify, merge, publish, distribute, sublicense,
  10 // and/or sell copies of the Software, and to permit persons to whom the Software
  11 // is furnished to do so, subject to the following conditions:
  12 //
  13 // The above copyright notice and this permission notice shall be included in
  14 // all copies or substantial portions of the Software.
  15 //
  16 // THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  17 // EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO
  18 // THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  19 // NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
  20 // LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
  21 // OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
  22 // WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  23
  24 #include <stdlib.h>
  25 #include <math.h>
  26 #include <assert.h>
  27 #include <string.h> //memcpy
  28 #include "qcmsint.h"
  29 #include "transform_util.h"
  30 #include "matrix.h"
  31
  32 static struct matrix build_lut_matrix(struct lutType *lut)
  33 {
  34         struct matrix result;
  35         if (lut) {
  36                 result.m[0][0] = s15Fixed16Number_to_float(lut->e00);
  37                 result.m[0][1] = s15Fixed16Number_to_float(lut->e01);
  38                 result.m[0][2] = s15Fixed16Number_to_float(lut->e02);
  39                 result.m[1][0] = s15Fixed16Number_to_float(lut->e10);
  40                 result.m[1][1] = s15Fixed16Number_to_float(lut->e11);
  41                 result.m[1][2] = s15Fixed16Number_to_float(lut->e12);
  42                 result.m[2][0] = s15Fixed16Number_to_float(lut->e20);
  43                 result.m[2][1] = s15Fixed16Number_to_float(lut->e21);
  44                 result.m[2][2] = s15Fixed16Number_to_float(lut->e22);
  45                 result.invalid = false;
  46         } else {
  47                 memset(&result, 0, sizeof(struct matrix));
  48                 result.invalid = true;
  49         }
  50         return result;
  51 }
  52
  53 static struct matrix build_mAB_matrix(struct lutmABType *lut)
  54 {
  55         struct matrix result;
  56         if (lut) {
  57                 result.m[0][0] = s15Fixed16Number_to_float(lut->e00);
  58                 result.m[0][1] = s15Fixed16Number_to_float(lut->e01);
  59                 result.m[0][2] = s15Fixed16Number_to_float(lut->e02);
  60                 result.m[1][0] = s15Fixed16Number_to_float(lut->e10);
  61                 result.m[1][1] = s15Fixed16Number_to_float(lut->e11);
  62                 result.m[1][2] = s15Fixed16Number_to_float(lut->e12);
  63                 result.m[2][0] = s15Fixed16Number_to_float(lut->e20);
  64                 result.m[2][1] = s15Fixed16Number_to_float(lut->e21);
  65                 result.m[2][2] = s15Fixed16Number_to_float(lut->e22);
  66                 result.invalid = false;
  67         } else {
  68                 memset(&result, 0, sizeof(struct matrix));
  69                 result.invalid = true;
  70         }
  71         return result;
  72 }
  73
  74 //Based on lcms cmsLab2XYZ
  75 #define f(t) (t <= (24.0f/116.0f)*(24.0f/116.0f)*(24.0f/116.0f)) ? ((841.0/108.0) * t + (16.0/116.0)) : pow(t,1.0/3.0)
  76 #define f_1(t) (t <= (24.0f/116.0f)) ? ((108.0/841.0) * (t - (16.0/116.0))) : (t * t * t)
  77 static void qcms_transform_module_LAB_to_XYZ(struct qcms_modular_transform *transform, float *src, float *dest, size_t length)
  78 {
  79         size_t i;
  80         // lcms: D50 XYZ values
  81         float WhitePointX = 0.9642f;
  82         float WhitePointY = 1.0f;
  83         float WhitePointZ = 0.8249f;
  84         for (i = 0; i < length; i++) {
  85                 float device_L = *src++ * 100.0f;
  86                 float device_a = *src++ * 255.0f - 128.0f;
  87                 float device_b = *src++ * 255.0f - 128.0f;
  88                 float y = (device_L + 16.0f) / 116.0f;
  89
  90                 float X = f_1((y + 0.002f * device_a)) * WhitePointX;
  91                 float Y = f_1(y) * WhitePointY;
  92                 float Z = f_1((y - 0.005f * device_b)) * WhitePointZ;
  93                 *dest++ = X / (1.0 + 32767.0/32768.0);
  94                 *dest++ = Y / (1.0 + 32767.0/32768.0);
  95                 *dest++ = Z / (1.0 + 32767.0/32768.0);
  96         }
  97 }
  98
  99 //Based on lcms cmsXYZ2Lab
 100 static void qcms_transform_module_XYZ_to_LAB(struct qcms_modular_transform *transform, float *src, float *dest, size_t length)
 101 {
 102         size_t i;
 103         // lcms: D50 XYZ values
 104         float WhitePointX = 0.9642f;
 105         float WhitePointY = 1.0f;
 106         float WhitePointZ = 0.8249f;
 107         for (i = 0; i < length; i++) {
 108                 float device_x = *src++ * (1.0 + 32767.0/32768.0) / WhitePointX;
 109                 float device_y = *src++ * (1.0 + 32767.0/32768.0) / WhitePointY;
 110                 float device_z = *src++ * (1.0 + 32767.0/32768.0) / WhitePointZ;
 111
 112                 float fx = f(device_x);
 113                 float fy = f(device_y);
 114                 float fz = f(device_z);
 115
 116                 float L = 116.0f*fy - 16.0f;
 117                 float a = 500.0f*(fx - fy);
 118                 float b = 200.0f*(fy - fz);
 119                 *dest++ = L / 100.0f;
 120                 *dest++ = (a+128.0f) / 255.0f;
 121                 *dest++ = (b+128.0f) / 255.0f;
 122         }
 123
 124 }
 125
 126 static void qcms_transform_module_clut_only(struct qcms_modular_transform *transform, float *src, float *dest, size_t length)
 127 {
 128         size_t i;
 129         int xy_len = 1;
 130         int x_len = transform->grid_size;
 131         int len = x_len * x_len;
 132         float* r_table = transform->r_clut;
 133         float* g_table = transform->g_clut;
 134         float* b_table = transform->b_clut;
 135
 136         for (i = 0; i < length; i++) {
 137                 float linear_r = *src++;
 138                 float linear_g = *src++;
 139                 float linear_b = *src++;
 140
 141                 int x = floor(linear_r * (transform->grid_size-1));
 142                 int y = floor(linear_g * (transform->grid_size-1));
 143                 int z = floor(linear_b * (transform->grid_size-1));
 144                 int x_n = ceil(linear_r * (transform->grid_size-1));
 145                 int y_n = ceil(linear_g * (transform->grid_size-1));
 146                 int z_n = ceil(linear_b * (transform->grid_size-1));
 147                 float x_d = linear_r * (transform->grid_size-1) - x;
 148                 float y_d = linear_g * (transform->grid_size-1) - y;
 149                 float z_d = linear_b * (transform->grid_size-1) - z;
 150
 151                 float r_x1 = lerp(CLU(r_table,x,y,z), CLU(r_table,x_n,y,z), x_d);
 152                 float r_x2 = lerp(CLU(r_table,x,y_n,z), CLU(r_table,x_n,y_n,z), x_d);
 153                 float r_y1 = lerp(r_x1, r_x2, y_d);
 154                 float r_x3 = lerp(CLU(r_table,x,y,z_n), CLU(r_table,x_n,y,z_n), x_d);
 155                 float r_x4 = lerp(CLU(r_table,x,y_n,z_n), CLU(r_table,x_n,y_n,z_n), x_d);
 156                 float r_y2 = lerp(r_x3, r_x4, y_d);
 157                 float clut_r = lerp(r_y1, r_y2, z_d);
 158
 159                 float g_x1 = lerp(CLU(g_table,x,y,z), CLU(g_table,x_n,y,z), x_d);
 160                 float g_x2 = lerp(CLU(g_table,x,y_n,z), CLU(g_table,x_n,y_n,z), x_d);
 161                 float g_y1 = lerp(g_x1, g_x2, y_d);
 162                 float g_x3 = lerp(CLU(g_table,x,y,z_n), CLU(g_table,x_n,y,z_n), x_d);
 163                 float g_x4 = lerp(CLU(g_table,x,y_n,z_n), CLU(g_table,x_n,y_n,z_n), x_d);
 164                 float g_y2 = lerp(g_x3, g_x4, y_d);
 165                 float clut_g = lerp(g_y1, g_y2, z_d);
 166
 167                 float b_x1 = lerp(CLU(b_table,x,y,z), CLU(b_table,x_n,y,z), x_d);
 168                 float b_x2 = lerp(CLU(b_table,x,y_n,z), CLU(b_table,x_n,y_n,z), x_d);
 169                 float b_y1 = lerp(b_x1, b_x2, y_d);
 170                 float b_x3 = lerp(CLU(b_table,x,y,z_n), CLU(b_table,x_n,y,z_n), x_d);
 171                 float b_x4 = lerp(CLU(b_table,x,y_n,z_n), CLU(b_table,x_n,y_n,z_n), x_d);
 172                 float b_y2 = lerp(b_x3, b_x4, y_d);
 173                 float clut_b = lerp(b_y1, b_y2, z_d);
 174
 175                 *dest++ = clamp_float(clut_r);
 176                 *dest++ = clamp_float(clut_g);
 177                 *dest++ = clamp_float(clut_b);
 178         }
 179 }
 180
 181 static void qcms_transform_module_clut(struct qcms_modular_transform *transform, float *src, float *dest, size_t length)
 182 {
 183         size_t i;
 184         int xy_len = 1;
 185         int x_len = transform->grid_size;
 186         int len = x_len * x_len;
 187         float* r_table = transform->r_clut;
 188         float* g_table = transform->g_clut;
 189         float* b_table = transform->b_clut;
 190         for (i = 0; i < length; i++) {
 191                 float device_r = *src++;
 192                 float device_g = *src++;
 193                 float device_b = *src++;
 194                 float linear_r = lut_interp_linear_float(device_r,
 195                                 transform->input_clut_table_r, transform->input_clut_table_length);
 196                 float linear_g = lut_interp_linear_float(device_g,
 197                                 transform->input_clut_table_g, transform->input_clut_table_length);
 198                 float linear_b = lut_interp_linear_float(device_b,
 199                                 transform->input_clut_table_b, transform->input_clut_table_length);
 200
 201                 int x = floor(linear_r * (transform->grid_size-1));
 202                 int y = floor(linear_g * (transform->grid_size-1));
 203                 int z = floor(linear_b * (transform->grid_size-1));
 204                 int x_n = ceil(linear_r * (transform->grid_size-1));
 205                 int y_n = ceil(linear_g * (transform->grid_size-1));
 206                 int z_n = ceil(linear_b * (transform->grid_size-1));
 207                 float x_d = linear_r * (transform->grid_size-1) - x;
 208                 float y_d = linear_g * (transform->grid_size-1) - y;
 209                 float z_d = linear_b * (transform->grid_size-1) - z;
 210
 211                 float r_x1 = lerp(CLU(r_table,x,y,z), CLU(r_table,x_n,y,z), x_d);
 212                 float r_x2 = lerp(CLU(r_table,x,y_n,z), CLU(r_table,x_n,y_n,z), x_d);
 213                 float r_y1 = lerp(r_x1, r_x2, y_d);
 214                 float r_x3 = lerp(CLU(r_table,x,y,z_n), CLU(r_table,x_n,y,z_n), x_d);
 215                 float r_x4 = lerp(CLU(r_table,x,y_n,z_n), CLU(r_table,x_n,y_n,z_n), x_d);
 216                 float r_y2 = lerp(r_x3, r_x4, y_d);
 217                 float clut_r = lerp(r_y1, r_y2, z_d);
 218
 219                 float g_x1 = lerp(CLU(g_table,x,y,z), CLU(g_table,x_n,y,z), x_d);
 220                 float g_x2 = lerp(CLU(g_table,x,y_n,z), CLU(g_table,x_n,y_n,z), x_d);
 221                 float g_y1 = lerp(g_x1, g_x2, y_d);
 222                 float g_x3 = lerp(CLU(g_table,x,y,z_n), CLU(g_table,x_n,y,z_n), x_d);
 223                 float g_x4 = lerp(CLU(g_table,x,y_n,z_n), CLU(g_table,x_n,y_n,z_n), x_d);
 224                 float g_y2 = lerp(g_x3, g_x4, y_d);
 225                 float clut_g = lerp(g_y1, g_y2, z_d);
 226
 227                 float b_x1 = lerp(CLU(b_table,x,y,z), CLU(b_table,x_n,y,z), x_d);
 228                 float b_x2 = lerp(CLU(b_table,x,y_n,z), CLU(b_table,x_n,y_n,z), x_d);
 229                 float b_y1 = lerp(b_x1, b_x2, y_d);
 230                 float b_x3 = lerp(CLU(b_table,x,y,z_n), CLU(b_table,x_n,y,z_n), x_d);
 231                 float b_x4 = lerp(CLU(b_table,x,y_n,z_n), CLU(b_table,x_n,y_n,z_n), x_d);
 232                 float b_y2 = lerp(b_x3, b_x4, y_d);
 233                 float clut_b = lerp(b_y1, b_y2, z_d);
 234
 235                 float pcs_r = lut_interp_linear_float(clut_r,
 236                                 transform->output_clut_table_r, transform->output_clut_table_length);
 237                 float pcs_g = lut_interp_linear_float(clut_g,
 238                                 transform->output_clut_table_g, transform->output_clut_table_length);
 239                 float pcs_b = lut_interp_linear_float(clut_b,
 240                                 transform->output_clut_table_b, transform->output_clut_table_length);
 241
 242                 *dest++ = clamp_float(pcs_r);
 243                 *dest++ = clamp_float(pcs_g);
 244                 *dest++ = clamp_float(pcs_b);
 245         }
 246 }
 247
 248 /* NOT USED
 249 static void qcms_transform_module_tetra_clut(struct qcms_modular_transform *transform, float *src, float *dest, size_t length)
 250 {
 251         size_t i;
 252         int xy_len = 1;
 253         int x_len = transform->grid_size;
 254         int len = x_len * x_len;
 255         float* r_table = transform->r_clut;
 256         float* g_table = transform->g_clut;
 257         float* b_table = transform->b_clut;
 258         float c0_r, c1_r, c2_r, c3_r;
 259         float c0_g, c1_g, c2_g, c3_g;
 260         float c0_b, c1_b, c2_b, c3_b;
 261         float clut_r, clut_g, clut_b;
 262         float pcs_r, pcs_g, pcs_b;
 263         for (i = 0; i < length; i++) {
 264                 float device_r = *src++;
 265                 float device_g = *src++;
 266                 float device_b = *src++;
 267                 float linear_r = lut_interp_linear_float(device_r,
 268                                 transform->input_clut_table_r, transform->input_clut_table_length);
 269                 float linear_g = lut_interp_linear_float(device_g,
 270                                 transform->input_clut_table_g, transform->input_clut_table_length);
 271                 float linear_b = lut_interp_linear_float(device_b,
 272                                 transform->input_clut_table_b, transform->input_clut_table_length);
 273
 274                 int x = floor(linear_r * (transform->grid_size-1));
 275                 int y = floor(linear_g * (transform->grid_size-1));
 276                 int z = floor(linear_b * (transform->grid_size-1));
 277                 int x_n = ceil(linear_r * (transform->grid_size-1));
 278                 int y_n = ceil(linear_g * (transform->grid_size-1));
 279                 int z_n = ceil(linear_b * (transform->grid_size-1));
 280                 float rx = linear_r * (transform->grid_size-1) - x;
 281                 float ry = linear_g * (transform->grid_size-1) - y;
 282                 float rz = linear_b * (transform->grid_size-1) - z;
 283
 284                 c0_r = CLU(r_table, x, y, z);
 285                 c0_g = CLU(g_table, x, y, z);
 286                 c0_b = CLU(b_table, x, y, z);
 287                 if( rx >= ry ) {
 288                         if (ry >= rz) { //rx >= ry && ry >= rz
 289                                 c1_r = CLU(r_table, x_n, y, z) - c0_r;
 290                                 c2_r = CLU(r_table, x_n, y_n, z) - CLU(r_table, x_n, y, z);
 291                                 c3_r = CLU(r_table, x_n, y_n, z_n) - CLU(r_table, x_n, y_n, z);
 292                                 c1_g = CLU(g_table, x_n, y, z) - c0_g;
 293                                 c2_g = CLU(g_table, x_n, y_n, z) - CLU(g_table, x_n, y, z);
 294                                 c3_g = CLU(g_table, x_n, y_n, z_n) - CLU(g_table, x_n, y_n, z);
 295                                 c1_b = CLU(b_table, x_n, y, z) - c0_b;
 296                                 c2_b = CLU(b_table, x_n, y_n, z) - CLU(b_table, x_n, y, z);
 297                                 c3_b = CLU(b_table, x_n, y_n, z_n) - CLU(b_table, x_n, y_n, z);
 298                         } else {
 299                                 if (rx >= rz) { //rx >= rz && rz >= ry
 300                                         c1_r = CLU(r_table, x_n, y, z) - c0_r;
 301                                         c2_r = CLU(r_table, x_n, y_n, z_n) - CLU(r_table, x_n, y, z_n);
 302                                         c3_r = CLU(r_table, x_n, y, z_n) - CLU(r_table, x_n, y, z);
 303                                         c1_g = CLU(g_table, x_n, y, z) - c0_g;
 304                                         c2_g = CLU(g_table, x_n, y_n, z_n) - CLU(g_table, x_n, y, z_n);
 305                                         c3_g = CLU(g_table, x_n, y, z_n) - CLU(g_table, x_n, y, z);
 306                                         c1_b = CLU(b_table, x_n, y, z) - c0_b;
 307                                         c2_b = CLU(b_table, x_n, y_n, z_n) - CLU(b_table, x_n, y, z_n);
 308                                         c3_b = CLU(b_table, x_n, y, z_n) - CLU(b_table, x_n, y, z);
 309                                 } else { //rz > rx && rx >= ry
 310                                         c1_r = CLU(r_table, x_n, y, z_n) - CLU(r_table, x, y, z_n);
 311                                         c2_r = CLU(r_table, x_n, y_n, z_n) - CLU(r_table, x_n, y, z_n);
 312                                         c3_r = CLU(r_table, x, y, z_n) - c0_r;
 313                                         c1_g = CLU(g_table, x_n, y, z_n) - CLU(g_table, x, y, z_n);
 314                                         c2_g = CLU(g_table, x_n, y_n, z_n) - CLU(g_table, x_n, y, z_n);
 315                                         c3_g = CLU(g_table, x, y, z_n) - c0_g;
 316                                         c1_b = CLU(b_table, x_n, y, z_n) - CLU(b_table, x, y, z_n);
 317                                         c2_b = CLU(b_table, x_n, y_n, z_n) - CLU(b_table, x_n, y, z_n);
 318                                         c3_b = CLU(b_table, x, y, z_n) - c0_b;
 319                                 }
 320                         }
 321                 } else {
 322                         if (rx >= rz) { //ry > rx && rx >= rz
 323                                 c1_r = CLU(r_table, x_n, y_n, z) - CLU(r_table, x, y_n, z);
 324                                 c2_r = CLU(r_table, x_n, y_n, z) - c0_r;
 325                                 c3_r = CLU(r_table, x_n, y_n, z_n) - CLU(r_table, x_n, y_n, z);
 326                                 c1_g = CLU(g_table, x_n, y_n, z) - CLU(g_table, x, y_n, z);
 327                                 c2_g = CLU(g_table, x_n, y_n, z) - c0_g;
 328                                 c3_g = CLU(g_table, x_n, y_n, z_n) - CLU(g_table, x_n, y_n, z);
 329                                 c1_b = CLU(b_table, x_n, y_n, z) - CLU(b_table, x, y_n, z);
 330                                 c2_b = CLU(b_table, x_n, y_n, z) - c0_b;
 331                                 c3_b = CLU(b_table, x_n, y_n, z_n) - CLU(b_table, x_n, y_n, z);
 332                         } else {
 333                                 if (ry >= rz) { //ry >= rz && rz > rx
 334                                         c1_r = CLU(r_table, x_n, y_n, z_n) - CLU(r_table, x, y_n, z_n);
 335                                         c2_r = CLU(r_table, x, y_n, z) - c0_r;
 336                                         c3_r = CLU(r_table, x, y_n, z_n) - CLU(r_table, x, y_n, z);
 337                                         c1_g = CLU(g_table, x_n, y_n, z_n) - CLU(g_table, x, y_n, z_n);
 338                                         c2_g = CLU(g_table, x, y_n, z) - c0_g;
 339                                         c3_g = CLU(g_table, x, y_n, z_n) - CLU(g_table, x, y_n, z);
 340                                         c1_b = CLU(b_table, x_n, y_n, z_n) - CLU(b_table, x, y_n, z_n);
 341                                         c2_b = CLU(b_table, x, y_n, z) - c0_b;
 342                                         c3_b = CLU(b_table, x, y_n, z_n) - CLU(b_table, x, y_n, z);
 343                                 } else { //rz > ry && ry > rx
 344                                         c1_r = CLU(r_table, x_n, y_n, z_n) - CLU(r_table, x, y_n, z_n);
 345                                         c2_r = CLU(r_table, x, y_n, z) - c0_r;
 346                                         c3_r = CLU(r_table, x_n, y_n, z_n) - CLU(r_table, x_n, y_n, z);
 347                                         c1_g = CLU(g_table, x_n, y_n, z_n) - CLU(g_table, x, y_n, z_n);
 348                                         c2_g = CLU(g_table, x, y_n, z) - c0_g;
 349                                         c3_g = CLU(g_table, x_n, y_n, z_n) - CLU(g_table, x_n, y_n, z);
 350                                         c1_b = CLU(b_table, x_n, y_n, z_n) - CLU(b_table, x, y_n, z_n);
 351                                         c2_b = CLU(b_table, x, y_n, z) - c0_b;
 352                                         c3_b = CLU(b_table, x_n, y_n, z_n) - CLU(b_table, x_n, y_n, z);
 353                                 }
 354                         }
 355                 }
 356
 357                 clut_r = c0_r + c1_r*rx + c2_r*ry + c3_r*rz;
 358                 clut_g = c0_g + c1_g*rx + c2_g*ry + c3_g*rz;
 359                 clut_b = c0_b + c1_b*rx + c2_b*ry + c3_b*rz;
 360
 361                 pcs_r = lut_interp_linear_float(clut_r,
 362                                 transform->output_clut_table_r, transform->output_clut_table_length);
 363                 pcs_g = lut_interp_linear_float(clut_g,
 364                                 transform->output_clut_table_g, transform->output_clut_table_length);
 365                 pcs_b = lut_interp_linear_float(clut_b,
 366                                 transform->output_clut_table_b, transform->output_clut_table_length);
 367                 *dest++ = clamp_float(pcs_r);
 368                 *dest++ = clamp_float(pcs_g);
 369                 *dest++ = clamp_float(pcs_b);
 370         }
 371 }
 372 */
 373
 374 static void qcms_transform_module_gamma_table(struct qcms_modular_transform *transform, float *src, float *dest, size_t length)
 375 {
 376         size_t i;
 377         float out_r, out_g, out_b;
 378         for (i = 0; i < length; i++) {
 379                 float in_r = *src++;
 380                 float in_g = *src++;
 381                 float in_b = *src++;
 382
 383                 out_r = lut_interp_linear_float(in_r, transform->input_clut_table_r, 256);
 384                 out_g = lut_interp_linear_float(in_g, transform->input_clut_table_g, 256);
 385                 out_b = lut_interp_linear_float(in_b, transform->input_clut_table_b, 256);
 386
 387                 *dest++ = clamp_float(out_r);
 388                 *dest++ = clamp_float(out_g);
 389                 *dest++ = clamp_float(out_b);
 390         }
 391 }
 392
 393 static void qcms_transform_module_gamma_lut(struct qcms_modular_transform *transform, float *src, float *dest, size_t length)
 394 {
 395         size_t i;
 396         float out_r, out_g, out_b;
 397         for (i = 0; i < length; i++) {
 398                 float in_r = *src++;
 399                 float in_g = *src++;
 400                 float in_b = *src++;
 401
 402                 out_r = lut_interp_linear(in_r,
 403                                 transform->output_gamma_lut_r, transform->output_gamma_lut_r_length);
 404                 out_g = lut_interp_linear(in_g,
 405                                 transform->output_gamma_lut_g, transform->output_gamma_lut_g_length);
 406                 out_b = lut_interp_linear(in_b,
 407                                 transform->output_gamma_lut_b, transform->output_gamma_lut_b_length);
 408
 409                 *dest++ = clamp_float(out_r);
 410                 *dest++ = clamp_float(out_g);
 411                 *dest++ = clamp_float(out_b);
 412         }
 413 }
 414
 415 static void qcms_transform_module_matrix_translate(struct qcms_modular_transform *transform, float *src, float *dest, size_t length)
 416 {
 417         size_t i;
 418         struct matrix mat;
 419
 420         /* store the results in column major mode
 421          * this makes doing the multiplication with sse easier */
 422         mat.m[0][0] = transform->matrix.m[0][0];
 423         mat.m[1][0] = transform->matrix.m[0][1];
 424         mat.m[2][0] = transform->matrix.m[0][2];
 425         mat.m[0][1] = transform->matrix.m[1][0];
 426         mat.m[1][1] = transform->matrix.m[1][1];
 427         mat.m[2][1] = transform->matrix.m[1][2];
 428         mat.m[0][2] = transform->matrix.m[2][0];
 429         mat.m[1][2] = transform->matrix.m[2][1];
 430         mat.m[2][2] = transform->matrix.m[2][2];
 431
 432         for (i = 0; i < length; i++) {
 433                 float in_r = *src++;
 434                 float in_g = *src++;
 435                 float in_b = *src++;
 436
 437                 float out_r = mat.m[0][0]*in_r + mat.m[1][0]*in_g + mat.m[2][0]*in_b + transform->tx;
 438                 float out_g = mat.m[0][1]*in_r + mat.m[1][1]*in_g + mat.m[2][1]*in_b + transform->ty;
 439                 float out_b = mat.m[0][2]*in_r + mat.m[1][2]*in_g + mat.m[2][2]*in_b + transform->tz;
 440
 441                 *dest++ = clamp_float(out_r);
 442                 *dest++ = clamp_float(out_g);
 443                 *dest++ = clamp_float(out_b);
 444         }
 445 }
 446
 447 static void qcms_transform_module_matrix(struct qcms_modular_transform *transform, float *src, float *dest, size_t length)
 448 {
 449         size_t i;
 450         struct matrix mat;
 451
 452         /* store the results in column major mode
 453          * this makes doing the multiplication with sse easier */
 454         mat.m[0][0] = transform->matrix.m[0][0];
 455         mat.m[1][0] = transform->matrix.m[0][1];
 456         mat.m[2][0] = transform->matrix.m[0][2];
 457         mat.m[0][1] = transform->matrix.m[1][0];
 458         mat.m[1][1] = transform->matrix.m[1][1];
 459         mat.m[2][1] = transform->matrix.m[1][2];
 460         mat.m[0][2] = transform->matrix.m[2][0];
 461         mat.m[1][2] = transform->matrix.m[2][1];
 462         mat.m[2][2] = transform->matrix.m[2][2];
 463
 464         for (i = 0; i < length; i++) {
 465                 float in_r = *src++;
 466                 float in_g = *src++;
 467                 float in_b = *src++;
 468
 469                 float out_r = mat.m[0][0]*in_r + mat.m[1][0]*in_g + mat.m[2][0]*in_b;
 470                 float out_g = mat.m[0][1]*in_r + mat.m[1][1]*in_g + mat.m[2][1]*in_b;
 471                 float out_b = mat.m[0][2]*in_r + mat.m[1][2]*in_g + mat.m[2][2]*in_b;
 472
 473                 *dest++ = clamp_float(out_r);
 474                 *dest++ = clamp_float(out_g);
 475                 *dest++ = clamp_float(out_b);
 476         }
 477 }
 478
 479 static struct qcms_modular_transform* qcms_modular_transform_alloc() {
 480         return calloc(1, sizeof(struct qcms_modular_transform));
 481 }
 482
 483 static void qcms_modular_transform_release(struct qcms_modular_transform *transform)
 484 {
 485         struct qcms_modular_transform *next_transform;
 486         while (transform != NULL) {
 487                 next_transform = transform->next_transform;
 488                 // clut may use a single block of memory.
 489                 // Perhaps we should remove this to simply the code.
 490                 if (transform->input_clut_table_r + transform->input_clut_table_length == transform->input_clut_table_g && transform->input_clut_table_g + transform->input_clut_table_length == transform->input_clut_table_b) {
 491                         if (transform->input_clut_table_r) free(transform->input_clut_table_r);
 492                 } else {
 493                         if (transform->input_clut_table_r) free(transform->input_clut_table_r);
 494                         if (transform->input_clut_table_g) free(transform->input_clut_table_g);
 495                         if (transform->input_clut_table_b) free(transform->input_clut_table_b);
 496                 }
 497                 if (transform->r_clut + 1 == transform->g_clut && transform->g_clut + 1 == transform->b_clut) {
 498                         if (transform->r_clut) free(transform->r_clut);
 499                 } else {
 500                         if (transform->r_clut) free(transform->r_clut);
 501                         if (transform->g_clut) free(transform->g_clut);
 502                         if (transform->b_clut) free(transform->b_clut);
 503                 }
 504                 if (transform->output_clut_table_r + transform->output_clut_table_length == transform->output_clut_table_g && transform->output_clut_table_g+ transform->output_clut_table_length == transform->output_clut_table_b) {
 505                         if (transform->output_clut_table_r) free(transform->output_clut_table_r);
 506                 } else {
 507                         if (transform->output_clut_table_r) free(transform->output_clut_table_r);
 508                         if (transform->output_clut_table_g) free(transform->output_clut_table_g);
 509                         if (transform->output_clut_table_b) free(transform->output_clut_table_b);
 510                 }
 511                 if (transform->output_gamma_lut_r) free(transform->output_gamma_lut_r);
 512                 if (transform->output_gamma_lut_g) free(transform->output_gamma_lut_g);
 513                 if (transform->output_gamma_lut_b) free(transform->output_gamma_lut_b);
 514                 free(transform);
 515                 transform = next_transform;
 516         }
 517 }
 518
 519 /* Set transform to be the next element in the linked list. */
 520 static void append_transform(struct qcms_modular_transform *transform, struct qcms_modular_transform ***next_transform)
 521 {
 522         **next_transform = transform;
 523         while (transform) {
 524                 *next_transform = &(transform->next_transform);
 525                 transform = transform->next_transform;
 526         }
 527 }
 528
 529 /* reverse the transformation list (used by mBA) */
 530 static struct qcms_modular_transform* reverse_transform(struct qcms_modular_transform *transform)
 531 {
 532         struct qcms_modular_transform *prev_transform = NULL;
 533         while (transform != NULL) {
 534                 struct qcms_modular_transform *next_transform = transform->next_transform;
 535                 transform->next_transform = prev_transform;
 536                 prev_transform = transform;
 537                 transform = next_transform;
 538         }
 539
 540         return prev_transform;
 541 }
 542
 543 #define EMPTY_TRANSFORM_LIST NULL
 544 static struct qcms_modular_transform* qcms_modular_transform_create_mAB(struct lutmABType *lut)
 545 {
 546         struct qcms_modular_transform *first_transform = NULL;
 547         struct qcms_modular_transform **next_transform = &first_transform;
 548         struct qcms_modular_transform *transform = NULL;
 549
 550         if (lut->a_curves[0] != NULL) {
 551                 size_t clut_length;
 552                 float *clut;
 553
 554                 // If the A curve is present this also implies the
 555                 // presence of a CLUT.
 556                 if (!lut->clut_table)
 557                         goto fail;
 558
 559                 // Prepare A curve.
 560                 transform = qcms_modular_transform_alloc();
 561                 if (!transform)
 562                         goto fail;
 563                 append_transform(transform, &next_transform);
 564                 transform->input_clut_table_r = build_input_gamma_table(lut->a_curves[0]);
 565                 transform->input_clut_table_g = build_input_gamma_table(lut->a_curves[1]);
 566                 transform->input_clut_table_b = build_input_gamma_table(lut->a_curves[2]);
 567                 transform->transform_module_fn = qcms_transform_module_gamma_table;
 568                 if (lut->num_grid_points[0] != lut->num_grid_points[1] ||
 569                         lut->num_grid_points[1] != lut->num_grid_points[2] ) {
 570                         //XXX: We don't currently support clut that are not squared!
 571                         goto fail;
 572                 }
 573
 574                 // Prepare CLUT
 575                 transform = qcms_modular_transform_alloc();
 576                 if (!transform)
 577                         goto fail;
 578                 append_transform(transform, &next_transform);
 579                 clut_length = sizeof(float)*pow(lut->num_grid_points[0], 3)*3;
 580                 clut = malloc(clut_length);
 581                 if (!clut)
 582                         goto fail;
 583                 memcpy(clut, lut->clut_table, clut_length);
 584                 transform->r_clut = clut + 0;
 585                 transform->g_clut = clut + 1;
 586                 transform->b_clut = clut + 2;
 587                 transform->grid_size = lut->num_grid_points[0];
 588                 transform->transform_module_fn = qcms_transform_module_clut_only;
 589         }
 590         if (lut->m_curves[0] != NULL) {
 591                 // M curve imples the presence of a Matrix
 592
 593                 // Prepare M curve
 594                 transform = qcms_modular_transform_alloc();
 595                 if (!transform)
 596                         goto fail;
 597                 append_transform(transform, &next_transform);
 598                 transform->input_clut_table_r = build_input_gamma_table(lut->m_curves[0]);
 599                 transform->input_clut_table_g = build_input_gamma_table(lut->m_curves[1]);
 600                 transform->input_clut_table_b = build_input_gamma_table(lut->m_curves[2]);
 601                 transform->transform_module_fn = qcms_transform_module_gamma_table;
 602
 603                 // Prepare Matrix
 604                 transform = qcms_modular_transform_alloc();
 605                 if (!transform)
 606                         goto fail;
 607                 append_transform(transform, &next_transform);
 608                 transform->matrix = build_mAB_matrix(lut);
 609                 if (transform->matrix.invalid)
 610                         goto fail;
 611                 transform->tx = s15Fixed16Number_to_float(lut->e03);
 612                 transform->ty = s15Fixed16Number_to_float(lut->e13);
 613                 transform->tz = s15Fixed16Number_to_float(lut->e23);
 614                 transform->transform_module_fn = qcms_transform_module_matrix_translate;
 615         }
 616         if (lut->b_curves[0] != NULL) {
 617                 // Prepare B curve
 618                 transform = qcms_modular_transform_alloc();
 619                 if (!transform)
 620                         goto fail;
 621                 append_transform(transform, &next_transform);
 622                 transform->input_clut_table_r = build_input_gamma_table(lut->b_curves[0]);
 623                 transform->input_clut_table_g = build_input_gamma_table(lut->b_curves[1]);
 624                 transform->input_clut_table_b = build_input_gamma_table(lut->b_curves[2]);
 625                 transform->transform_module_fn = qcms_transform_module_gamma_table;
 626         } else {
 627                 // B curve is mandatory
 628                 goto fail;
 629         }
 630
 631         if (lut->reversed) {
 632                 // mBA are identical to mAB except that the transformation order
 633                 // is reversed
 634                 first_transform = reverse_transform(first_transform);
 635         }
 636
 637         return first_transform;
 638 fail:
 639         qcms_modular_transform_release(first_transform);
 640         return NULL;
 641 }
 642
 643 static struct qcms_modular_transform* qcms_modular_transform_create_lut(struct lutType *lut)
 644 {
 645         struct qcms_modular_transform *first_transform = NULL;
 646         struct qcms_modular_transform **next_transform = &first_transform;
 647         struct qcms_modular_transform *transform = NULL;
 648
 649         size_t in_curve_len, clut_length, out_curve_len;
 650         float *in_curves, *clut, *out_curves;
 651
 652         // Prepare Matrix
 653         transform = qcms_modular_transform_alloc();
 654         if (!transform)
 655                 goto fail;
 656         append_transform(transform, &next_transform);
 657         transform->matrix = build_lut_matrix(lut);
 658         if (transform->matrix.invalid)
 659                 goto fail;
 660         transform->transform_module_fn = qcms_transform_module_matrix;
 661
 662         // Prepare input curves
 663         transform = qcms_modular_transform_alloc();
 664         if (!transform)
 665                 goto fail;
 666         append_transform(transform, &next_transform);
 667         in_curve_len = sizeof(float)*lut->num_input_table_entries * 3;
 668         in_curves = malloc(in_curve_len);
 669         if (!in_curves)
 670                 goto fail;
 671         memcpy(in_curves, lut->input_table, in_curve_len);
 672         transform->input_clut_table_r = in_curves + lut->num_input_table_entries * 0;
 673         transform->input_clut_table_g = in_curves + lut->num_input_table_entries * 1;
 674         transform->input_clut_table_b = in_curves + lut->num_input_table_entries * 2;
 675         transform->input_clut_table_length = lut->num_input_table_entries;
 676
 677         // Prepare table
 678         clut_length = sizeof(float)*pow(lut->num_clut_grid_points, 3)*3;
 679         clut = malloc(clut_length);
 680         if (!clut)
 681                 goto fail;
 682         memcpy(clut, lut->clut_table, clut_length);
 683         transform->r_clut = clut + 0;
 684         transform->g_clut = clut + 1;
 685         transform->b_clut = clut + 2;
 686         transform->grid_size = lut->num_clut_grid_points;
 687
 688         // Prepare output curves
 689         out_curve_len = sizeof(float) * lut->num_output_table_entries * 3;
 690         out_curves = malloc(out_curve_len);
 691         if (!out_curves)
 692                 goto fail;
 693         memcpy(out_curves, lut->output_table, out_curve_len);
 694         transform->output_clut_table_r = out_curves + lut->num_output_table_entries * 0;
 695         transform->output_clut_table_g = out_curves + lut->num_output_table_entries * 1;
 696         transform->output_clut_table_b = out_curves + lut->num_output_table_entries * 2;
 697         transform->output_clut_table_length = lut->num_output_table_entries;
 698         transform->transform_module_fn = qcms_transform_module_clut;
 699
 700         return first_transform;
 701 fail:
 702         qcms_modular_transform_release(first_transform);
 703         return NULL;
 704 }
 705
 706 struct qcms_modular_transform* qcms_modular_transform_create_input(qcms_profile *in)
 707 {
 708         struct qcms_modular_transform *first_transform = NULL;
 709         struct qcms_modular_transform **next_transform = &first_transform;
 710
 711         if (in->A2B0) {
 712                 struct qcms_modular_transform *lut_transform;
 713                 lut_transform = qcms_modular_transform_create_lut(in->A2B0);
 714                 if (!lut_transform)
 715                         goto fail;
 716                 append_transform(lut_transform, &next_transform);
 717         } else if (in->mAB && in->mAB->num_in_channels == 3 && in->mAB->num_out_channels == 3) {
 718                 struct qcms_modular_transform *mAB_transform;
 719                 mAB_transform = qcms_modular_transform_create_mAB(in->mAB);
 720                 if (!mAB_transform)
 721                         goto fail;
 722                 append_transform(mAB_transform, &next_transform);
 723
 724         } else {
 725                 struct qcms_modular_transform *transform;
 726
 727                 transform = qcms_modular_transform_alloc();
 728                 if (!transform)
 729                         goto fail;
 730                 append_transform(transform, &next_transform);
 731                 transform->input_clut_table_r = build_input_gamma_table(in->redTRC);
 732                 transform->input_clut_table_g = build_input_gamma_table(in->greenTRC);
 733                 transform->input_clut_table_b = build_input_gamma_table(in->blueTRC);
 734                 transform->transform_module_fn = qcms_transform_module_gamma_table;
 735                 if (!transform->input_clut_table_r || !transform->input_clut_table_g ||
 736                                 !transform->input_clut_table_b) {
 737                         goto fail;
 738                 }
 739
 740                 transform = qcms_modular_transform_alloc();
 741                 if (!transform)
 742                         goto fail;
 743                 append_transform(transform, &next_transform);
 744                 transform->matrix.m[0][0] = 1/1.999969482421875f;
 745                 transform->matrix.m[0][1] = 0.f;
 746                 transform->matrix.m[0][2] = 0.f;
 747                 transform->matrix.m[1][0] = 0.f;
 748                 transform->matrix.m[1][1] = 1/1.999969482421875f;
 749                 transform->matrix.m[1][2] = 0.f;
 750                 transform->matrix.m[2][0] = 0.f;
 751                 transform->matrix.m[2][1] = 0.f;
 752                 transform->matrix.m[2][2] = 1/1.999969482421875f;
 753                 transform->matrix.invalid = false;
 754                 transform->transform_module_fn = qcms_transform_module_matrix;
 755
 756                 transform = qcms_modular_transform_alloc();
 757                 if (!transform)
 758                         goto fail;
 759                 append_transform(transform, &next_transform);
 760                 transform->matrix = build_colorant_matrix(in);
 761                 transform->transform_module_fn = qcms_transform_module_matrix;
 762         }
 763
 764         return first_transform;
 765 fail:
 766         qcms_modular_transform_release(first_transform);
 767         return EMPTY_TRANSFORM_LIST;
 768 }
 769 static struct qcms_modular_transform* qcms_modular_transform_create_output(qcms_profile *out)
 770 {
 771         struct qcms_modular_transform *first_transform = NULL;
 772         struct qcms_modular_transform **next_transform = &first_transform;
 773
 774         if (out->B2A0) {
 775                 struct qcms_modular_transform *lut_transform;
 776                 lut_transform = qcms_modular_transform_create_lut(out->B2A0);
 777                 if (!lut_transform)
 778                         goto fail;
 779                 append_transform(lut_transform, &next_transform);
 780         } else if (out->mBA && out->mBA->num_in_channels == 3 && out->mBA->num_out_channels == 3) {
 781                 struct qcms_modular_transform *lut_transform;
 782                 lut_transform = qcms_modular_transform_create_mAB(out->mBA);
 783                 if (!lut_transform)
 784                         goto fail;
 785                 append_transform(lut_transform, &next_transform);
 786         } else if (out->redTRC && out->greenTRC && out->blueTRC) {
 787                 struct qcms_modular_transform *transform;
 788
 789                 transform = qcms_modular_transform_alloc();
 790                 if (!transform)
 791                         goto fail;
 792                 append_transform(transform, &next_transform);
 793                 transform->matrix = matrix_invert(build_colorant_matrix(out));
 794                 transform->transform_module_fn = qcms_transform_module_matrix;
 795
 796                 transform = qcms_modular_transform_alloc();
 797                 if (!transform)
 798                         goto fail;
 799                 append_transform(transform, &next_transform);
 800                 transform->matrix.m[0][0] = 1.999969482421875f;
 801                 transform->matrix.m[0][1] = 0.f;
 802                 transform->matrix.m[0][2] = 0.f;
 803                 transform->matrix.m[1][0] = 0.f;
 804                 transform->matrix.m[1][1] = 1.999969482421875f;
 805                 transform->matrix.m[1][2] = 0.f;
 806                 transform->matrix.m[2][0] = 0.f;
 807                 transform->matrix.m[2][1] = 0.f;
 808                 transform->matrix.m[2][2] = 1.999969482421875f;
 809                 transform->matrix.invalid = false;
 810                 transform->transform_module_fn = qcms_transform_module_matrix;
 811
 812                 transform = qcms_modular_transform_alloc();
 813                 if (!transform)
 814                         goto fail;
 815                 append_transform(transform, &next_transform);
 816                 build_output_lut(out->redTRC, &transform->output_gamma_lut_r,
 817                         &transform->output_gamma_lut_r_length);
 818                 build_output_lut(out->greenTRC, &transform->output_gamma_lut_g,
 819                         &transform->output_gamma_lut_g_length);
 820                 build_output_lut(out->blueTRC, &transform->output_gamma_lut_b,
 821                         &transform->output_gamma_lut_b_length);
 822                 transform->transform_module_fn = qcms_transform_module_gamma_lut;
 823
 824                 if (!transform->output_gamma_lut_r || !transform->output_gamma_lut_g ||
 825                                 !transform->output_gamma_lut_b) {
 826                         goto fail;
 827                 }
 828         } else {
 829                 assert(0 && "Unsupported output profile workflow.");
 830                 return NULL;
 831         }
 832
 833         return first_transform;
 834 fail:
 835         qcms_modular_transform_release(first_transform);
 836         return EMPTY_TRANSFORM_LIST;
 837 }
 838
 839 /* Not Completed
 840 // Simplify the transformation chain to an equivalent transformation chain
 841 static struct qcms_modular_transform* qcms_modular_transform_reduce(struct qcms_modular_transform *transform)
 842 {
 843         struct qcms_modular_transform *first_transform = NULL;
 844         struct qcms_modular_transform *curr_trans = transform;
 845         struct qcms_modular_transform *prev_trans = NULL;
 846         while (curr_trans) {
 847                 struct qcms_modular_transform *next_trans = curr_trans->next_transform;
 848                 if (curr_trans->transform_module_fn == qcms_transform_module_matrix) {
 849                         if (next_trans && next_trans->transform_module_fn == qcms_transform_module_matrix) {
 850                                 curr_trans->matrix = matrix_multiply(curr_trans->matrix, next_trans->matrix);
 851                                 goto remove_next;
 852                         }
 853                 }
 854                 if (curr_trans->transform_module_fn == qcms_transform_module_gamma_table) {
 855                         bool isLinear = true;
 856                         uint16_t i;
 857                         for (i = 0; isLinear && i < 256; i++) {
 858                                 isLinear &= (int)(curr_trans->input_clut_table_r[i] * 255) == i;
 859                                 isLinear &= (int)(curr_trans->input_clut_table_g[i] * 255) == i;
 860                                 isLinear &= (int)(curr_trans->input_clut_table_b[i] * 255) == i;
 861                         }
 862                         goto remove_current;
 863                 }
 864
 865 next_transform:
 866                 if (!next_trans) break;
 867                 prev_trans = curr_trans;
 868                 curr_trans = next_trans;
 869                 continue;
 870 remove_current:
 871                 if (curr_trans == transform) {
 872                         //Update head
 873                         transform = next_trans;
 874                 } else {
 875                         prev_trans->next_transform = next_trans;
 876                 }
 877                 curr_trans->next_transform = NULL;
 878                 qcms_modular_transform_release(curr_trans);
 879                 //return transform;
 880                 return qcms_modular_transform_reduce(transform);
 881 remove_next:
 882                 curr_trans->next_transform = next_trans->next_transform;
 883                 next_trans->next_transform = NULL;
 884                 qcms_modular_transform_release(next_trans);
 885                 continue;
 886         }
 887         return transform;
 888 }
 889 */
 890
 891 static struct qcms_modular_transform* qcms_modular_transform_create(qcms_profile *in, qcms_profile *out)
 892 {
 893         struct qcms_modular_transform *first_transform = NULL;
 894         struct qcms_modular_transform **next_transform = &first_transform;
 895
 896         if (in->color_space == RGB_SIGNATURE) {
 897                 struct qcms_modular_transform* rgb_to_pcs;
 898                 rgb_to_pcs = qcms_modular_transform_create_input(in);
 899                 if (!rgb_to_pcs)
 900                         goto fail;
 901                 append_transform(rgb_to_pcs, &next_transform);
 902         } else {
 903                 assert(0 && "input color space not supported");
 904                 goto fail;
 905         }
 906
 907         if (in->pcs == LAB_SIGNATURE && out->pcs == XYZ_SIGNATURE) {
 908                 struct qcms_modular_transform* lab_to_pcs;
 909                 lab_to_pcs = qcms_modular_transform_alloc();
 910                 if (!lab_to_pcs)
 911                         goto fail;
 912                 append_transform(lab_to_pcs, &next_transform);
 913                 lab_to_pcs->transform_module_fn = qcms_transform_module_LAB_to_XYZ;
 914         }
 915
 916         // This does not improve accuracy in practice, something is wrong here.
 917         //if (in->chromaticAdaption.invalid == false) {
 918         //      struct qcms_modular_transform* chromaticAdaption;
 919         //      chromaticAdaption = qcms_modular_transform_alloc();
 920         //      if (!chromaticAdaption)
 921         //              goto fail;
 922         //      append_transform(chromaticAdaption, &next_transform);
 923         //      chromaticAdaption->matrix = matrix_invert(in->chromaticAdaption);
 924         //      chromaticAdaption->transform_module_fn = qcms_transform_module_matrix;
 925         //}
 926
 927         if (in->pcs == XYZ_SIGNATURE && out->pcs == LAB_SIGNATURE) {
 928                 struct qcms_modular_transform* pcs_to_lab;
 929                 pcs_to_lab = qcms_modular_transform_alloc();
 930                 if (!pcs_to_lab)
 931                         goto fail;
 932                 append_transform(pcs_to_lab, &next_transform);
 933                 pcs_to_lab->transform_module_fn = qcms_transform_module_XYZ_to_LAB;
 934         }
 935
 936         if (out->color_space == RGB_SIGNATURE) {
 937                 struct qcms_modular_transform* pcs_to_rgb;
 938                 pcs_to_rgb = qcms_modular_transform_create_output(out);
 939                 if (!pcs_to_rgb)
 940                         goto fail;
 941                 append_transform(pcs_to_rgb, &next_transform);
 942         } else {
 943                 assert(0 && "output color space not supported");
 944                 goto fail;
 945         }
 946         // Not Completed
 947         //return qcms_modular_transform_reduce(first_transform);
 948         return first_transform;
 949 fail:
 950         qcms_modular_transform_release(first_transform);
 951         return EMPTY_TRANSFORM_LIST;
 952 }
 953
 954 static float* qcms_modular_transform_data(struct qcms_modular_transform *transform, float *src, float *dest, size_t len)
 955 {
 956         while (transform != NULL) {
 957                 // Keep swaping src/dest when performing a transform to use less memory.
 958                 float *new_src = dest;
 959                 const transform_module_fn_t transform_fn = transform->transform_module_fn;
 960                 if (transform_fn != qcms_transform_module_gamma_table &&
 961                     transform_fn != qcms_transform_module_gamma_lut &&
 962                     transform_fn != qcms_transform_module_clut &&
 963                     transform_fn != qcms_transform_module_clut_only &&
 964                     transform_fn != qcms_transform_module_matrix &&
 965                     transform_fn != qcms_transform_module_matrix_translate &&
 966                     transform_fn != qcms_transform_module_LAB_to_XYZ &&
 967                     transform_fn != qcms_transform_module_XYZ_to_LAB) {
 968                         assert(0 && "Unsupported transform module");
 969                         return NULL;
 970                 }
 971                 transform->transform_module_fn(transform,src,dest,len);
 972                 dest = src;
 973                 src = new_src;
 974                 transform = transform->next_transform;
 975         }
 976         // The results end up in the src buffer because of the switching
 977         return src;
 978 }
 979
 980 float* qcms_chain_transform(qcms_profile *in, qcms_profile *out, float *src, float *dest, size_t lutSize)
 981 {
 982         struct qcms_modular_transform *transform_list = qcms_modular_transform_create(in, out);
 983         if (transform_list != NULL) {
 984                 float *lut = qcms_modular_transform_data(transform_list, src, dest, lutSize/3);
 985                 qcms_modular_transform_release(transform_list);
 986                 return lut;
 987         }
 988         return NULL;
 989 }