DXR is a code search and navigation tool aimed at making sense of large projects. It supports full-text and regex searches as well as structural queries.

Mercurial (920bcf17a9e1)

VCS Links

Line Code
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586
//  qcms
//  Copyright (C) 2009 Mozilla Corporation
//  Copyright (C) 1998-2007 Marti Maria
//
// Permission is hereby granted, free of charge, to any person obtaining 
// a copy of this software and associated documentation files (the "Software"), 
// to deal in the Software without restriction, including without limitation 
// the rights to use, copy, modify, merge, publish, distribute, sublicense, 
// and/or sell copies of the Software, and to permit persons to whom the Software 
// is furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in 
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 
// EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO 
// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 
// NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE 
// LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION 
// OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION 
// WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

#include <stdlib.h>
#include <math.h>
#include <assert.h>
#include "qcmsint.h"

#if defined(_M_IX86) || defined(__i386__)
#define X86
#endif

//XXX: could use a bettername
typedef uint16_t uint16_fract_t;

/* value must be a value between 0 and 1 */
//XXX: is the above a good restriction to have?
float lut_interp_linear(double value, uint16_t *table, int length)
{
	int upper, lower;
	value = value * (length - 1);
	upper = ceil(value);
	lower = floor(value);
	//XXX: can we be more performant here?
	value = table[upper]*(1. - (upper - value)) + table[lower]*(upper - value);
	/* scale the value */
	return value * (1./65535.);
}

/* same as above but takes and returns a uint16_t value representing a range from 0..1 */
uint16_t lut_interp_linear16(uint16_t input_value, uint16_t *table, int length)
{
	uint32_t value = (input_value * (length - 1));
	uint32_t upper = (value + 65534) / 65535; /* equivalent to ceil(value/65535) */
	uint32_t lower = value / 65535;           /* equivalent to floor(value/65535) */
	uint32_t interp = value % 65535;

	value = (table[upper]*(interp) + table[lower]*(65535 - interp))/65535;

	return value;
}

void compute_curve_gamma_table_type1(float gamma_table[256], double gamma)
{
	unsigned int i;
	for (i = 0; i < 256; i++) {
		gamma_table[i] = pow(i/255., gamma);
	}
}

void compute_curve_gamma_table_type2(float gamma_table[256], uint16_t *table, int length)
{
	unsigned int i;
	for (i = 0; i < 256; i++) {
		gamma_table[i] = lut_interp_linear(i/255., table, length);
	}
}

void compute_curve_gamma_table_type0(float gamma_table[256])
{
	unsigned int i;
	for (i = 0; i < 256; i++) {
		gamma_table[i] = i/255.;
	}
}

unsigned char clamp_u8(float v)
{
	if (v > 255.)
		return 255;
	else if (v < 0)
		return 0;
	else
		return floor(v+.5);
}

struct vector {
	float v[3];
};

struct matrix {
	float m[3][3];
	bool invalid;
};

struct vector matrix_eval(struct matrix mat, struct vector v)
{
	struct vector result;
	result.v[0] = mat.m[0][0]*v.v[0] + mat.m[0][1]*v.v[1] + mat.m[0][2]*v.v[2];
	result.v[1] = mat.m[1][0]*v.v[0] + mat.m[1][1]*v.v[1] + mat.m[1][2]*v.v[2];
	result.v[2] = mat.m[2][0]*v.v[0] + mat.m[2][1]*v.v[1] + mat.m[2][2]*v.v[2];
	return result;
}

//XXX: should probably pass by reference and we could
//probably reuse this computation in matrix_invert
float matrix_det(struct matrix mat)
{
	float det;
	det = mat.m[0][0]*mat.m[1][1]*mat.m[2][2] +
		mat.m[0][1]*mat.m[1][2]*mat.m[2][0] +
		mat.m[0][2]*mat.m[1][0]*mat.m[2][1] -
		mat.m[0][0]*mat.m[1][2]*mat.m[2][1] -
		mat.m[0][1]*mat.m[1][0]*mat.m[2][2] -
		mat.m[0][2]*mat.m[1][1]*mat.m[2][0];
	return det;
}

/* from pixman and cairo and Mathematics for Game Programmers */
/* lcms uses gauss-jordan elimination with partial pivoting which is
 * less efficient and not as numerically stable. See Mathematics for
 * Game Programmers. */
struct matrix matrix_invert(struct matrix mat)
{
	struct matrix dest_mat;
	int i,j;
	static int a[3] = { 2, 2, 1 };
	static int b[3] = { 1, 0, 0 };

	/* inv  (A) = 1/det (A) * adj (A) */
	float det = matrix_det(mat);

	if (det == 0) {
		dest_mat.invalid = true;
	} else {
		dest_mat.invalid = false;
	}

	det = 1/det;

	for (j = 0; j < 3; j++) {
		for (i = 0; i < 3; i++) {
			double p;
			int ai = a[i];
			int aj = a[j];
			int bi = b[i];
			int bj = b[j];

			p = mat.m[ai][aj] * mat.m[bi][bj] -
				mat.m[ai][bj] * mat.m[bi][aj];
			if (((i + j) & 1) != 0)
				p = -p;

			dest_mat.m[j][i] = det * p;
		}
	}
	return dest_mat;
}

struct matrix matrix_identity(void)
{
	struct matrix i;
	i.m[0][0] = 1;
	i.m[0][1] = 0;
	i.m[0][2] = 0;
	i.m[1][0] = 0;
	i.m[1][1] = 1;
	i.m[1][2] = 0;
	i.m[2][0] = 0;
	i.m[2][1] = 0;
	i.m[2][2] = 1;
	i.invalid = false;
	return i;
}

static struct matrix matrix_invalid(void)
{
	struct matrix inv = matrix_identity();
	inv.invalid = true;
	return inv;
}


/* from pixman */
/* MAT3per... */
struct matrix matrix_multiply(struct matrix a, struct matrix b)
{
	struct matrix result;
	int dx, dy;
	int o;
	for (dy = 0; dy < 3; dy++) {
		for (dx = 0; dx < 3; dx++) {
			double v = 0;
			for (o = 0; o < 3; o++) {
				v += a.m[dy][o] * b.m[o][dx];
			}
			result.m[dy][dx] = v;
		}
	}
	result.invalid = a.invalid || b.invalid;
	return result;
}

float u8Fixed8Number_to_float(uint16_t x)
{
	// 0x0000 = 0.
	// 0x0100 = 1.
	// 0xffff = 255  + 255/256
	return x/256.;
}

float *build_input_gamma_table(struct curveType *TRC)
{
	float *gamma_table = malloc(sizeof(float)*256);
	if (gamma_table) {
		if (TRC->count == 0) {
			compute_curve_gamma_table_type0(gamma_table);
		} else if (TRC->count == 1) {
			compute_curve_gamma_table_type1(gamma_table, u8Fixed8Number_to_float(TRC->data[0]));
		} else {
			compute_curve_gamma_table_type2(gamma_table, TRC->data, TRC->count);
		}
	}
	return gamma_table;
}

struct matrix build_colorant_matrix(qcms_profile *p)
{
	struct matrix result;
	result.m[0][0] = s15Fixed16Number_to_float(p->redColorant.X);
	result.m[0][1] = s15Fixed16Number_to_float(p->greenColorant.X);
	result.m[0][2] = s15Fixed16Number_to_float(p->blueColorant.X);
	result.m[1][0] = s15Fixed16Number_to_float(p->redColorant.Y);
	result.m[1][1] = s15Fixed16Number_to_float(p->greenColorant.Y);
	result.m[1][2] = s15Fixed16Number_to_float(p->blueColorant.Y);
	result.m[2][0] = s15Fixed16Number_to_float(p->redColorant.Z);
	result.m[2][1] = s15Fixed16Number_to_float(p->greenColorant.Z);
	result.m[2][2] = s15Fixed16Number_to_float(p->blueColorant.Z);
	result.invalid = false;
	return result;
}

/* The following code is copied nearly directly from lcms.
 * I think it could be much better. For example, Argyll seems to have better code in
 * icmTable_lookup_bwd and icmTable_setup_bwd. However, for now this is a quick way
 * to a working solution and allows for easy comparing with lcms. */
uint16_fract_t lut_inverse_interp16(uint16_t Value, uint16_t LutTable[], int length)
{
        int l = 1;
        int r = 0x10000;
        int x = 0, res;       // 'int' Give spacing for negative values
        int NumZeroes, NumPoles;
        int cell0, cell1;
        double val2;
        double y0, y1, x0, x1;
        double a, b, f;

        // July/27 2001 - Expanded to handle degenerated curves with an arbitrary
        // number of elements containing 0 at the begining of the table (Zeroes)
        // and another arbitrary number of poles (FFFFh) at the end.
        // First the zero and pole extents are computed, then value is compared.

        NumZeroes = 0;
        while (LutTable[NumZeroes] == 0 && NumZeroes < length-1)
                        NumZeroes++;

        // There are no zeros at the beginning and we are trying to find a zero, so
        // return anything. It seems zero would be the less destructive choice
	/* I'm not sure that this makes sense, but oh well... */
        if (NumZeroes == 0 && Value == 0)
            return 0;

        NumPoles = 0;
        while (LutTable[length-1- NumPoles] == 0xFFFF && NumPoles < length-1)
                        NumPoles++;

        // Does the curve belong to this case?
        if (NumZeroes > 1 || NumPoles > 1)
        {               
                int a, b;

                // Identify if value fall downto 0 or FFFF zone             
                if (Value == 0) return 0;
               // if (Value == 0xFFFF) return 0xFFFF;

                // else restrict to valid zone

                a = ((NumZeroes-1) * 0xFFFF) / (length-1);               
                b = ((length-1 - NumPoles) * 0xFFFF) / (length-1);
                                                                
                l = a - 1;
                r = b + 1;
        }


        // Seems not a degenerated case... apply binary search

        while (r > l) {

                x = (l + r) / 2;

		res = (int) lut_interp_linear16((uint16_fract_t) (x-1), LutTable, length);

                if (res == Value) {

                    // Found exact match. 
                    
                    return (uint16_fract_t) (x - 1);
                }

                if (res > Value) r = x - 1;
                else l = x + 1;
        }

        // Not found, should we interpolate?

                
        // Get surrounding nodes
        
        val2 = (length-1) * ((double) (x - 1) / 65535.0);

        cell0 = (int) floor(val2);
        cell1 = (int) ceil(val2);
           
        if (cell0 == cell1) return (uint16_fract_t) x;

        y0 = LutTable[cell0] ;
        x0 = (65535.0 * cell0) / (length-1); 

        y1 = LutTable[cell1] ;
        x1 = (65535.0 * cell1) / (length-1);

        a = (y1 - y0) / (x1 - x0);
        b = y0 - a * x0;

        if (fabs(a) < 0.01) return (uint16_fract_t) x;

        f = ((Value - b) / a);

        if (f < 0.0) return (uint16_fract_t) 0;
        if (f >= 65535.0) return (uint16_fract_t) 0xFFFF;

        return (uint16_fract_t) floor(f + 0.5);                        
        
}

// Build a White point, primary chromas transfer matrix from RGB to CIE XYZ
// This is just an approximation, I am not handling all the non-linear
// aspects of the RGB to XYZ process, and assumming that the gamma correction
// has transitive property in the tranformation chain.
//
// the alghoritm:
//
//            - First I build the absolute conversion matrix using
//              primaries in XYZ. This matrix is next inverted
//            - Then I eval the source white point across this matrix
//              obtaining the coeficients of the transformation
//            - Then, I apply these coeficients to the original matrix
static struct matrix build_RGB_to_XYZ_transfer_matrix(qcms_CIE_xyY white, qcms_CIE_xyYTRIPLE primrs)
{
	struct matrix primaries;
	struct matrix primaries_invert;
	struct matrix result;
	struct vector white_point;
	struct vector coefs;

	double xn, yn;
	double xr, yr;
	double xg, yg;
	double xb, yb;

	xn = white.x;
	yn = white.y;

	if (yn == 0.0)
		return matrix_invalid();

	xr = primrs.red.x;
	yr = primrs.red.y;
	xg = primrs.green.x;
	yg = primrs.green.y;
	xb = primrs.blue.x;
	yb = primrs.blue.y;

	primaries.m[0][0] = xr;
	primaries.m[0][1] = xg;
	primaries.m[0][2] = xb;

	primaries.m[1][0] = yr;
	primaries.m[1][1] = yg;
	primaries.m[1][2] = yb;

	primaries.m[2][0] = 1 - xr - yr;
	primaries.m[2][1] = 1 - xg - yg;
	primaries.m[2][2] = 1 - xb - yb;
	primaries.invalid = false;

	white_point.v[0] = xn/yn;
	white_point.v[1] = 1.;
	white_point.v[2] = (1.0-xn-yn)/yn;

	primaries_invert = matrix_invert(primaries);

	coefs = matrix_eval(primaries_invert, white_point);

	result.m[0][0] = coefs.v[0]*xr;
	result.m[0][1] = coefs.v[1]*xg;
	result.m[0][2] = coefs.v[2]*xb;

	result.m[1][0] = coefs.v[0]*yr;
	result.m[1][1] = coefs.v[1]*yg;
	result.m[1][2] = coefs.v[2]*yb;

	result.m[2][0] = coefs.v[0]*(1.-xr-yr);
	result.m[2][1] = coefs.v[1]*(1.-xg-yg);
	result.m[2][2] = coefs.v[2]*(1.-xb-yb);
	result.invalid = primaries_invert.invalid;

	return result;
}

struct CIE_XYZ {
	double X;
	double Y;
	double Z;
};

/* CIE Illuminant D50 */
static const struct CIE_XYZ D50_XYZ = {
	0.9642,
	1.0000,
	0.8249
};

/* from lcms: xyY2XYZ()
 * corresponds to argyll: icmYxy2XYZ() */
static struct CIE_XYZ xyY2XYZ(qcms_CIE_xyY source)
{
	struct CIE_XYZ dest;
	dest.X = (source.x / source.y) * source.Y;
	dest.Y = source.Y;
	dest.Z = ((1 - source.x - source.y) / source.y) * source.Y;
	return dest;
}

/* from lcms: ComputeChromaticAdaption */
// Compute chromatic adaption matrix using chad as cone matrix
static struct matrix
compute_chromatic_adaption(struct CIE_XYZ source_white_point,
                           struct CIE_XYZ dest_white_point,
                           struct matrix chad)
{
	struct matrix chad_inv;
	struct vector cone_source_XYZ, cone_source_rgb;
	struct vector cone_dest_XYZ, cone_dest_rgb;
	struct matrix cone, tmp;

	tmp = chad;
	chad_inv = matrix_invert(tmp);

	cone_source_XYZ.v[0] = source_white_point.X;
	cone_source_XYZ.v[1] = source_white_point.Y;
	cone_source_XYZ.v[2] = source_white_point.Z;

	cone_dest_XYZ.v[0] = dest_white_point.X;
	cone_dest_XYZ.v[1] = dest_white_point.Y;
	cone_dest_XYZ.v[2] = dest_white_point.Z;

	cone_source_rgb = matrix_eval(chad, cone_source_XYZ);
	cone_dest_rgb   = matrix_eval(chad, cone_dest_XYZ);

	cone.m[0][0] = cone_dest_rgb.v[0]/cone_source_rgb.v[0];
	cone.m[0][1] = 0;
	cone.m[0][2] = 0;
	cone.m[1][0] = 0;
	cone.m[1][1] = cone_dest_rgb.v[1]/cone_source_rgb.v[1];
	cone.m[1][2] = 0;
	cone.m[2][0] = 0;
	cone.m[2][1] = 0;
	cone.m[2][2] = cone_dest_rgb.v[2]/cone_source_rgb.v[2];
	cone.invalid = false;

	// Normalize
	return matrix_multiply(chad_inv, matrix_multiply(cone, chad));
}

/* from lcms: cmsAdaptionMatrix */
// Returns the final chrmatic adaptation from illuminant FromIll to Illuminant ToIll
// Bradford is assumed
static struct matrix
adaption_matrix(struct CIE_XYZ source_illumination, struct CIE_XYZ target_illumination)
{
	struct matrix lam_rigg = {{ // Bradford matrix
	                         {  0.8951,  0.2664, -0.1614 },
	                         { -0.7502,  1.7135,  0.0367 },
	                         {  0.0389, -0.0685,  1.0296 }
	                         }};
	return compute_chromatic_adaption(source_illumination, target_illumination, lam_rigg);
}

/* from lcms: cmsAdaptMatrixToD50 */
static struct matrix adapt_matrix_to_D50(struct matrix r, qcms_CIE_xyY source_white_pt)
{
	struct CIE_XYZ Dn;
	struct matrix Bradford;

	if (source_white_pt.y == 0.0)
		return matrix_invalid();

	Dn = xyY2XYZ(source_white_pt);

	Bradford = adaption_matrix(Dn, D50_XYZ);
	return matrix_multiply(Bradford, r);
}

qcms_bool set_rgb_colorants(qcms_profile *profile, qcms_CIE_xyY white_point, qcms_CIE_xyYTRIPLE primaries)
{
	struct matrix colorants;
	colorants = build_RGB_to_XYZ_transfer_matrix(white_point, primaries);
	colorants = adapt_matrix_to_D50(colorants, white_point);

	if (colorants.invalid)
		return false;

	/* note: there's a transpose type of operation going on here */
	profile->redColorant.X = double_to_s15Fixed16Number(colorants.m[0][0]);
	profile->redColorant.Y = double_to_s15Fixed16Number(colorants.m[1][0]);
	profile->redColorant.Z = double_to_s15Fixed16Number(colorants.m[2][0]);

	profile->greenColorant.X = double_to_s15Fixed16Number(colorants.m[0][1]);
	profile->greenColorant.Y = double_to_s15Fixed16Number(colorants.m[1][1]);
	profile->greenColorant.Z = double_to_s15Fixed16Number(colorants.m[2][1]);

	profile->blueColorant.X = double_to_s15Fixed16Number(colorants.m[0][2]);
	profile->blueColorant.Y = double_to_s15Fixed16Number(colorants.m[1][2]);
	profile->blueColorant.Z = double_to_s15Fixed16Number(colorants.m[2][2]);

	return true;
}

/*
 The number of entries needed to invert a lookup table should not
 necessarily be the same as the original number of entries.  This is
 especially true of lookup tables that have a small number of entries.

 For example:
 Using a table like:
    {0, 3104, 14263, 34802, 65535}
 invert_lut will produce an inverse of:
    {3, 34459, 47529, 56801, 65535}
 which has an maximum error of about 9855 (pixel difference of ~38.346)

 For now, we punt the decision of output size to the caller. */
static uint16_t *invert_lut(uint16_t *table, int length, int out_length)
{
	int i;
	/* for now we invert the lut by creating a lut of size out_length
	 * and attempting to lookup a value for each entry using lut_inverse_interp16 */
	uint16_t *output = malloc(sizeof(uint16_t)*out_length);
	if (!output)
		return NULL;

	for (i = 0; i < out_length; i++) {
		double x = ((double) i * 65535.) / (double) (out_length - 1);
		uint16_fract_t input = floor(x + .5);
		output[i] = lut_inverse_interp16(input, table, length);
	}
	return output;
}

static uint16_t *build_linear_table(int length)
{
	int i;
	uint16_t *output = malloc(sizeof(uint16_t)*length);
	if (!output)
		return NULL;

	for (i = 0; i < length; i++) {
		double x = ((double) i * 65535.) / (double) (length - 1);
		uint16_fract_t input = floor(x + .5);
		output[i] = input;
	}
	return output;
}

static uint16_t *build_pow_table(float gamma, int length)
{
	int i;
	uint16_t *output = malloc(sizeof(uint16_t)*length);
	if (!output)
		return NULL;

	for (i = 0; i < length; i++) {
		uint16_fract_t result;
		double x = ((double) i) / (double) (length - 1);
		x = pow(x, gamma);
                //XXX turn this conversion into a function
		result = floor(x*65535. + .5);
		output[i] = result;
	}
	return output;
}

static float clamp_float(float a)
{
	if (a > 1.)
		return 1.;
	else if (a < 0)
		return 0;
	else
		return a;
}

#if 0
static void qcms_transform_data_rgb_out_pow(qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length)
{
	int i;
	float (*mat)[4] = transform->matrix;
	for (i=0; i<length; i++) {
		unsigned char device_r = *src++;
		unsigned char device_g = *src++;
		unsigned char device_b = *src++;

		float linear_r = transform->input_gamma_table_r[device_r];
		float linear_g = transform->input_gamma_table_g[device_g];
		float linear_b = transform->input_gamma_table_b[device_b];

		float out_linear_r = mat[0][0]*linear_r + mat[1][0]*linear_g + mat[2][0]*linear_b;
		float out_linear_g = mat[0][1]*linear_r + mat[1][1]*linear_g + mat[2][1]*linear_b;
		float out_linear_b = mat[0][2]*linear_r + mat[1][2]*linear_g + mat[2][2]*linear_b;

		float out_device_r = pow(out_linear_r, transform->out_gamma_r);
		float out_device_g = pow(out_linear_g, transform->out_gamma_g);
		float out_device_b = pow(out_linear_b, transform->out_gamma_b);

		*dest++ = clamp_u8(255*out_device_r);
		*dest++ = clamp_u8(255*out_device_g);
		*dest++ = clamp_u8(255*out_device_b);
	}
}
#endif

static void qcms_transform_data_gray_out_lut(qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length)
{
	int i;
	for (i = 0; i < length; i++) {
		float out_device_r, out_device_g, out_device_b;
		unsigned char device = *src++;

		float linear = transform->input_gamma_table_gray[device];

                out_device_r = lut_interp_linear(linear, transform->output_gamma_lut_r, transform->output_gamma_lut_r_length);
		out_device_g = lut_interp_linear(linear, transform->output_gamma_lut_g, transform->output_gamma_lut_g_length);
		out_device_b = lut_interp_linear(linear, transform->output_gamma_lut_b, transform->output_gamma_lut_b_length);

		*dest++ = clamp_u8(out_device_r*255);
		*dest++ = clamp_u8(out_device_g*255);
		*dest++ = clamp_u8(out_device_b*255);
	}
}

/* Alpha is not corrected.
   A rationale for this is found in Alvy Ray's "Should Alpha Be Nonlinear If
   RGB Is?" Tech Memo 17 (December 14, 1998).
	See: ftp://ftp.alvyray.com/Acrobat/17_Nonln.pdf
*/

static void qcms_transform_data_graya_out_lut(qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length)
{
	int i;
	for (i = 0; i < length; i++) {
		float out_device_r, out_device_g, out_device_b;
		unsigned char device = *src++;
		unsigned char alpha = *src++;

		float linear = transform->input_gamma_table_gray[device];

                out_device_r = lut_interp_linear(linear, transform->output_gamma_lut_r, transform->output_gamma_lut_r_length);
		out_device_g = lut_interp_linear(linear, transform->output_gamma_lut_g, transform->output_gamma_lut_g_length);
		out_device_b = lut_interp_linear(linear, transform->output_gamma_lut_b, transform->output_gamma_lut_b_length);

		*dest++ = clamp_u8(out_device_r*255);
		*dest++ = clamp_u8(out_device_g*255);
		*dest++ = clamp_u8(out_device_b*255);
		*dest++ = alpha;
	}
}


static void qcms_transform_data_gray_out_precache(qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length)
{
	int i;
	for (i = 0; i < length; i++) {
		unsigned char device = *src++;
		uint16_t gray;

		float linear = transform->input_gamma_table_gray[device];

		/* we could round here... */
		gray = linear * 65535.;

		*dest++ = transform->output_table_r->data[gray];
		*dest++ = transform->output_table_g->data[gray];
		*dest++ = transform->output_table_b->data[gray];
	}
}

static void qcms_transform_data_graya_out_precache(qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length)
{
	int i;
	for (i = 0; i < length; i++) {
		unsigned char device = *src++;
		unsigned char alpha = *src++;
		uint16_t gray;

		float linear = transform->input_gamma_table_gray[device];

		/* we could round here... */
		gray = linear * 65535.;

		*dest++ = transform->output_table_r->data[gray];
		*dest++ = transform->output_table_g->data[gray];
		*dest++ = transform->output_table_b->data[gray];
		*dest++ = alpha;
	}
}

static const ALIGN float floatScale = 65536.0f;
static const ALIGN float * const floatScaleAddr = &floatScale; // Win32 ASM doesn't know how to take addressOf inline

static const ALIGN float clampMaxValue = ((float) (65536 - 1)) / 65536.0f;

#ifdef X86
#if 0
#include <emmintrin.h>
void qcms_transform_data_rgb_out_lut_sse_intrin(qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length)
{
	int i;
	float (*mat)[4] = transform->matrix;
        char input_back[32];
	/* Ensure we have a buffer that's 16 byte aligned regardless of the original
	 * stack alignment. We can't use __attribute__((aligned(16))) or __declspec(align(32))
	 * because they don't work on stack variables. gcc 4.4 does do the right thing 
	 * on x86 but that's too new for us right now. For more info: gcc bug #16660 */
        float *input = (float*)(((uintptr_t)&input_back[16]) & ~0xf);
        /* share input and output locations to save having to keep the
         * locations in separate registers */
        uint32_t* output = (uint32_t*)input;
	for (i=0; i<length; i++) {
		const float *clampMax = &clampMaxValue;

		unsigned char device_r = *src++;
		unsigned char device_g = *src++;
		unsigned char device_b = *src++;

		__m128 xmm1 = _mm_load_ps(mat[0]);
		__m128 xmm2 = _mm_load_ps(mat[1]);
		__m128 xmm3 = _mm_load_ps(mat[2]);

		__m128 vec_r = _mm_load_ss(&transform->input_gamma_table_r[device_r]);
		vec_r = _mm_shuffle_ps(vec_r, vec_r, 0);
		__m128 vec_g = _mm_load_ss(&transform->input_gamma_table_r[device_g]);
		vec_g = _mm_shuffle_ps(vec_g, vec_g, 0);
		__m128 vec_b = _mm_load_ss(&transform->input_gamma_table_r[device_b]);
		vec_b = _mm_shuffle_ps(vec_b, vec_b, 0);

		vec_r = _mm_mul_ps(vec_r, xmm1);
		vec_g = _mm_mul_ps(vec_g, xmm2);
		vec_b = _mm_mul_ps(vec_b, xmm3);

		vec_r = _mm_add_ps(vec_r, _mm_add_ps(vec_g, vec_b));

		__m128 max = _mm_load_ss(&clampMax);
		max = _mm_shuffle_ps(max, max, 0);
		__m128 min = _mm_setzero_ps();

		vec_r = _mm_max_ps(min, vec_r);
		vec_r = _mm_min_ps(max, vec_r);

		__m128 scale = _mm_load_ss(&floatScale);
		scale = _mm_shuffle_ps(scale, scale, 0);
		__m128 result = _mm_mul_ps(vec_r, scale);

		__m128i out = _mm_cvtps_epi32(result);
		_mm_store_si128((__m128i*)input, out);

		*dest++ = transform->output_table_r->data[output[0]];
		*dest++ = transform->output_table_g->data[output[1]];
		*dest++ = transform->output_table_b->data[output[2]];
	}
}
#endif
static void qcms_transform_data_rgb_out_lut_sse(qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length)
{
	int i;
	float (*mat)[4] = transform->matrix;
        char input_back[32];
	/* Ensure we have a buffer that's 16 byte aligned regardless of the original
	 * stack alignment. We can't use __attribute__((aligned(16))) or __declspec(align(32))
	 * because they don't work on stack variables. gcc 4.4 does do the right thing 
	 * on x86 but that's too new for us right now. For more info: gcc bug #16660 */
        float *input = (float*)(((uintptr_t)&input_back[16]) & ~0xf);
        /* share input and output locations to save having to keep the
         * locations in separate registers */
        uint32_t* output = (uint32_t*)input;

        input[3] = 0; /* initialize the unused 4th element of the input array */
	for (i = 0; i < length; i++) {
		const float *clampMax = &clampMaxValue;

		unsigned char device_r = *src++;
		unsigned char device_g = *src++;
		unsigned char device_b = *src++;

		input[0] = transform->input_gamma_table_r[device_r];
		input[1] = transform->input_gamma_table_g[device_g];
		input[2] = transform->input_gamma_table_b[device_b];

#ifdef __GNUC__
		__asm(
                      "movaps (%0), %%xmm1;\n\t"          // Move the first matrix column to xmm1
                      "movaps 16(%0), %%xmm2;\n\t"        // Move the second matrix column to xmm2
                      "movaps 32(%0), %%xmm3;\n\t"        // move the third matrix column to xmm3
                      "movaps (%3), %%xmm0;\n\t"        // Move the vector to xmm0

                                                          // Note - We have to copy and then shuffle because of the weird
                                                          // semantics of shufps
                                                          //
                      "movaps %%xmm0, %%xmm4;\n\t"        // Copy the vector to xmm4
                      "shufps $0, %%xmm4, %%xmm4;\n\t"    // Shuffle to repeat the first vector element repeated 4 times
                      "mulps %%xmm4, %%xmm1;\n\t"         // Multiply the first vector element by the first matrix column
                      "movaps %%xmm0, %%xmm5; \n\t"       // Copy the vector to xmm5
                      "shufps $0x55, %%xmm5, %%xmm5;\n\t" // Shuffle to repeat the second vector element repeated 4 times
                      "mulps %%xmm5, %%xmm2;\n\t"         // Multiply the second vector element by the seccond matrix column 
                      "movaps %%xmm0, %%xmm6;\n\t"        // Copy the vector to xmm6
                      "shufps $0xAA, %%xmm6, %%xmm6;\n\t" // Shuffle to repeat the third vector element repeated 4 times
                      "mulps %%xmm6, %%xmm3;\n\t"         // Multiply the third vector element by the third matrix column

                      "addps %%xmm3, %%xmm2;\n\t"         // Sum (second + third) columns
                      "addps %%xmm2, %%xmm1;\n\t"         // Sum ((second + third) + first) columns

                      "movss (%1), %%xmm7;\n\t"        // load the floating point representation of 65535/65536 
                      "shufps $0, %%xmm7, %%xmm7;\n\t" // move it into all of the four slots
                      "minps %%xmm7, %%xmm1;\n\t"      // clamp the vector to 1.0 max
                      "xorps %%xmm6, %%xmm6;\n\t"       // get us cleared bitpatern, which is 0.0f
                      "maxps %%xmm6, %%xmm1;\n\t"      // clamp the vector to 0.0 min
                      "movss (%2), %%xmm5;\n\t"        // load the floating point scale factor
                      "shufps $0, %%xmm5, %%xmm5;\n\t" // put it in all four slots
                      "mulps %%xmm5, %%xmm1;\n\t"      // multiply by the scale factor
                      "cvtps2dq %%xmm1, %%xmm1;\n\t"   // convert to integers
                      "movdqa %%xmm1, (%3);\n\t"       // store

                      : 
                      : "r" (mat), "r" (clampMax), "r" (&floatScale), "r" (input)
                      : "memory"
/* older versions of gcc don't know about these registers so only include them as constraints
   if gcc knows about them */
#ifdef __SSE2__
                        , "%xmm0", "%xmm1", "%xmm2", "%xmm3", "%xmm4", "%xmm5", "%xmm6", "%xmm7"
#endif
                      );
#else
                __asm {
                      mov      eax, mat
                      mov      ecx, clampMax
                      mov      edx, floatScaleAddr
		      mov      ebx, input

                      movaps   xmm1, [eax]
                      movaps   xmm2, [eax + 16]
                      movaps   xmm3, [eax + 32]
                      movaps   xmm0, [ebx]

                      movaps   xmm4, xmm0
                      shufps   xmm4, xmm4, 0
                      mulps    xmm1, xmm4
                      movaps   xmm5, xmm0
                      shufps   xmm5, xmm5, 0x55
                      mulps    xmm2, xmm5
                      movaps   xmm6, xmm0
                      shufps   xmm6, xmm6, 0xAA
                      mulps    xmm3, xmm6

                      addps    xmm2, xmm3
                      addps    xmm1, xmm2

                      movss    xmm7, [ecx]
                      shufps   xmm7, xmm7, 0
                      minps    xmm1, xmm7
                      xorps    xmm6, xmm6
                      maxps    xmm1, xmm6
                      movss    xmm5, [edx]
                      shufps   xmm5, xmm5, 0
                      mulps    xmm1, xmm5
                      cvtps2dq xmm1, xmm1
                      movdqa   [ebx], xmm1
                }
#endif

		*dest++ = transform->output_table_r->data[output[0]];
		*dest++ = transform->output_table_g->data[output[1]];
		*dest++ = transform->output_table_b->data[output[2]];
	}
}

static void qcms_transform_data_rgba_out_lut_sse(qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length)
{
	int i;
	float (*mat)[4] = transform->matrix;
        char input_back[32];
	/* align input on 16 byte boundary */
        float *input = (float*)(((uintptr_t)&input_back[16]) & ~0xf);
        /* share input and output locations to save having to keep the
         * locations in separate registers */
        uint32_t* output = (uint32_t*)input;
	for (i = 0; i < length; i++) {
		const float *clampMax = &clampMaxValue;

		unsigned char device_r = *src++;
		unsigned char device_g = *src++;
		unsigned char device_b = *src++;
		unsigned char alpha = *src++;

		input[0] = transform->input_gamma_table_r[device_r];
		input[1] = transform->input_gamma_table_g[device_g];
		input[2] = transform->input_gamma_table_b[device_b];

#ifdef __GNUC__
		__asm(
                      "movaps (%0), %%xmm1;\n\t"          // Move the first matrix column to xmm1
                      "movaps 16(%0), %%xmm2;\n\t"        // Move the second matrix column to xmm2
                      "movaps 32(%0), %%xmm3;\n\t"        // move the third matrix column to xmm3
                      "movaps (%3), %%xmm0;\n\t"        // Move the vector to xmm0

                                                          // Note - We have to copy and then shuffle because of the weird
                                                          // semantics of shufps
                                                          //
                      "movaps %%xmm0, %%xmm4;\n\t"        // Copy the vector to xmm4
                      "shufps $0, %%xmm4, %%xmm4;\n\t"    // Shuffle to repeat the first vector element repeated 4 times
                      "mulps %%xmm4, %%xmm1;\n\t"         // Multiply the first vector element by the first matrix column
                      "movaps %%xmm0, %%xmm5; \n\t"       // Copy the vector to xmm5
                      "shufps $0x55, %%xmm5, %%xmm5;\n\t" // Shuffle to repeat the second vector element repeated 4 times
                      "mulps %%xmm5, %%xmm2;\n\t"         // Multiply the second vector element by the seccond matrix column 
                      "movaps %%xmm0, %%xmm6;\n\t"        // Copy the vector to xmm6
                      "shufps $0xAA, %%xmm6, %%xmm6;\n\t" // Shuffle to repeat the third vector element repeated 4 times
                      "mulps %%xmm6, %%xmm3;\n\t"         // Multiply the third vector element by the third matrix column

                      "addps %%xmm3, %%xmm2;\n\t"         // Sum (second + third) columns
                      "addps %%xmm2, %%xmm1;\n\t"         // Sum ((second + third) + first) columns

                      "movss (%1), %%xmm7;\n\t"        // load the floating point representation of 65535/65536 
                      "shufps $0, %%xmm7, %%xmm7;\n\t" // move it into all of the four slots
                      "minps %%xmm7, %%xmm1;\n\t"      // clamp the vector to 1.0 max
                      "xorps %%xmm6, %%xmm6;\n\t"       // get us cleared bitpatern, which is 0.0f
                      "maxps %%xmm6, %%xmm1;\n\t"      // clamp the vector to 0.0 min
                      "movss (%2), %%xmm5;\n\t"        // load the floating point scale factor
                      "shufps $0, %%xmm5, %%xmm5;\n\t" // put it in all four slots
                      "mulps %%xmm5, %%xmm1;\n\t"      // multiply by the scale factor
                      "cvtps2dq %%xmm1, %%xmm1;\n\t"   // convert to integers
                      "movdqa %%xmm1, (%3);\n\t"       // store

                      : 
                      : "r" (mat), "r" (clampMax), "r" (&floatScale), "r" (input)
                      : "memory"
/* older versions of gcc don't know about these registers so only include them as constraints
   if gcc knows about them */
#ifdef __SSE2__
                        , "%xmm0", "%xmm1", "%xmm2", "%xmm3", "%xmm4", "%xmm5", "%xmm6", "%xmm7"
#endif
                      );
#else
                __asm {
                      mov      eax, mat
                      mov      ecx, clampMax
                      mov      edx, floatScaleAddr
		      mov      ebx, input

                      movaps   xmm1, [eax]
                      movaps   xmm2, [eax + 16]
                      movaps   xmm3, [eax + 32]
                      movaps   xmm0, [ebx]

                      movaps   xmm4, xmm0
                      shufps   xmm4, xmm4, 0
                      mulps    xmm1, xmm4
                      movaps   xmm5, xmm0
                      shufps   xmm5, xmm5, 0x55
                      mulps    xmm2, xmm5
                      movaps   xmm6, xmm0
                      shufps   xmm6, xmm6, 0xAA
                      mulps    xmm3, xmm6

                      addps    xmm2, xmm3
                      addps    xmm1, xmm2

                      movss    xmm7, [ecx]
                      shufps   xmm7, xmm7, 0
                      minps    xmm1, xmm7
                      xorps    xmm6, xmm6
                      maxps    xmm1, xmm6
                      movss    xmm5, [edx]
                      shufps   xmm5, xmm5, 0
                      mulps    xmm1, xmm5
                      cvtps2dq xmm1, xmm1
                      movdqa   [ebx], xmm1
                }
#endif

		*dest++ = transform->output_table_r->data[output[0]];
		*dest++ = transform->output_table_g->data[output[1]];
		*dest++ = transform->output_table_b->data[output[2]];
		*dest++ = alpha;
	}
}
#endif

static void qcms_transform_data_rgb_out_lut_precache(qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length)
{
	int i;
	float (*mat)[4] = transform->matrix;
	for (i = 0; i < length; i++) {
		unsigned char device_r = *src++;
		unsigned char device_g = *src++;
		unsigned char device_b = *src++;
		uint16_t r, g, b;

		float linear_r = transform->input_gamma_table_r[device_r];
		float linear_g = transform->input_gamma_table_g[device_g];
		float linear_b = transform->input_gamma_table_b[device_b];

		float out_linear_r = mat[0][0]*linear_r + mat[1][0]*linear_g + mat[2][0]*linear_b;
		float out_linear_g = mat[0][1]*linear_r + mat[1][1]*linear_g + mat[2][1]*linear_b;
		float out_linear_b = mat[0][2]*linear_r + mat[1][2]*linear_g + mat[2][2]*linear_b;

		out_linear_r = clamp_float(out_linear_r);
		out_linear_g = clamp_float(out_linear_g);
		out_linear_b = clamp_float(out_linear_b);

		/* we could round here... */
		r = out_linear_r * 65535.;
		g = out_linear_g * 65535.;
		b = out_linear_b * 65535.;

		*dest++ = transform->output_table_r->data[r];
		*dest++ = transform->output_table_g->data[g];
		*dest++ = transform->output_table_b->data[b];
	}
}

static void qcms_transform_data_rgba_out_lut_precache(qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length)
{
	int i;
	float (*mat)[4] = transform->matrix;
	for (i = 0; i < length; i++) {
		unsigned char device_r = *src++;
		unsigned char device_g = *src++;
		unsigned char device_b = *src++;
		unsigned char alpha = *src++;
		uint16_t r, g, b;

		float linear_r = transform->input_gamma_table_r[device_r];
		float linear_g = transform->input_gamma_table_g[device_g];
		float linear_b = transform->input_gamma_table_b[device_b];

		float out_linear_r = mat[0][0]*linear_r + mat[1][0]*linear_g + mat[2][0]*linear_b;
		float out_linear_g = mat[0][1]*linear_r + mat[1][1]*linear_g + mat[2][1]*linear_b;
		float out_linear_b = mat[0][2]*linear_r + mat[1][2]*linear_g + mat[2][2]*linear_b;

		out_linear_r = clamp_float(out_linear_r);
		out_linear_g = clamp_float(out_linear_g);
		out_linear_b = clamp_float(out_linear_b);

		/* we could round here... */
		r = out_linear_r * 65535.;
		g = out_linear_g * 65535.;
		b = out_linear_b * 65535.;

		*dest++ = transform->output_table_r->data[r];
		*dest++ = transform->output_table_g->data[g];
		*dest++ = transform->output_table_b->data[b];
		*dest++ = alpha;
	}
}

static void qcms_transform_data_rgb_out_lut(qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length)
{
	int i;
	float (*mat)[4] = transform->matrix;
	for (i = 0; i < length; i++) {
		unsigned char device_r = *src++;
		unsigned char device_g = *src++;
		unsigned char device_b = *src++;
		float out_device_r, out_device_g, out_device_b;

		float linear_r = transform->input_gamma_table_r[device_r];
		float linear_g = transform->input_gamma_table_g[device_g];
		float linear_b = transform->input_gamma_table_b[device_b];

		float out_linear_r = mat[0][0]*linear_r + mat[1][0]*linear_g + mat[2][0]*linear_b;
		float out_linear_g = mat[0][1]*linear_r + mat[1][1]*linear_g + mat[2][1]*linear_b;
		float out_linear_b = mat[0][2]*linear_r + mat[1][2]*linear_g + mat[2][2]*linear_b;

		out_linear_r = clamp_float(out_linear_r);
		out_linear_g = clamp_float(out_linear_g);
		out_linear_b = clamp_float(out_linear_b);

		out_device_r = lut_interp_linear(out_linear_r, transform->output_gamma_lut_r, transform->output_gamma_lut_r_length);
		out_device_g = lut_interp_linear(out_linear_g, transform->output_gamma_lut_g, transform->output_gamma_lut_g_length);
		out_device_b = lut_interp_linear(out_linear_b, transform->output_gamma_lut_b, transform->output_gamma_lut_b_length);

		*dest++ = clamp_u8(out_device_r*255);
		*dest++ = clamp_u8(out_device_g*255);
		*dest++ = clamp_u8(out_device_b*255);
	}
}

static void qcms_transform_data_rgba_out_lut(qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length)
{
	int i;
	float (*mat)[4] = transform->matrix;
	for (i = 0; i < length; i++) {
		unsigned char device_r = *src++;
		unsigned char device_g = *src++;
		unsigned char device_b = *src++;
		unsigned char alpha = *src++;
		float out_device_r, out_device_g, out_device_b;

		float linear_r = transform->input_gamma_table_r[device_r];
		float linear_g = transform->input_gamma_table_g[device_g];
		float linear_b = transform->input_gamma_table_b[device_b];

		float out_linear_r = mat[0][0]*linear_r + mat[1][0]*linear_g + mat[2][0]*linear_b;
		float out_linear_g = mat[0][1]*linear_r + mat[1][1]*linear_g + mat[2][1]*linear_b;
		float out_linear_b = mat[0][2]*linear_r + mat[1][2]*linear_g + mat[2][2]*linear_b;

		out_linear_r = clamp_float(out_linear_r);
		out_linear_g = clamp_float(out_linear_g);
		out_linear_b = clamp_float(out_linear_b);

		out_device_r = lut_interp_linear(out_linear_r, transform->output_gamma_lut_r, transform->output_gamma_lut_r_length);
		out_device_g = lut_interp_linear(out_linear_g, transform->output_gamma_lut_g, transform->output_gamma_lut_g_length);
		out_device_b = lut_interp_linear(out_linear_b, transform->output_gamma_lut_b, transform->output_gamma_lut_b_length);

		*dest++ = clamp_u8(out_device_r*255);
		*dest++ = clamp_u8(out_device_g*255);
		*dest++ = clamp_u8(out_device_b*255);
		*dest++ = alpha;
	}
}

#if 0
static void qcms_transform_data_rgb_out_linear(qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length)
{
	int i;
	float (*mat)[4] = transform->matrix;
	for (i = 0; i < length; i++) {
		unsigned char device_r = *src++;
		unsigned char device_g = *src++;
		unsigned char device_b = *src++;

		float linear_r = transform->input_gamma_table_r[device_r];
		float linear_g = transform->input_gamma_table_g[device_g];
		float linear_b = transform->input_gamma_table_b[device_b];

		float out_linear_r = mat[0][0]*linear_r + mat[1][0]*linear_g + mat[2][0]*linear_b;
		float out_linear_g = mat[0][1]*linear_r + mat[1][1]*linear_g + mat[2][1]*linear_b;
		float out_linear_b = mat[0][2]*linear_r + mat[1][2]*linear_g + mat[2][2]*linear_b;

		*dest++ = clamp_u8(out_linear_r*255);
		*dest++ = clamp_u8(out_linear_g*255);
		*dest++ = clamp_u8(out_linear_b*255);
	}
}
#endif

static struct precache_output *precache_reference(struct precache_output *p)
{
	p->ref_count++;
	return p;
}

static struct precache_output *precache_create()
{
	struct precache_output *p = malloc(sizeof(struct precache_output));
	if (p)
		p->ref_count = 1;
	return p;
}

void precache_release(struct precache_output *p)
{
	if (--p->ref_count == 0) {
		free(p);
	}
}

#ifdef HAS_POSIX_MEMALIGN
static qcms_transform *transform_alloc(void)
{
	qcms_transform *t;
	if (!posix_memalign(&t, 16, sizeof(*t))) {
		return t;
	} else {
		return NULL;
	}
}
static void transform_free(qcms_transform *t)
{
	free(t);
}
#else
static qcms_transform *transform_alloc(void)
{
	/* transform needs to be aligned on a 16byte boundrary */
	char *original_block = calloc(sizeof(qcms_transform) + sizeof(void*) + 16, 1);
	/* make room for a pointer to the block returned by calloc */
	void *transform_start = original_block + sizeof(void*);
	/* align transform_start */
	qcms_transform *transform_aligned = (qcms_transform*)(((uintptr_t)transform_start + 15) & ~0xf);

	/* store a pointer to the block returned by calloc so that we can free it later */
	void **(original_block_ptr) = (void**)transform_aligned;
	if (!original_block)
		return NULL;
	original_block_ptr--;
	*original_block_ptr = original_block;

	return transform_aligned;
}
static void transform_free(qcms_transform *t)
{
	/* get at the pointer to the unaligned block returned by calloc */
	void **p = (void**)t;
	p--;
	free(*p);
}
#endif

void qcms_transform_release(qcms_transform *t)
{
	/* ensure we only free the gamma tables once even if there are
	 * multiple references to the same data */

	if (t->output_table_r)
		precache_release(t->output_table_r);
	if (t->output_table_g)
		precache_release(t->output_table_g);
	if (t->output_table_b)
		precache_release(t->output_table_b);

	free(t->input_gamma_table_r);
	if (t->input_gamma_table_g != t->input_gamma_table_r)
		free(t->input_gamma_table_g);
	if (t->input_gamma_table_g != t->input_gamma_table_r &&
	    t->input_gamma_table_g != t->input_gamma_table_b)
		free(t->input_gamma_table_b);

	free(t->input_gamma_table_gray);

	free(t->output_gamma_lut_r);
	free(t->output_gamma_lut_g);
	free(t->output_gamma_lut_b);

	transform_free(t);
}

static void compute_precache_pow(uint8_t *output, float gamma)
{
	uint32_t v = 0;
	for (v = 0; v <= 0xffff; v++) {
		//XXX: don't do integer/float conversion... and round?
		output[v] = 255. * pow(v/65535., gamma);
	}
}

void compute_precache_lut(uint8_t *output, uint16_t *table, int length)
{
	uint32_t v = 0;
	for (v = 0; v <= 0xffff; v++) {
		//XXX: don't do integer/float conversion... round?
		output[v] = lut_interp_linear16(v, table, length) >> 8;
	}
}

void compute_precache_linear(uint8_t *output)
{
	uint32_t v = 0;
	for (v = 0; v <= 0xffff; v++) {
		//XXX: round?
		output[v] = v >> 8;
	}
}

qcms_bool compute_precache(struct curveType *trc, uint8_t *output)
{
	if (trc->count == 0) {
		compute_precache_linear(output);
	} else if (trc->count == 1) {
		compute_precache_pow(output, 1./u8Fixed8Number_to_float(trc->data[0]));
	} else {
		uint16_t *inverted;
		int inverted_size = trc->count;
		//XXX: the choice of a minimum of 256 here is not backed by any theory, measurement or data, however it is what lcms uses.
		// the maximum number we would need is 65535 because that's the accuracy used for computing the precache table
		if (inverted_size < 256)
			inverted_size = 256;

		inverted = invert_lut(trc->data, trc->count, inverted_size);
		if (!inverted)
			return false;
		compute_precache_lut(output, inverted, inverted_size);
		free(inverted);
	}
	return true;
}


// Determine if we can build with SSE2 (this was partly copied from jmorecfg.h in
// mozilla/jpeg)
 // -------------------------------------------------------------------------
#if defined(_M_IX86) && defined(_MSC_VER)
#define HAS_CPUID
/* Get us a CPUID function. Avoid clobbering EBX because sometimes it's the PIC
   register - I'm not sure if that ever happens on windows, but cpuid isn't
   on the critical path so we just preserve the register to be safe and to be
   consistent with the non-windows version. */
static void cpuid(uint32_t fxn, uint32_t *a, uint32_t *b, uint32_t *c, uint32_t *d) {
       uint32_t a_, b_, c_, d_;
       __asm {
              xchg   ebx, esi
              mov    eax, fxn
              cpuid
              mov    a_, eax
              mov    b_, ebx
              mov    c_, ecx
              mov    d_, edx
              xchg   ebx, esi
       }
       *a = a_;
       *b = b_;
       *c = c_;
       *d = d_;
}
#elif defined(__GNUC__) && defined(__i386__)
#define HAS_CPUID
/* Get us a CPUID function. We can't use ebx because it's the PIC register on
   some platforms, so we use ESI instead and save ebx to avoid clobbering it. */
static void cpuid(uint32_t fxn, uint32_t *a, uint32_t *b, uint32_t *c, uint32_t *d) {

	uint32_t a_, b_, c_, d_;
       __asm__ __volatile__ ("xchgl %%ebx, %%esi; cpuid; xchgl %%ebx, %%esi;" 
                             : "=a" (a_), "=S" (b_), "=c" (c_), "=d" (d_) : "a" (fxn));
	   *a = a_;
	   *b = b_;
	   *c = c_;
	   *d = d_;
}
#endif

// -------------------------Runtime SSE2 Detection-----------------------------

#define SSE2_EDX_MASK (1UL << 26)
static qcms_bool sse2_available(void)
{
#ifdef HAS_CPUID
       static int has_sse2 = -1;
       uint32_t a, b, c, d;
       uint32_t function = 0x00000001;

       if (has_sse2 == -1) {
              has_sse2 = 0;
	      cpuid(function, &a, &b, &c, &d);
              if (d & SSE2_EDX_MASK)
                     has_sse2 = 1;
              else
                     has_sse2 = 0;
       }

       return has_sse2;
#endif
       return false;
}

void build_output_lut(struct curveType *trc,
		uint16_t **output_gamma_lut, size_t *output_gamma_lut_length)
{
	if (trc->count == 0) {
		*output_gamma_lut = build_linear_table(4096);
		*output_gamma_lut_length = 4096;
	} else if (trc->count == 1) {
		float gamma = 1./u8Fixed8Number_to_float(trc->data[0]);
		*output_gamma_lut = build_pow_table(gamma, 4096);
		*output_gamma_lut_length = 4096;
	} else {
		//XXX: the choice of a minimum of 256 here is not backed by any theory, measurement or data, however it is what lcms uses.
		*output_gamma_lut_length = trc->count;
		if (*output_gamma_lut_length < 256)
			*output_gamma_lut_length = 256;

		*output_gamma_lut = invert_lut(trc->data, trc->count, *output_gamma_lut_length);
	}

}

void qcms_profile_precache_output_transform(qcms_profile *profile)
{
	/* we only support precaching on rgb profiles */
	if (profile->color_space != RGB_SIGNATURE)
		return;

	if (!profile->output_table_r) {
		profile->output_table_r = precache_create();
		if (profile->output_table_r &&
				!compute_precache(profile->redTRC, profile->output_table_r->data)) {
			precache_release(profile->output_table_r);
			profile->output_table_r = NULL;
		}
	}
	if (!profile->output_table_g) {
		profile->output_table_g = precache_create();
		if (profile->output_table_g &&
				!compute_precache(profile->greenTRC, profile->output_table_g->data)) {
			precache_release(profile->output_table_g);
			profile->output_table_g = NULL;
		}
	}
	if (!profile->output_table_b) {
		profile->output_table_b = precache_create();
		if (profile->output_table_b &&
				!compute_precache(profile->blueTRC, profile->output_table_b->data)) {
			precache_release(profile->output_table_g);
			profile->output_table_g = NULL;
		}
	}
}

#define NO_MEM_TRANSFORM NULL

qcms_transform* qcms_transform_create(
		qcms_profile *in, qcms_data_type in_type,
		qcms_profile* out, qcms_data_type out_type,
		qcms_intent intent)
{
	bool precache = false;

        qcms_transform *transform = transform_alloc();
        if (!transform) {
		return NULL;
	}
	if (out_type != QCMS_DATA_RGB_8 &&
                out_type != QCMS_DATA_RGBA_8) {
            assert(0 && "output type");
	    free(transform);
            return NULL;
        }

	if (out->output_table_r &&
			out->output_table_g &&
			out->output_table_b) {
		precache = true;
	}

	if (precache) {
		transform->output_table_r = precache_reference(out->output_table_r);
		transform->output_table_g = precache_reference(out->output_table_g);
		transform->output_table_b = precache_reference(out->output_table_b);
	} else {
		build_output_lut(out->redTRC, &transform->output_gamma_lut_r, &transform->output_gamma_lut_r_length);
		build_output_lut(out->greenTRC, &transform->output_gamma_lut_g, &transform->output_gamma_lut_g_length);
		build_output_lut(out->blueTRC, &transform->output_gamma_lut_b, &transform->output_gamma_lut_b_length);
		if (!transform->output_gamma_lut_r || !transform->output_gamma_lut_g || !transform->output_gamma_lut_b) {
			qcms_transform_release(transform);
			return NO_MEM_TRANSFORM;
		}
	}

        if (in->color_space == RGB_SIGNATURE) {
            struct matrix in_matrix, out_matrix, result;

            if (in_type != QCMS_DATA_RGB_8 &&
                    in_type != QCMS_DATA_RGBA_8){
                assert(0 && "input type");
		free(transform);
                return NULL;
            }
	    if (precache) {
#ifdef X86
		    if (sse2_available()) {
			    if (in_type == QCMS_DATA_RGB_8)
				    transform->transform_fn = qcms_transform_data_rgb_out_lut_sse;
			    else
				    transform->transform_fn = qcms_transform_data_rgba_out_lut_sse;

		    } else
#endif
		    {
			    if (in_type == QCMS_DATA_RGB_8)
				    transform->transform_fn = qcms_transform_data_rgb_out_lut_precache;
			    else
				    transform->transform_fn = qcms_transform_data_rgba_out_lut_precache;
		    }
	    } else {
		    if (in_type == QCMS_DATA_RGB_8)
			    transform->transform_fn = qcms_transform_data_rgb_out_lut;
		    else
			    transform->transform_fn = qcms_transform_data_rgba_out_lut;
	    }

            //XXX: avoid duplicating tables if we can
            transform->input_gamma_table_r = build_input_gamma_table(in->redTRC);
            transform->input_gamma_table_g = build_input_gamma_table(in->greenTRC);
            transform->input_gamma_table_b = build_input_gamma_table(in->blueTRC);

	    if (!transform->input_gamma_table_r || !transform->input_gamma_table_g || !transform->input_gamma_table_b) {
		    qcms_transform_release(transform);
		    return NO_MEM_TRANSFORM;
	    }

            /* build combined colorant matrix */
            in_matrix = build_colorant_matrix(in);
            out_matrix = build_colorant_matrix(out);
            out_matrix = matrix_invert(out_matrix);
            if (out_matrix.invalid) {
                qcms_transform_release(transform);
                return NULL;
            }
            result = matrix_multiply(out_matrix, in_matrix);

            /* store the results in column major mode
             * this makes doing the multiplication with sse easier */
            transform->matrix[0][0] = result.m[0][0];
            transform->matrix[1][0] = result.m[0][1];
            transform->matrix[2][0] = result.m[0][2];
            transform->matrix[0][1] = result.m[1][0];
            transform->matrix[1][1] = result.m[1][1];
            transform->matrix[2][1] = result.m[1][2];
            transform->matrix[0][2] = result.m[2][0];
            transform->matrix[1][2] = result.m[2][1];
            transform->matrix[2][2] = result.m[2][2];

        } else if (in->color_space == GRAY_SIGNATURE) {
            if (in_type != QCMS_DATA_GRAY_8 &&
                    in_type != QCMS_DATA_GRAYA_8){
                assert(0 && "input type");
		free(transform);
                return NULL;
            }

            transform->input_gamma_table_gray = build_input_gamma_table(in->grayTRC);
	    if (!transform->input_gamma_table_gray) {
		    qcms_transform_release(transform);
		    return NO_MEM_TRANSFORM;
	    }

	    if (precache) {
		    if (in_type == QCMS_DATA_GRAY_8) {
			    transform->transform_fn = qcms_transform_data_gray_out_precache;
		    } else {
			    transform->transform_fn = qcms_transform_data_graya_out_precache;
		    }
	    } else {
		    if (in_type == QCMS_DATA_GRAY_8) {
			    transform->transform_fn = qcms_transform_data_gray_out_lut;
		    } else {
			    transform->transform_fn = qcms_transform_data_graya_out_lut;
		    }
	    }
	} else {
		assert(0 && "unexpected colorspace");
	}
	return transform;
}

void qcms_transform_data(qcms_transform *transform, void *src, void *dest, size_t length)
{
	transform->transform_fn(transform, src, dest, length);
}