ViewVC Help
View File | Revision Log | Show Annotations | View Changeset | Root Listing
root/REPOS_ERICCA/magic/outil/src/robustpredicates.cpp
Revision: 481
Committed: Tue Jan 28 16:10:58 2014 UTC (11 years, 5 months ago) by francois
Original Path: magic/lib/outil/src/robustpredicates.cpp
File size: 180636 byte(s)
Log Message:
unification de la facon d'ecrire les fichiers tous en minuscules

File Contents

# Content
1 /*****************************************************************************/
2 /* */
3 /* Routines for Arbitrary Precision Floating-point Arithmetic */
4 /* and Fast Robust Geometric Predicates */
5 /* (predicates.c) */
6 /* */
7 /* May 18, 1996 */
8 /* */
9 /* Placed in the public domain by */
10 /* Jonathan Richard Shewchuk */
11 /* School of Computer Science */
12 /* Carnegie Mellon University */
13 /* 5000 Forbes Avenue */
14 /* Pittsburgh, Pennsylvania 15213-3891 */
15 /* jrs@cs.cmu.edu */
16 /* */
17 /* This file contains C implementation of algorithms for exact addition */
18 /* and multiplication of floating-point numbers, and predicates for */
19 /* robustly performing the orientation and incircle tests used in */
20 /* computational geometry. The algorithms and underlying theory are */
21 /* described in Jonathan Richard Shewchuk. "Adaptive Precision Floating- */
22 /* Point Arithmetic and Fast Robust Geometric Predicates." Technical */
23 /* Report CMU-CS-96-140, School of Computer Science, Carnegie Mellon */
24 /* University, Pittsburgh, Pennsylvania, May 1996. (Submitted to */
25 /* Discrete & Computational Geometry.) */
26 /* */
27 /* This file, the paper listed above, and other information are available */
28 /* from the Web page http://www.cs.cmu.edu/~quake/robust.html . */
29 /* */
30 /*****************************************************************************/
31
32 /*****************************************************************************/
33 /* */
34 /* Using this code: */
35 /* */
36 /* First, read the short or long version of the paper (from the Web page */
37 /* above). */
38 /* */
39 /* Be sure to call exactinit() once, before calling any of the arithmetic */
40 /* functions or geometric predicates. Also be sure to turn on the */
41 /* optimizer when compiling this file. */
42 /* */
43 /* */
44 /* Several geometric predicates are defined. Their parameters are all */
45 /* points. Each point is an array of two or three floating-point */
46 /* numbers. The geometric predicates, described in the papers, are */
47 /* */
48 /* orient2d(pa, pb, pc) */
49 /* orient2dfast(pa, pb, pc) */
50 /* orient3d(pa, pb, pc, pd) */
51 /* orient3dfast(pa, pb, pc, pd) */
52 /* incircle(pa, pb, pc, pd) */
53 /* incirclefast(pa, pb, pc, pd) */
54 /* insphere(pa, pb, pc, pd, pe) */
55 /* inspherefast(pa, pb, pc, pd, pe) */
56 /* */
57 /* Those with suffix "fast" are approximate, non-robust versions. Those */
58 /* without the suffix are adaptive precision, robust versions. There */
59 /* are also versions with the suffices "exact" and "slow", which are */
60 /* non-adaptive, exact arithmetic versions, which I use only for timings */
61 /* in my arithmetic papers. */
62 /* */
63 /* */
64 /* An expansion is represented by an array of floating-point numbers, */
65 /* sorted from smallest to largest magnitude (possibly with interspersed */
66 /* zeros). The length of each expansion is stored as a separate integer, */
67 /* and each arithmetic function returns an integer which is the length */
68 /* of the expansion it created. */
69 /* */
70 /* Several arithmetic functions are defined. Their parameters are */
71 /* */
72 /* e, f Input expansions */
73 /* elen, flen Lengths of input expansions (must be >= 1) */
74 /* h Output expansion */
75 /* b Input scalar */
76 /* */
77 /* The arithmetic functions are */
78 /* */
79 /* grow_expansion(elen, e, b, h) */
80 /* grow_expansion_zeroelim(elen, e, b, h) */
81 /* expansion_sum(elen, e, flen, f, h) */
82 /* expansion_sum_zeroelim1(elen, e, flen, f, h) */
83 /* expansion_sum_zeroelim2(elen, e, flen, f, h) */
84 /* fast_expansion_sum(elen, e, flen, f, h) */
85 /* fast_expansion_sum_zeroelim(elen, e, flen, f, h) */
86 /* linear_expansion_sum(elen, e, flen, f, h) */
87 /* linear_expansion_sum_zeroelim(elen, e, flen, f, h) */
88 /* scale_expansion(elen, e, b, h) */
89 /* scale_expansion_zeroelim(elen, e, b, h) */
90 /* compress(elen, e, h) */
91 /* */
92 /* All of these are described in the long version of the paper; some are */
93 /* described in the short version. All return an integer that is the */
94 /* length of h. Those with suffix _zeroelim perform zero elimination, */
95 /* and are recommended over their counterparts. The procedure */
96 /* fast_expansion_sum_zeroelim() (or linear_expansion_sum_zeroelim() on */
97 /* processors that do not use the round-to-even tiebreaking rule) is */
98 /* recommended over expansion_sum_zeroelim(). Each procedure has a */
99 /* little note next to it (in the code below) that tells you whether or */
100 /* not the output expansion may be the same array as one of the input */
101 /* expansions. */
102 /* */
103 /* */
104 /* If you look around below, you'll also find macros for a bunch of */
105 /* simple unrolled arithmetic operations, and procedures for printing */
106 /* expansions (commented out because they don't work with all C */
107 /* compilers) and for generating random floating-point numbers whose */
108 /* significand bits are all random. Most of the macros have undocumented */
109 /* requirements that certain of their parameters should not be the same */
110 /* variable; for safety, better to make sure all the parameters are */
111 /* distinct variables. Feel free to send email to jrs@cs.cmu.edu if you */
112 /* have questions. */
113 /* */
114 /*****************************************************************************/
115
116 #include <stdio.h>
117 #include <stdlib.h>
118 #include <math.h>
119 #ifdef CPU86
120 #include <float.h>
121 #endif /* CPU86 */
122 #ifdef LINUX
123 #include <fpu_control.h>
124 #endif /* LINUX */
125
126 namespace robustPredicates
127 {
128
129 /* On some machines, the exact arithmetic routines might be defeated by the */
130 /* use of internal extended precision floating-point registers. Sometimes */
131 /* this problem can be fixed by defining certain values to be volatile, */
132 /* thus forcing them to be stored to memory and rounded off. This isn't */
133 /* a great solution, though, as it slows the arithmetic down. */
134 /* */
135 /* To try this out, write "#define INEXACT volatile" below. Normally, */
136 /* however, INEXACT should be defined to be nothing. ("#define INEXACT".) */
137
138 #define INEXACT /* Nothing */
139 /* #define INEXACT volatile */
140
141 #define REAL double /* float or double */
142 #define REALPRINT doubleprint
143 #define REALRAND doublerand
144 #define NARROWRAND narrowdoublerand
145 #define UNIFORMRAND uniformdoublerand
146
147 /* Which of the following two methods of finding the absolute values is */
148 /* fastest is compiler-dependent. A few compilers can inline and optimize */
149 /* the fabs() call; but most will incur the overhead of a function call, */
150 /* which is disastrously slow. A faster way on IEEE machines might be to */
151 /* mask the appropriate bit, but that's difficult to do in C. */
152
153 #define Absolute(a) ((a) >= 0.0 ? (a) : -(a))
154 /* #define Absolute(a) fabs(a) */
155
156 /* Many of the operations are broken up into two pieces, a main part that */
157 /* performs an approximate operation, and a "tail" that computes the */
158 /* roundoff error of that operation. */
159 /* */
160 /* The operations Fast_Two_Sum(), Fast_Two_Diff(), Two_Sum(), Two_Diff(), */
161 /* Split(), and Two_Product() are all implemented as described in the */
162 /* reference. Each of these macros requires certain variables to be */
163 /* defined in the calling routine. The variables `bvirt', `c', `abig', */
164 /* `_i', `_j', `_k', `_l', `_m', and `_n' are declared `INEXACT' because */
165 /* they store the result of an operation that may incur roundoff error. */
166 /* The input parameter `x' (or the highest numbered `x_' parameter) must */
167 /* also be declared `INEXACT'. */
168
169 #define Fast_Two_Sum_Tail(a, b, x, y) \
170 bvirt = x - a; \
171 y = b - bvirt
172
173 #define Fast_Two_Sum(a, b, x, y) \
174 x = (REAL) (a + b); \
175 Fast_Two_Sum_Tail(a, b, x, y)
176
177 #define Fast_Two_Diff_Tail(a, b, x, y) \
178 bvirt = a - x; \
179 y = bvirt - b
180
181 #define Fast_Two_Diff(a, b, x, y) \
182 x = (REAL) (a - b); \
183 Fast_Two_Diff_Tail(a, b, x, y)
184
185 #define Two_Sum_Tail(a, b, x, y) \
186 bvirt = (REAL) (x - a); \
187 avirt = x - bvirt; \
188 bround = b - bvirt; \
189 around = a - avirt; \
190 y = around + bround
191
192 #define Two_Sum(a, b, x, y) \
193 x = (REAL) (a + b); \
194 Two_Sum_Tail(a, b, x, y)
195
196 #define Two_Diff_Tail(a, b, x, y) \
197 bvirt = (REAL) (a - x); \
198 avirt = x + bvirt; \
199 bround = bvirt - b; \
200 around = a - avirt; \
201 y = around + bround
202
203 #define Two_Diff(a, b, x, y) \
204 x = (REAL) (a - b); \
205 Two_Diff_Tail(a, b, x, y)
206
207 #define Split(a, ahi, alo) \
208 c = (REAL) (splitter * a); \
209 abig = (REAL) (c - a); \
210 ahi = c - abig; \
211 alo = a - ahi
212
213 #define Two_Product_Tail(a, b, x, y) \
214 Split(a, ahi, alo); \
215 Split(b, bhi, blo); \
216 err1 = x - (ahi * bhi); \
217 err2 = err1 - (alo * bhi); \
218 err3 = err2 - (ahi * blo); \
219 y = (alo * blo) - err3
220
221 #define Two_Product(a, b, x, y) \
222 x = (REAL) (a * b); \
223 Two_Product_Tail(a, b, x, y)
224
225 /* Two_Product_Presplit() is Two_Product() where one of the inputs has */
226 /* already been split. Avoids redundant splitting. */
227
228 #define Two_Product_Presplit(a, b, bhi, blo, x, y) \
229 x = (REAL) (a * b); \
230 Split(a, ahi, alo); \
231 err1 = x - (ahi * bhi); \
232 err2 = err1 - (alo * bhi); \
233 err3 = err2 - (ahi * blo); \
234 y = (alo * blo) - err3
235
236 /* Two_Product_2Presplit() is Two_Product() where both of the inputs have */
237 /* already been split. Avoids redundant splitting. */
238
239 #define Two_Product_2Presplit(a, ahi, alo, b, bhi, blo, x, y) \
240 x = (REAL) (a * b); \
241 err1 = x - (ahi * bhi); \
242 err2 = err1 - (alo * bhi); \
243 err3 = err2 - (ahi * blo); \
244 y = (alo * blo) - err3
245
246 /* Square() can be done more quickly than Two_Product(). */
247
248 #define Square_Tail(a, x, y) \
249 Split(a, ahi, alo); \
250 err1 = x - (ahi * ahi); \
251 err3 = err1 - ((ahi + ahi) * alo); \
252 y = (alo * alo) - err3
253
254 #define Square(a, x, y) \
255 x = (REAL) (a * a); \
256 Square_Tail(a, x, y)
257
258 /* Macros for summing expansions of various fixed lengths. These are all */
259 /* unrolled versions of Expansion_Sum(). */
260
261 #define Two_One_Sum(a1, a0, b, x2, x1, x0) \
262 Two_Sum(a0, b , _i, x0); \
263 Two_Sum(a1, _i, x2, x1)
264
265 #define Two_One_Diff(a1, a0, b, x2, x1, x0) \
266 Two_Diff(a0, b , _i, x0); \
267 Two_Sum( a1, _i, x2, x1)
268
269 #define Two_Two_Sum(a1, a0, b1, b0, x3, x2, x1, x0) \
270 Two_One_Sum(a1, a0, b0, _j, _0, x0); \
271 Two_One_Sum(_j, _0, b1, x3, x2, x1)
272
273 #define Two_Two_Diff(a1, a0, b1, b0, x3, x2, x1, x0) \
274 Two_One_Diff(a1, a0, b0, _j, _0, x0); \
275 Two_One_Diff(_j, _0, b1, x3, x2, x1)
276
277 #define Four_One_Sum(a3, a2, a1, a0, b, x4, x3, x2, x1, x0) \
278 Two_One_Sum(a1, a0, b , _j, x1, x0); \
279 Two_One_Sum(a3, a2, _j, x4, x3, x2)
280
281 #define Four_Two_Sum(a3, a2, a1, a0, b1, b0, x5, x4, x3, x2, x1, x0) \
282 Four_One_Sum(a3, a2, a1, a0, b0, _k, _2, _1, _0, x0); \
283 Four_One_Sum(_k, _2, _1, _0, b1, x5, x4, x3, x2, x1)
284
285 #define Four_Four_Sum(a3, a2, a1, a0, b4, b3, b1, b0, x7, x6, x5, x4, x3, x2, \
286 x1, x0) \
287 Four_Two_Sum(a3, a2, a1, a0, b1, b0, _l, _2, _1, _0, x1, x0); \
288 Four_Two_Sum(_l, _2, _1, _0, b4, b3, x7, x6, x5, x4, x3, x2)
289
290 #define Eight_One_Sum(a7, a6, a5, a4, a3, a2, a1, a0, b, x8, x7, x6, x5, x4, \
291 x3, x2, x1, x0) \
292 Four_One_Sum(a3, a2, a1, a0, b , _j, x3, x2, x1, x0); \
293 Four_One_Sum(a7, a6, a5, a4, _j, x8, x7, x6, x5, x4)
294
295 #define Eight_Two_Sum(a7, a6, a5, a4, a3, a2, a1, a0, b1, b0, x9, x8, x7, \
296 x6, x5, x4, x3, x2, x1, x0) \
297 Eight_One_Sum(a7, a6, a5, a4, a3, a2, a1, a0, b0, _k, _6, _5, _4, _3, _2, \
298 _1, _0, x0); \
299 Eight_One_Sum(_k, _6, _5, _4, _3, _2, _1, _0, b1, x9, x8, x7, x6, x5, x4, \
300 x3, x2, x1)
301
302 #define Eight_Four_Sum(a7, a6, a5, a4, a3, a2, a1, a0, b4, b3, b1, b0, x11, \
303 x10, x9, x8, x7, x6, x5, x4, x3, x2, x1, x0) \
304 Eight_Two_Sum(a7, a6, a5, a4, a3, a2, a1, a0, b1, b0, _l, _6, _5, _4, _3, \
305 _2, _1, _0, x1, x0); \
306 Eight_Two_Sum(_l, _6, _5, _4, _3, _2, _1, _0, b4, b3, x11, x10, x9, x8, \
307 x7, x6, x5, x4, x3, x2)
308
309 /* Macros for multiplying expansions of various fixed lengths. */
310
311 #define Two_One_Product(a1, a0, b, x3, x2, x1, x0) \
312 Split(b, bhi, blo); \
313 Two_Product_Presplit(a0, b, bhi, blo, _i, x0); \
314 Two_Product_Presplit(a1, b, bhi, blo, _j, _0); \
315 Two_Sum(_i, _0, _k, x1); \
316 Fast_Two_Sum(_j, _k, x3, x2)
317
318 #define Four_One_Product(a3, a2, a1, a0, b, x7, x6, x5, x4, x3, x2, x1, x0) \
319 Split(b, bhi, blo); \
320 Two_Product_Presplit(a0, b, bhi, blo, _i, x0); \
321 Two_Product_Presplit(a1, b, bhi, blo, _j, _0); \
322 Two_Sum(_i, _0, _k, x1); \
323 Fast_Two_Sum(_j, _k, _i, x2); \
324 Two_Product_Presplit(a2, b, bhi, blo, _j, _0); \
325 Two_Sum(_i, _0, _k, x3); \
326 Fast_Two_Sum(_j, _k, _i, x4); \
327 Two_Product_Presplit(a3, b, bhi, blo, _j, _0); \
328 Two_Sum(_i, _0, _k, x5); \
329 Fast_Two_Sum(_j, _k, x7, x6)
330
331 #define Two_Two_Product(a1, a0, b1, b0, x7, x6, x5, x4, x3, x2, x1, x0) \
332 Split(a0, a0hi, a0lo); \
333 Split(b0, bhi, blo); \
334 Two_Product_2Presplit(a0, a0hi, a0lo, b0, bhi, blo, _i, x0); \
335 Split(a1, a1hi, a1lo); \
336 Two_Product_2Presplit(a1, a1hi, a1lo, b0, bhi, blo, _j, _0); \
337 Two_Sum(_i, _0, _k, _1); \
338 Fast_Two_Sum(_j, _k, _l, _2); \
339 Split(b1, bhi, blo); \
340 Two_Product_2Presplit(a0, a0hi, a0lo, b1, bhi, blo, _i, _0); \
341 Two_Sum(_1, _0, _k, x1); \
342 Two_Sum(_2, _k, _j, _1); \
343 Two_Sum(_l, _j, _m, _2); \
344 Two_Product_2Presplit(a1, a1hi, a1lo, b1, bhi, blo, _j, _0); \
345 Two_Sum(_i, _0, _n, _0); \
346 Two_Sum(_1, _0, _i, x2); \
347 Two_Sum(_2, _i, _k, _1); \
348 Two_Sum(_m, _k, _l, _2); \
349 Two_Sum(_j, _n, _k, _0); \
350 Two_Sum(_1, _0, _j, x3); \
351 Two_Sum(_2, _j, _i, _1); \
352 Two_Sum(_l, _i, _m, _2); \
353 Two_Sum(_1, _k, _i, x4); \
354 Two_Sum(_2, _i, _k, x5); \
355 Two_Sum(_m, _k, x7, x6)
356
357 /* An expansion of length two can be squared more quickly than finding the */
358 /* product of two different expansions of length two, and the result is */
359 /* guaranteed to have no more than six (rather than eight) components. */
360
361 #define Two_Square(a1, a0, x5, x4, x3, x2, x1, x0) \
362 Square(a0, _j, x0); \
363 _0 = a0 + a0; \
364 Two_Product(a1, _0, _k, _1); \
365 Two_One_Sum(_k, _1, _j, _l, _2, x1); \
366 Square(a1, _j, _1); \
367 Two_Two_Sum(_j, _1, _l, _2, x5, x4, x3, x2)
368
369 /* splitter = 2^ceiling(p / 2) + 1. Used to split floats in half. */
370 static REAL splitter;
371 static REAL epsilon; /* = 2^(-p). Used to estimate roundoff errors. */
372 /* A set of coefficients used to calculate maximum roundoff errors. */
373 static REAL resulterrbound;
374 static REAL ccwerrboundA, ccwerrboundB, ccwerrboundC;
375 static REAL o3derrboundA, o3derrboundB, o3derrboundC;
376 static REAL iccerrboundA, iccerrboundB, iccerrboundC;
377 static REAL isperrboundA, isperrboundB, isperrboundC;
378
379 /*****************************************************************************/
380 /* */
381 /* doubleprint() Print the bit representation of a double. */
382 /* */
383 /* Useful for debugging exact arithmetic routines. */
384 /* */
385 /*****************************************************************************/
386
387 /*
388 void doubleprint(number)
389 double number;
390 {
391 unsigned long long no;
392 unsigned long long sign, expo;
393 int exponent;
394 int i, bottomi;
395
396 no = *(unsigned long long *) &number;
397 sign = no & 0x8000000000000000ll;
398 expo = (no >> 52) & 0x7ffll;
399 exponent = (int) expo;
400 exponent = exponent - 1023;
401 if (sign) {
402 printf("-");
403 } else {
404 printf(" ");
405 }
406 if (exponent == -1023) {
407 printf(
408 "0.0000000000000000000000000000000000000000000000000000_ ( )");
409 } else {
410 printf("1.");
411 bottomi = -1;
412 for (i = 0; i < 52; i++) {
413 if (no & 0x0008000000000000ll) {
414 printf("1");
415 bottomi = i;
416 } else {
417 printf("0");
418 }
419 no <<= 1;
420 }
421 printf("_%d (%d)", exponent, exponent - 1 - bottomi);
422 }
423 }
424 */
425
426 /*****************************************************************************/
427 /* */
428 /* floatprint() Print the bit representation of a float. */
429 /* */
430 /* Useful for debugging exact arithmetic routines. */
431 /* */
432 /*****************************************************************************/
433
434 /*
435 void floatprint(number)
436 float number;
437 {
438 unsigned no;
439 unsigned sign, expo;
440 int exponent;
441 int i, bottomi;
442
443 no = *(unsigned *) &number;
444 sign = no & 0x80000000;
445 expo = (no >> 23) & 0xff;
446 exponent = (int) expo;
447 exponent = exponent - 127;
448 if (sign) {
449 printf("-");
450 } else {
451 printf(" ");
452 }
453 if (exponent == -127) {
454 printf("0.00000000000000000000000_ ( )");
455 } else {
456 printf("1.");
457 bottomi = -1;
458 for (i = 0; i < 23; i++) {
459 if (no & 0x00400000) {
460 printf("1");
461 bottomi = i;
462 } else {
463 printf("0");
464 }
465 no <<= 1;
466 }
467 printf("_%3d (%3d)", exponent, exponent - 1 - bottomi);
468 }
469 }
470 */
471
472 /*****************************************************************************/
473 /* */
474 /* expansion_print() Print the bit representation of an expansion. */
475 /* */
476 /* Useful for debugging exact arithmetic routines. */
477 /* */
478 /*****************************************************************************/
479
480 /*
481 void expansion_print(elen, e)
482 int elen;
483 REAL *e;
484 {
485 int i;
486
487 for (i = elen - 1; i >= 0; i--) {
488 REALPRINT(e[i]);
489 if (i > 0) {
490 printf(" +\n");
491 } else {
492 printf("\n");
493 }
494 }
495 }
496 */
497
498 /*****************************************************************************/
499 /* */
500 /* doublerand() Generate a double with random 53-bit significand and a */
501 /* random exponent in [0, 511]. */
502 /* */
503 /*****************************************************************************/
504
505 /*
506 double doublerand()
507 {
508 double result;
509 double expo;
510 long a, b, c;
511 long i;
512
513 a = random();
514 b = random();
515 c = random();
516 result = (double) (a - 1073741824) * 8388608.0 + (double) (b >> 8);
517 for (i = 512, expo = 2; i <= 131072; i *= 2, expo = expo * expo) {
518 if (c & i) {
519 result *= expo;
520 }
521 }
522 return result;
523 }
524 */
525
526 /*****************************************************************************/
527 /* */
528 /* narrowdoublerand() Generate a double with random 53-bit significand */
529 /* and a random exponent in [0, 7]. */
530 /* */
531 /*****************************************************************************/
532
533 /*
534 double narrowdoublerand()
535 {
536 double result;
537 double expo;
538 long a, b, c;
539 long i;
540
541 a = random();
542 b = random();
543 c = random();
544 result = (double) (a - 1073741824) * 8388608.0 + (double) (b >> 8);
545 for (i = 512, expo = 2; i <= 2048; i *= 2, expo = expo * expo) {
546 if (c & i) {
547 result *= expo;
548 }
549 }
550 return result;
551 }
552 */
553
554 /*****************************************************************************/
555 /* */
556 /* uniformdoublerand() Generate a double with random 53-bit significand. */
557 /* */
558 /*****************************************************************************/
559
560 /*
561 double uniformdoublerand()
562 {
563 double result;
564 long a, b;
565
566 a = random();
567 b = random();
568 result = (double) (a - 1073741824) * 8388608.0 + (double) (b >> 8);
569 return result;
570 }
571 */
572
573 /*****************************************************************************/
574 /* */
575 /* floatrand() Generate a float with random 24-bit significand and a */
576 /* random exponent in [0, 63]. */
577 /* */
578 /*****************************************************************************/
579
580 /*
581 float floatrand()
582 {
583 float result;
584 float expo;
585 long a, c;
586 long i;
587
588 a = random();
589 c = random();
590 result = (float) ((a - 1073741824) >> 6);
591 for (i = 512, expo = 2; i <= 16384; i *= 2, expo = expo * expo) {
592 if (c & i) {
593 result *= expo;
594 }
595 }
596 return result;
597 }
598 */
599
600 /*****************************************************************************/
601 /* */
602 /* narrowfloatrand() Generate a float with random 24-bit significand and */
603 /* a random exponent in [0, 7]. */
604 /* */
605 /*****************************************************************************/
606
607 /*
608 float narrowfloatrand()
609 {
610 float result;
611 float expo;
612 long a, c;
613 long i;
614
615 a = random();
616 c = random();
617 result = (float) ((a - 1073741824) >> 6);
618 for (i = 512, expo = 2; i <= 2048; i *= 2, expo = expo * expo) {
619 if (c & i) {
620 result *= expo;
621 }
622 }
623 return result;
624 }
625 */
626
627 /*****************************************************************************/
628 /* */
629 /* uniformfloatrand() Generate a float with random 24-bit significand. */
630 /* */
631 /*****************************************************************************/
632
633 /*
634 float uniformfloatrand()
635 {
636 float result;
637 long a;
638
639 a = random();
640 result = (float) ((a - 1073741824) >> 6);
641 return result;
642 }
643 */
644
645 /*****************************************************************************/
646 /* */
647 /* exactinit() Initialize the variables used for exact arithmetic. */
648 /* */
649 /* `epsilon' is the largest power of two such that 1.0 + epsilon = 1.0 in */
650 /* floating-point arithmetic. `epsilon' bounds the relative roundoff */
651 /* error. It is used for floating-point error analysis. */
652 /* */
653 /* `splitter' is used to split floating-point numbers into two half- */
654 /* length significands for exact multiplication. */
655 /* */
656 /* I imagine that a highly optimizing compiler might be too smart for its */
657 /* own good, and somehow cause this routine to fail, if it pretends that */
658 /* floating-point arithmetic is too much like real arithmetic. */
659 /* */
660 /* Don't change this routine unless you fully understand it. */
661 /* */
662 /*****************************************************************************/
663
664 REAL exactinit()
665 {
666 REAL half;
667 REAL check, lastcheck;
668 int every_other;
669 #ifdef LINUX
670 int cword;
671 #endif /* LINUX */
672
673 #ifdef CPU86
674 #ifdef SINGLE
675 _control87(_PC_24, _MCW_PC); /* Set FPU control word for single precision. */
676 #else /* not SINGLE */
677 _control87(_PC_53, _MCW_PC); /* Set FPU control word for double precision. */
678 #endif /* not SINGLE */
679 #endif /* CPU86 */
680 #ifdef LINUX
681 #ifdef SINGLE
682 /* cword = 4223; */
683 cword = 4210; /* set FPU control word for single precision */
684 #else /* not SINGLE */
685 /* cword = 4735; */
686 cword = 4722; /* set FPU control word for double precision */
687 #endif /* not SINGLE */
688 _FPU_SETCW(cword);
689 #endif /* LINUX */
690
691 every_other = 1;
692 half = 0.5;
693 epsilon = 1.0;
694 splitter = 1.0;
695 check = 1.0;
696 /* Repeatedly divide `epsilon' by two until it is too small to add to */
697 /* one without causing roundoff. (Also check if the sum is equal to */
698 /* the previous sum, for machines that round up instead of using exact */
699 /* rounding. Not that this library will work on such machines anyway. */
700 do {
701 lastcheck = check;
702 epsilon *= half;
703 if (every_other) {
704 splitter *= 2.0;
705 }
706 every_other = !every_other;
707 check = 1.0 + epsilon;
708 } while ((check != 1.0) && (check != lastcheck));
709 splitter += 1.0;
710
711 /* Error bounds for orientation and incircle tests. */
712 resulterrbound = (3.0 + 8.0 * epsilon) * epsilon;
713 ccwerrboundA = (3.0 + 16.0 * epsilon) * epsilon;
714 ccwerrboundB = (2.0 + 12.0 * epsilon) * epsilon;
715 ccwerrboundC = (9.0 + 64.0 * epsilon) * epsilon * epsilon;
716 o3derrboundA = (7.0 + 56.0 * epsilon) * epsilon;
717 o3derrboundB = (3.0 + 28.0 * epsilon) * epsilon;
718 o3derrboundC = (26.0 + 288.0 * epsilon) * epsilon * epsilon;
719 iccerrboundA = (10.0 + 96.0 * epsilon) * epsilon;
720 iccerrboundB = (4.0 + 48.0 * epsilon) * epsilon;
721 iccerrboundC = (44.0 + 576.0 * epsilon) * epsilon * epsilon;
722 isperrboundA = (16.0 + 224.0 * epsilon) * epsilon;
723 isperrboundB = (5.0 + 72.0 * epsilon) * epsilon;
724 isperrboundC = (71.0 + 1408.0 * epsilon) * epsilon * epsilon;
725
726 return epsilon; /* Added by H. Si 30 Juli, 2004. */
727 }
728
729 /*****************************************************************************/
730 /* */
731 /* grow_expansion() Add a scalar to an expansion. */
732 /* */
733 /* Sets h = e + b. See the long version of my paper for details. */
734 /* */
735 /* Maintains the nonoverlapping property. If round-to-even is used (as */
736 /* with IEEE 754), maintains the strongly nonoverlapping and nonadjacent */
737 /* properties as well. (That is, if e has one of these properties, so */
738 /* will h.) */
739 /* */
740 /*****************************************************************************/
741
742 int grow_expansion(int elen, REAL *e, REAL b, REAL *h)
743 /* e and h can be the same. */
744 {
745 REAL Q;
746 INEXACT REAL Qnew;
747 int eindex;
748 REAL enow;
749 INEXACT REAL bvirt;
750 REAL avirt, bround, around;
751
752 Q = b;
753 for (eindex = 0; eindex < elen; eindex++) {
754 enow = e[eindex];
755 Two_Sum(Q, enow, Qnew, h[eindex]);
756 Q = Qnew;
757 }
758 h[eindex] = Q;
759 return eindex + 1;
760 }
761
762 /*****************************************************************************/
763 /* */
764 /* grow_expansion_zeroelim() Add a scalar to an expansion, eliminating */
765 /* zero components from the output expansion. */
766 /* */
767 /* Sets h = e + b. See the long version of my paper for details. */
768 /* */
769 /* Maintains the nonoverlapping property. If round-to-even is used (as */
770 /* with IEEE 754), maintains the strongly nonoverlapping and nonadjacent */
771 /* properties as well. (That is, if e has one of these properties, so */
772 /* will h.) */
773 /* */
774 /*****************************************************************************/
775
776 int grow_expansion_zeroelim(int elen, REAL *e, REAL b, REAL *h)
777 /* e and h can be the same. */
778 {
779 REAL Q, hh;
780 INEXACT REAL Qnew;
781 int eindex, hindex;
782 REAL enow;
783 INEXACT REAL bvirt;
784 REAL avirt, bround, around;
785
786 hindex = 0;
787 Q = b;
788 for (eindex = 0; eindex < elen; eindex++) {
789 enow = e[eindex];
790 Two_Sum(Q, enow, Qnew, hh);
791 Q = Qnew;
792 if (hh != 0.0) {
793 h[hindex++] = hh;
794 }
795 }
796 if ((Q != 0.0) || (hindex == 0)) {
797 h[hindex++] = Q;
798 }
799 return hindex;
800 }
801
802 /*****************************************************************************/
803 /* */
804 /* expansion_sum() Sum two expansions. */
805 /* */
806 /* Sets h = e + f. See the long version of my paper for details. */
807 /* */
808 /* Maintains the nonoverlapping property. If round-to-even is used (as */
809 /* with IEEE 754), maintains the nonadjacent property as well. (That is, */
810 /* if e has one of these properties, so will h.) Does NOT maintain the */
811 /* strongly nonoverlapping property. */
812 /* */
813 /*****************************************************************************/
814
815 int expansion_sum(int elen, REAL *e, int flen, REAL *f, REAL *h)
816 /* e and h can be the same, but f and h cannot. */
817 {
818 REAL Q;
819 INEXACT REAL Qnew;
820 int findex, hindex, hlast;
821 REAL hnow;
822 INEXACT REAL bvirt;
823 REAL avirt, bround, around;
824
825 Q = f[0];
826 for (hindex = 0; hindex < elen; hindex++) {
827 hnow = e[hindex];
828 Two_Sum(Q, hnow, Qnew, h[hindex]);
829 Q = Qnew;
830 }
831 h[hindex] = Q;
832 hlast = hindex;
833 for (findex = 1; findex < flen; findex++) {
834 Q = f[findex];
835 for (hindex = findex; hindex <= hlast; hindex++) {
836 hnow = h[hindex];
837 Two_Sum(Q, hnow, Qnew, h[hindex]);
838 Q = Qnew;
839 }
840 h[++hlast] = Q;
841 }
842 return hlast + 1;
843 }
844
845 /*****************************************************************************/
846 /* */
847 /* expansion_sum_zeroelim1() Sum two expansions, eliminating zero */
848 /* components from the output expansion. */
849 /* */
850 /* Sets h = e + f. See the long version of my paper for details. */
851 /* */
852 /* Maintains the nonoverlapping property. If round-to-even is used (as */
853 /* with IEEE 754), maintains the nonadjacent property as well. (That is, */
854 /* if e has one of these properties, so will h.) Does NOT maintain the */
855 /* strongly nonoverlapping property. */
856 /* */
857 /*****************************************************************************/
858
859 int expansion_sum_zeroelim1(int elen, REAL *e, int flen, REAL *f, REAL *h)
860 /* e and h can be the same, but f and h cannot. */
861 {
862 REAL Q;
863 INEXACT REAL Qnew;
864 int index, findex, hindex, hlast;
865 REAL hnow;
866 INEXACT REAL bvirt;
867 REAL avirt, bround, around;
868
869 Q = f[0];
870 for (hindex = 0; hindex < elen; hindex++) {
871 hnow = e[hindex];
872 Two_Sum(Q, hnow, Qnew, h[hindex]);
873 Q = Qnew;
874 }
875 h[hindex] = Q;
876 hlast = hindex;
877 for (findex = 1; findex < flen; findex++) {
878 Q = f[findex];
879 for (hindex = findex; hindex <= hlast; hindex++) {
880 hnow = h[hindex];
881 Two_Sum(Q, hnow, Qnew, h[hindex]);
882 Q = Qnew;
883 }
884 h[++hlast] = Q;
885 }
886 hindex = -1;
887 for (index = 0; index <= hlast; index++) {
888 hnow = h[index];
889 if (hnow != 0.0) {
890 h[++hindex] = hnow;
891 }
892 }
893 if (hindex == -1) {
894 return 1;
895 } else {
896 return hindex + 1;
897 }
898 }
899
900 /*****************************************************************************/
901 /* */
902 /* expansion_sum_zeroelim2() Sum two expansions, eliminating zero */
903 /* components from the output expansion. */
904 /* */
905 /* Sets h = e + f. See the long version of my paper for details. */
906 /* */
907 /* Maintains the nonoverlapping property. If round-to-even is used (as */
908 /* with IEEE 754), maintains the nonadjacent property as well. (That is, */
909 /* if e has one of these properties, so will h.) Does NOT maintain the */
910 /* strongly nonoverlapping property. */
911 /* */
912 /*****************************************************************************/
913
914 int expansion_sum_zeroelim2(int elen, REAL *e, int flen, REAL *f, REAL *h)
915 /* e and h can be the same, but f and h cannot. */
916 {
917 REAL Q, hh;
918 INEXACT REAL Qnew;
919 int eindex, findex, hindex, hlast;
920 REAL enow;
921 INEXACT REAL bvirt;
922 REAL avirt, bround, around;
923
924 hindex = 0;
925 Q = f[0];
926 for (eindex = 0; eindex < elen; eindex++) {
927 enow = e[eindex];
928 Two_Sum(Q, enow, Qnew, hh);
929 Q = Qnew;
930 if (hh != 0.0) {
931 h[hindex++] = hh;
932 }
933 }
934 h[hindex] = Q;
935 hlast = hindex;
936 for (findex = 1; findex < flen; findex++) {
937 hindex = 0;
938 Q = f[findex];
939 for (eindex = 0; eindex <= hlast; eindex++) {
940 enow = h[eindex];
941 Two_Sum(Q, enow, Qnew, hh);
942 Q = Qnew;
943 if (hh != 0) {
944 h[hindex++] = hh;
945 }
946 }
947 h[hindex] = Q;
948 hlast = hindex;
949 }
950 return hlast + 1;
951 }
952
953 /*****************************************************************************/
954 /* */
955 /* fast_expansion_sum() Sum two expansions. */
956 /* */
957 /* Sets h = e + f. See the long version of my paper for details. */
958 /* */
959 /* If round-to-even is used (as with IEEE 754), maintains the strongly */
960 /* nonoverlapping property. (That is, if e is strongly nonoverlapping, h */
961 /* will be also.) Does NOT maintain the nonoverlapping or nonadjacent */
962 /* properties. */
963 /* */
964 /*****************************************************************************/
965
966 int fast_expansion_sum(int elen, REAL *e, int flen, REAL *f, REAL *h)
967 /* h cannot be e or f. */
968 {
969 REAL Q;
970 INEXACT REAL Qnew;
971 INEXACT REAL bvirt;
972 REAL avirt, bround, around;
973 int eindex, findex, hindex;
974 REAL enow, fnow;
975
976 enow = e[0];
977 fnow = f[0];
978 eindex = findex = 0;
979 if ((fnow > enow) == (fnow > -enow)) {
980 Q = enow;
981 enow = e[++eindex];
982 } else {
983 Q = fnow;
984 fnow = f[++findex];
985 }
986 hindex = 0;
987 if ((eindex < elen) && (findex < flen)) {
988 if ((fnow > enow) == (fnow > -enow)) {
989 Fast_Two_Sum(enow, Q, Qnew, h[0]);
990 enow = e[++eindex];
991 } else {
992 Fast_Two_Sum(fnow, Q, Qnew, h[0]);
993 fnow = f[++findex];
994 }
995 Q = Qnew;
996 hindex = 1;
997 while ((eindex < elen) && (findex < flen)) {
998 if ((fnow > enow) == (fnow > -enow)) {
999 Two_Sum(Q, enow, Qnew, h[hindex]);
1000 enow = e[++eindex];
1001 } else {
1002 Two_Sum(Q, fnow, Qnew, h[hindex]);
1003 fnow = f[++findex];
1004 }
1005 Q = Qnew;
1006 hindex++;
1007 }
1008 }
1009 while (eindex < elen) {
1010 Two_Sum(Q, enow, Qnew, h[hindex]);
1011 enow = e[++eindex];
1012 Q = Qnew;
1013 hindex++;
1014 }
1015 while (findex < flen) {
1016 Two_Sum(Q, fnow, Qnew, h[hindex]);
1017 fnow = f[++findex];
1018 Q = Qnew;
1019 hindex++;
1020 }
1021 h[hindex] = Q;
1022 return hindex + 1;
1023 }
1024
1025 /*****************************************************************************/
1026 /* */
1027 /* fast_expansion_sum_zeroelim() Sum two expansions, eliminating zero */
1028 /* components from the output expansion. */
1029 /* */
1030 /* Sets h = e + f. See the long version of my paper for details. */
1031 /* */
1032 /* If round-to-even is used (as with IEEE 754), maintains the strongly */
1033 /* nonoverlapping property. (That is, if e is strongly nonoverlapping, h */
1034 /* will be also.) Does NOT maintain the nonoverlapping or nonadjacent */
1035 /* properties. */
1036 /* */
1037 /*****************************************************************************/
1038
1039 int fast_expansion_sum_zeroelim(int elen, REAL *e, int flen, REAL *f, REAL *h)
1040 /* h cannot be e or f. */
1041 {
1042 REAL Q;
1043 INEXACT REAL Qnew;
1044 INEXACT REAL hh;
1045 INEXACT REAL bvirt;
1046 REAL avirt, bround, around;
1047 int eindex, findex, hindex;
1048 REAL enow, fnow;
1049
1050 enow = e[0];
1051 fnow = f[0];
1052 eindex = findex = 0;
1053 if ((fnow > enow) == (fnow > -enow)) {
1054 Q = enow;
1055 enow = e[++eindex];
1056 } else {
1057 Q = fnow;
1058 fnow = f[++findex];
1059 }
1060 hindex = 0;
1061 if ((eindex < elen) && (findex < flen)) {
1062 if ((fnow > enow) == (fnow > -enow)) {
1063 Fast_Two_Sum(enow, Q, Qnew, hh);
1064 enow = e[++eindex];
1065 } else {
1066 Fast_Two_Sum(fnow, Q, Qnew, hh);
1067 fnow = f[++findex];
1068 }
1069 Q = Qnew;
1070 if (hh != 0.0) {
1071 h[hindex++] = hh;
1072 }
1073 while ((eindex < elen) && (findex < flen)) {
1074 if ((fnow > enow) == (fnow > -enow)) {
1075 Two_Sum(Q, enow, Qnew, hh);
1076 enow = e[++eindex];
1077 } else {
1078 Two_Sum(Q, fnow, Qnew, hh);
1079 fnow = f[++findex];
1080 }
1081 Q = Qnew;
1082 if (hh != 0.0) {
1083 h[hindex++] = hh;
1084 }
1085 }
1086 }
1087 while (eindex < elen) {
1088 Two_Sum(Q, enow, Qnew, hh);
1089 enow = e[++eindex];
1090 Q = Qnew;
1091 if (hh != 0.0) {
1092 h[hindex++] = hh;
1093 }
1094 }
1095 while (findex < flen) {
1096 Two_Sum(Q, fnow, Qnew, hh);
1097 fnow = f[++findex];
1098 Q = Qnew;
1099 if (hh != 0.0) {
1100 h[hindex++] = hh;
1101 }
1102 }
1103 if ((Q != 0.0) || (hindex == 0)) {
1104 h[hindex++] = Q;
1105 }
1106 return hindex;
1107 }
1108
1109 /*****************************************************************************/
1110 /* */
1111 /* linear_expansion_sum() Sum two expansions. */
1112 /* */
1113 /* Sets h = e + f. See either version of my paper for details. */
1114 /* */
1115 /* Maintains the nonoverlapping property. (That is, if e is */
1116 /* nonoverlapping, h will be also.) */
1117 /* */
1118 /*****************************************************************************/
1119
1120 int linear_expansion_sum(int elen, REAL *e, int flen, REAL *f, REAL *h)
1121 /* h cannot be e or f. */
1122 {
1123 REAL Q, q;
1124 INEXACT REAL Qnew;
1125 INEXACT REAL R;
1126 INEXACT REAL bvirt;
1127 REAL avirt, bround, around;
1128 int eindex, findex, hindex;
1129 REAL enow, fnow;
1130 REAL g0;
1131
1132 enow = e[0];
1133 fnow = f[0];
1134 eindex = findex = 0;
1135 if ((fnow > enow) == (fnow > -enow)) {
1136 g0 = enow;
1137 enow = e[++eindex];
1138 } else {
1139 g0 = fnow;
1140 fnow = f[++findex];
1141 }
1142 if ((eindex < elen) && ((findex >= flen)
1143 || ((fnow > enow) == (fnow > -enow)))) {
1144 Fast_Two_Sum(enow, g0, Qnew, q);
1145 enow = e[++eindex];
1146 } else {
1147 Fast_Two_Sum(fnow, g0, Qnew, q);
1148 fnow = f[++findex];
1149 }
1150 Q = Qnew;
1151 for (hindex = 0; hindex < elen + flen - 2; hindex++) {
1152 if ((eindex < elen) && ((findex >= flen)
1153 || ((fnow > enow) == (fnow > -enow)))) {
1154 Fast_Two_Sum(enow, q, R, h[hindex]);
1155 enow = e[++eindex];
1156 } else {
1157 Fast_Two_Sum(fnow, q, R, h[hindex]);
1158 fnow = f[++findex];
1159 }
1160 Two_Sum(Q, R, Qnew, q);
1161 Q = Qnew;
1162 }
1163 h[hindex] = q;
1164 h[hindex + 1] = Q;
1165 return hindex + 2;
1166 }
1167
1168 /*****************************************************************************/
1169 /* */
1170 /* linear_expansion_sum_zeroelim() Sum two expansions, eliminating zero */
1171 /* components from the output expansion. */
1172 /* */
1173 /* Sets h = e + f. See either version of my paper for details. */
1174 /* */
1175 /* Maintains the nonoverlapping property. (That is, if e is */
1176 /* nonoverlapping, h will be also.) */
1177 /* */
1178 /*****************************************************************************/
1179
1180 int linear_expansion_sum_zeroelim(int elen, REAL *e, int flen, REAL *f,
1181 REAL *h)
1182 /* h cannot be e or f. */
1183 {
1184 REAL Q, q, hh;
1185 INEXACT REAL Qnew;
1186 INEXACT REAL R;
1187 INEXACT REAL bvirt;
1188 REAL avirt, bround, around;
1189 int eindex, findex, hindex;
1190 int count;
1191 REAL enow, fnow;
1192 REAL g0;
1193
1194 enow = e[0];
1195 fnow = f[0];
1196 eindex = findex = 0;
1197 hindex = 0;
1198 if ((fnow > enow) == (fnow > -enow)) {
1199 g0 = enow;
1200 enow = e[++eindex];
1201 } else {
1202 g0 = fnow;
1203 fnow = f[++findex];
1204 }
1205 if ((eindex < elen) && ((findex >= flen)
1206 || ((fnow > enow) == (fnow > -enow)))) {
1207 Fast_Two_Sum(enow, g0, Qnew, q);
1208 enow = e[++eindex];
1209 } else {
1210 Fast_Two_Sum(fnow, g0, Qnew, q);
1211 fnow = f[++findex];
1212 }
1213 Q = Qnew;
1214 for (count = 2; count < elen + flen; count++) {
1215 if ((eindex < elen) && ((findex >= flen)
1216 || ((fnow > enow) == (fnow > -enow)))) {
1217 Fast_Two_Sum(enow, q, R, hh);
1218 enow = e[++eindex];
1219 } else {
1220 Fast_Two_Sum(fnow, q, R, hh);
1221 fnow = f[++findex];
1222 }
1223 Two_Sum(Q, R, Qnew, q);
1224 Q = Qnew;
1225 if (hh != 0) {
1226 h[hindex++] = hh;
1227 }
1228 }
1229 if (q != 0) {
1230 h[hindex++] = q;
1231 }
1232 if ((Q != 0.0) || (hindex == 0)) {
1233 h[hindex++] = Q;
1234 }
1235 return hindex;
1236 }
1237
1238 /*****************************************************************************/
1239 /* */
1240 /* scale_expansion() Multiply an expansion by a scalar. */
1241 /* */
1242 /* Sets h = be. See either version of my paper for details. */
1243 /* */
1244 /* Maintains the nonoverlapping property. If round-to-even is used (as */
1245 /* with IEEE 754), maintains the strongly nonoverlapping and nonadjacent */
1246 /* properties as well. (That is, if e has one of these properties, so */
1247 /* will h.) */
1248 /* */
1249 /*****************************************************************************/
1250
1251 int scale_expansion(int elen, REAL *e, REAL b, REAL *h)
1252 /* e and h cannot be the same. */
1253 {
1254 INEXACT REAL Q;
1255 INEXACT REAL sum;
1256 INEXACT REAL product1;
1257 REAL product0;
1258 int eindex, hindex;
1259 REAL enow;
1260 INEXACT REAL bvirt;
1261 REAL avirt, bround, around;
1262 INEXACT REAL c;
1263 INEXACT REAL abig;
1264 REAL ahi, alo, bhi, blo;
1265 REAL err1, err2, err3;
1266
1267 Split(b, bhi, blo);
1268 Two_Product_Presplit(e[0], b, bhi, blo, Q, h[0]);
1269 hindex = 1;
1270 for (eindex = 1; eindex < elen; eindex++) {
1271 enow = e[eindex];
1272 Two_Product_Presplit(enow, b, bhi, blo, product1, product0);
1273 Two_Sum(Q, product0, sum, h[hindex]);
1274 hindex++;
1275 Two_Sum(product1, sum, Q, h[hindex]);
1276 hindex++;
1277 }
1278 h[hindex] = Q;
1279 return elen + elen;
1280 }
1281
1282 /*****************************************************************************/
1283 /* */
1284 /* scale_expansion_zeroelim() Multiply an expansion by a scalar, */
1285 /* eliminating zero components from the */
1286 /* output expansion. */
1287 /* */
1288 /* Sets h = be. See either version of my paper for details. */
1289 /* */
1290 /* Maintains the nonoverlapping property. If round-to-even is used (as */
1291 /* with IEEE 754), maintains the strongly nonoverlapping and nonadjacent */
1292 /* properties as well. (That is, if e has one of these properties, so */
1293 /* will h.) */
1294 /* */
1295 /*****************************************************************************/
1296
1297 int scale_expansion_zeroelim(int elen, REAL *e, REAL b, REAL *h)
1298 /* e and h cannot be the same. */
1299 {
1300 INEXACT REAL Q, sum;
1301 REAL hh;
1302 INEXACT REAL product1;
1303 REAL product0;
1304 int eindex, hindex;
1305 REAL enow;
1306 INEXACT REAL bvirt;
1307 REAL avirt, bround, around;
1308 INEXACT REAL c;
1309 INEXACT REAL abig;
1310 REAL ahi, alo, bhi, blo;
1311 REAL err1, err2, err3;
1312
1313 Split(b, bhi, blo);
1314 Two_Product_Presplit(e[0], b, bhi, blo, Q, hh);
1315 hindex = 0;
1316 if (hh != 0) {
1317 h[hindex++] = hh;
1318 }
1319 for (eindex = 1; eindex < elen; eindex++) {
1320 enow = e[eindex];
1321 Two_Product_Presplit(enow, b, bhi, blo, product1, product0);
1322 Two_Sum(Q, product0, sum, hh);
1323 if (hh != 0) {
1324 h[hindex++] = hh;
1325 }
1326 Fast_Two_Sum(product1, sum, Q, hh);
1327 if (hh != 0) {
1328 h[hindex++] = hh;
1329 }
1330 }
1331 if ((Q != 0.0) || (hindex == 0)) {
1332 h[hindex++] = Q;
1333 }
1334 return hindex;
1335 }
1336
1337 /*****************************************************************************/
1338 /* */
1339 /* compress() Compress an expansion. */
1340 /* */
1341 /* See the long version of my paper for details. */
1342 /* */
1343 /* Maintains the nonoverlapping property. If round-to-even is used (as */
1344 /* with IEEE 754), then any nonoverlapping expansion is converted to a */
1345 /* nonadjacent expansion. */
1346 /* */
1347 /*****************************************************************************/
1348
1349 int compress(int elen, REAL *e, REAL *h)
1350 /* e and h may be the same. */
1351 {
1352 REAL Q, q;
1353 INEXACT REAL Qnew;
1354 int eindex, hindex;
1355 INEXACT REAL bvirt;
1356 REAL enow, hnow;
1357 int top, bottom;
1358
1359 bottom = elen - 1;
1360 Q = e[bottom];
1361 for (eindex = elen - 2; eindex >= 0; eindex--) {
1362 enow = e[eindex];
1363 Fast_Two_Sum(Q, enow, Qnew, q);
1364 if (q != 0) {
1365 h[bottom--] = Qnew;
1366 Q = q;
1367 } else {
1368 Q = Qnew;
1369 }
1370 }
1371 top = 0;
1372 for (hindex = bottom + 1; hindex < elen; hindex++) {
1373 hnow = h[hindex];
1374 Fast_Two_Sum(hnow, Q, Qnew, q);
1375 if (q != 0) {
1376 h[top++] = q;
1377 }
1378 Q = Qnew;
1379 }
1380 h[top] = Q;
1381 return top + 1;
1382 }
1383
1384 /*****************************************************************************/
1385 /* */
1386 /* estimate() Produce a one-word estimate of an expansion's value. */
1387 /* */
1388 /* See either version of my paper for details. */
1389 /* */
1390 /*****************************************************************************/
1391
1392 REAL estimate(int elen, REAL *e)
1393 {
1394 REAL Q;
1395 int eindex;
1396
1397 Q = e[0];
1398 for (eindex = 1; eindex < elen; eindex++) {
1399 Q += e[eindex];
1400 }
1401 return Q;
1402 }
1403
1404 /*****************************************************************************/
1405 /* */
1406 /* orient2dfast() Approximate 2D orientation test. Nonrobust. */
1407 /* orient2dexact() Exact 2D orientation test. Robust. */
1408 /* orient2dslow() Another exact 2D orientation test. Robust. */
1409 /* orient2d() Adaptive exact 2D orientation test. Robust. */
1410 /* */
1411 /* Return a positive value if the points pa, pb, and pc occur */
1412 /* in counterclockwise order; a negative value if they occur */
1413 /* in clockwise order; and zero if they are collinear. The */
1414 /* result is also a rough approximation of twice the signed */
1415 /* area of the triangle defined by the three points. */
1416 /* */
1417 /* Only the first and last routine should be used; the middle two are for */
1418 /* timings. */
1419 /* */
1420 /* The last three use exact arithmetic to ensure a correct answer. The */
1421 /* result returned is the determinant of a matrix. In orient2d() only, */
1422 /* this determinant is computed adaptively, in the sense that exact */
1423 /* arithmetic is used only to the degree it is needed to ensure that the */
1424 /* returned value has the correct sign. Hence, orient2d() is usually quite */
1425 /* fast, but will run more slowly when the input points are collinear or */
1426 /* nearly so. */
1427 /* */
1428 /*****************************************************************************/
1429
1430 REAL orient2dfast(REAL *pa, REAL *pb, REAL *pc)
1431 {
1432 REAL acx, bcx, acy, bcy;
1433
1434 acx = pa[0] - pc[0];
1435 bcx = pb[0] - pc[0];
1436 acy = pa[1] - pc[1];
1437 bcy = pb[1] - pc[1];
1438 return acx * bcy - acy * bcx;
1439 }
1440
1441 REAL orient2dexact(REAL *pa, REAL *pb, REAL *pc)
1442 {
1443 INEXACT REAL axby1, axcy1, bxcy1, bxay1, cxay1, cxby1;
1444 REAL axby0, axcy0, bxcy0, bxay0, cxay0, cxby0;
1445 REAL aterms[4], bterms[4], cterms[4];
1446 INEXACT REAL aterms3, bterms3, cterms3;
1447 REAL v[8], w[12];
1448 int vlength, wlength;
1449
1450 INEXACT REAL bvirt;
1451 REAL avirt, bround, around;
1452 INEXACT REAL c;
1453 INEXACT REAL abig;
1454 REAL ahi, alo, bhi, blo;
1455 REAL err1, err2, err3;
1456 INEXACT REAL _i, _j;
1457 REAL _0;
1458
1459 Two_Product(pa[0], pb[1], axby1, axby0);
1460 Two_Product(pa[0], pc[1], axcy1, axcy0);
1461 Two_Two_Diff(axby1, axby0, axcy1, axcy0,
1462 aterms3, aterms[2], aterms[1], aterms[0]);
1463 aterms[3] = aterms3;
1464
1465 Two_Product(pb[0], pc[1], bxcy1, bxcy0);
1466 Two_Product(pb[0], pa[1], bxay1, bxay0);
1467 Two_Two_Diff(bxcy1, bxcy0, bxay1, bxay0,
1468 bterms3, bterms[2], bterms[1], bterms[0]);
1469 bterms[3] = bterms3;
1470
1471 Two_Product(pc[0], pa[1], cxay1, cxay0);
1472 Two_Product(pc[0], pb[1], cxby1, cxby0);
1473 Two_Two_Diff(cxay1, cxay0, cxby1, cxby0,
1474 cterms3, cterms[2], cterms[1], cterms[0]);
1475 cterms[3] = cterms3;
1476
1477 vlength = fast_expansion_sum_zeroelim(4, aterms, 4, bterms, v);
1478 wlength = fast_expansion_sum_zeroelim(vlength, v, 4, cterms, w);
1479
1480 return w[wlength - 1];
1481 }
1482
1483 REAL orient2dslow(REAL *pa, REAL *pb, REAL *pc)
1484 {
1485 INEXACT REAL acx, acy, bcx, bcy;
1486 REAL acxtail, acytail;
1487 REAL bcxtail, bcytail;
1488 REAL negate, negatetail;
1489 REAL axby[8], bxay[8];
1490 INEXACT REAL axby7, bxay7;
1491 REAL deter[16];
1492 int deterlen;
1493
1494 INEXACT REAL bvirt;
1495 REAL avirt, bround, around;
1496 INEXACT REAL c;
1497 INEXACT REAL abig;
1498 REAL a0hi, a0lo, a1hi, a1lo, bhi, blo;
1499 REAL err1, err2, err3;
1500 INEXACT REAL _i, _j, _k, _l, _m, _n;
1501 REAL _0, _1, _2;
1502
1503 Two_Diff(pa[0], pc[0], acx, acxtail);
1504 Two_Diff(pa[1], pc[1], acy, acytail);
1505 Two_Diff(pb[0], pc[0], bcx, bcxtail);
1506 Two_Diff(pb[1], pc[1], bcy, bcytail);
1507
1508 Two_Two_Product(acx, acxtail, bcy, bcytail,
1509 axby7, axby[6], axby[5], axby[4],
1510 axby[3], axby[2], axby[1], axby[0]);
1511 axby[7] = axby7;
1512 negate = -acy;
1513 negatetail = -acytail;
1514 Two_Two_Product(bcx, bcxtail, negate, negatetail,
1515 bxay7, bxay[6], bxay[5], bxay[4],
1516 bxay[3], bxay[2], bxay[1], bxay[0]);
1517 bxay[7] = bxay7;
1518
1519 deterlen = fast_expansion_sum_zeroelim(8, axby, 8, bxay, deter);
1520
1521 return deter[deterlen - 1];
1522 }
1523
1524 REAL orient2dadapt(REAL *pa, REAL *pb, REAL *pc, REAL detsum)
1525 {
1526 INEXACT REAL acx, acy, bcx, bcy;
1527 REAL acxtail, acytail, bcxtail, bcytail;
1528 INEXACT REAL detleft, detright;
1529 REAL detlefttail, detrighttail;
1530 REAL det, errbound;
1531 REAL B[4], C1[8], C2[12], D[16];
1532 INEXACT REAL B3;
1533 int C1length, C2length, Dlength;
1534 REAL u[4];
1535 INEXACT REAL u3;
1536 INEXACT REAL s1, t1;
1537 REAL s0, t0;
1538
1539 INEXACT REAL bvirt;
1540 REAL avirt, bround, around;
1541 INEXACT REAL c;
1542 INEXACT REAL abig;
1543 REAL ahi, alo, bhi, blo;
1544 REAL err1, err2, err3;
1545 INEXACT REAL _i, _j;
1546 REAL _0;
1547
1548 acx = (REAL) (pa[0] - pc[0]);
1549 bcx = (REAL) (pb[0] - pc[0]);
1550 acy = (REAL) (pa[1] - pc[1]);
1551 bcy = (REAL) (pb[1] - pc[1]);
1552
1553 Two_Product(acx, bcy, detleft, detlefttail);
1554 Two_Product(acy, bcx, detright, detrighttail);
1555
1556 Two_Two_Diff(detleft, detlefttail, detright, detrighttail,
1557 B3, B[2], B[1], B[0]);
1558 B[3] = B3;
1559
1560 det = estimate(4, B);
1561 errbound = ccwerrboundB * detsum;
1562 if ((det >= errbound) || (-det >= errbound)) {
1563 return det;
1564 }
1565
1566 Two_Diff_Tail(pa[0], pc[0], acx, acxtail);
1567 Two_Diff_Tail(pb[0], pc[0], bcx, bcxtail);
1568 Two_Diff_Tail(pa[1], pc[1], acy, acytail);
1569 Two_Diff_Tail(pb[1], pc[1], bcy, bcytail);
1570
1571 if ((acxtail == 0.0) && (acytail == 0.0)
1572 && (bcxtail == 0.0) && (bcytail == 0.0)) {
1573 return det;
1574 }
1575
1576 errbound = ccwerrboundC * detsum + resulterrbound * Absolute(det);
1577 det += (acx * bcytail + bcy * acxtail)
1578 - (acy * bcxtail + bcx * acytail);
1579 if ((det >= errbound) || (-det >= errbound)) {
1580 return det;
1581 }
1582
1583 Two_Product(acxtail, bcy, s1, s0);
1584 Two_Product(acytail, bcx, t1, t0);
1585 Two_Two_Diff(s1, s0, t1, t0, u3, u[2], u[1], u[0]);
1586 u[3] = u3;
1587 C1length = fast_expansion_sum_zeroelim(4, B, 4, u, C1);
1588
1589 Two_Product(acx, bcytail, s1, s0);
1590 Two_Product(acy, bcxtail, t1, t0);
1591 Two_Two_Diff(s1, s0, t1, t0, u3, u[2], u[1], u[0]);
1592 u[3] = u3;
1593 C2length = fast_expansion_sum_zeroelim(C1length, C1, 4, u, C2);
1594
1595 Two_Product(acxtail, bcytail, s1, s0);
1596 Two_Product(acytail, bcxtail, t1, t0);
1597 Two_Two_Diff(s1, s0, t1, t0, u3, u[2], u[1], u[0]);
1598 u[3] = u3;
1599 Dlength = fast_expansion_sum_zeroelim(C2length, C2, 4, u, D);
1600
1601 return(D[Dlength - 1]);
1602 }
1603
1604 REAL orient2d(REAL *pa, REAL *pb, REAL *pc)
1605 {
1606 REAL detleft, detright, det;
1607 REAL detsum, errbound;
1608
1609 detleft = (pa[0] - pc[0]) * (pb[1] - pc[1]);
1610 detright = (pa[1] - pc[1]) * (pb[0] - pc[0]);
1611 det = detleft - detright;
1612
1613 if (detleft > 0.0) {
1614 if (detright <= 0.0) {
1615 return det;
1616 } else {
1617 detsum = detleft + detright;
1618 }
1619 } else if (detleft < 0.0) {
1620 if (detright >= 0.0) {
1621 return det;
1622 } else {
1623 detsum = -detleft - detright;
1624 }
1625 } else {
1626 return det;
1627 }
1628
1629 errbound = ccwerrboundA * detsum;
1630 if ((det >= errbound) || (-det >= errbound)) {
1631 return det;
1632 }
1633
1634 return orient2dadapt(pa, pb, pc, detsum);
1635 }
1636
1637 /*****************************************************************************/
1638 /* */
1639 /* orient3dfast() Approximate 3D orientation test. Nonrobust. */
1640 /* orient3dexact() Exact 3D orientation test. Robust. */
1641 /* orient3dslow() Another exact 3D orientation test. Robust. */
1642 /* orient3d() Adaptive exact 3D orientation test. Robust. */
1643 /* */
1644 /* Return a positive value if the point pd lies below the */
1645 /* plane passing through pa, pb, and pc; "below" is defined so */
1646 /* that pa, pb, and pc appear in counterclockwise order when */
1647 /* viewed from above the plane. Returns a negative value if */
1648 /* pd lies above the plane. Returns zero if the points are */
1649 /* coplanar. The result is also a rough approximation of six */
1650 /* times the signed volume of the tetrahedron defined by the */
1651 /* four points. */
1652 /* */
1653 /* Only the first and last routine should be used; the middle two are for */
1654 /* timings. */
1655 /* */
1656 /* The last three use exact arithmetic to ensure a correct answer. The */
1657 /* result returned is the determinant of a matrix. In orient3d() only, */
1658 /* this determinant is computed adaptively, in the sense that exact */
1659 /* arithmetic is used only to the degree it is needed to ensure that the */
1660 /* returned value has the correct sign. Hence, orient3d() is usually quite */
1661 /* fast, but will run more slowly when the input points are coplanar or */
1662 /* nearly so. */
1663 /* */
1664 /*****************************************************************************/
1665
1666 REAL orient3dfast(REAL *pa, REAL *pb, REAL *pc, REAL *pd)
1667 {
1668 REAL adx, bdx, cdx;
1669 REAL ady, bdy, cdy;
1670 REAL adz, bdz, cdz;
1671
1672 adx = pa[0] - pd[0];
1673 bdx = pb[0] - pd[0];
1674 cdx = pc[0] - pd[0];
1675 ady = pa[1] - pd[1];
1676 bdy = pb[1] - pd[1];
1677 cdy = pc[1] - pd[1];
1678 adz = pa[2] - pd[2];
1679 bdz = pb[2] - pd[2];
1680 cdz = pc[2] - pd[2];
1681
1682 return adx * (bdy * cdz - bdz * cdy)
1683 + bdx * (cdy * adz - cdz * ady)
1684 + cdx * (ady * bdz - adz * bdy);
1685 }
1686
1687 REAL orient3dexact(REAL *pa, REAL *pb, REAL *pc, REAL *pd)
1688 {
1689 INEXACT REAL axby1, bxcy1, cxdy1, dxay1, axcy1, bxdy1;
1690 INEXACT REAL bxay1, cxby1, dxcy1, axdy1, cxay1, dxby1;
1691 REAL axby0, bxcy0, cxdy0, dxay0, axcy0, bxdy0;
1692 REAL bxay0, cxby0, dxcy0, axdy0, cxay0, dxby0;
1693 REAL ab[4], bc[4], cd[4], da[4], ac[4], bd[4];
1694 REAL temp8[8];
1695 int templen;
1696 REAL abc[12], bcd[12], cda[12], dab[12];
1697 int abclen, bcdlen, cdalen, dablen;
1698 REAL adet[24], bdet[24], cdet[24], ddet[24];
1699 int alen, blen, clen, dlen;
1700 REAL abdet[48], cddet[48];
1701 int ablen, cdlen;
1702 REAL deter[96];
1703 int deterlen;
1704 int i;
1705
1706 INEXACT REAL bvirt;
1707 REAL avirt, bround, around;
1708 INEXACT REAL c;
1709 INEXACT REAL abig;
1710 REAL ahi, alo, bhi, blo;
1711 REAL err1, err2, err3;
1712 INEXACT REAL _i, _j;
1713 REAL _0;
1714
1715 Two_Product(pa[0], pb[1], axby1, axby0);
1716 Two_Product(pb[0], pa[1], bxay1, bxay0);
1717 Two_Two_Diff(axby1, axby0, bxay1, bxay0, ab[3], ab[2], ab[1], ab[0]);
1718
1719 Two_Product(pb[0], pc[1], bxcy1, bxcy0);
1720 Two_Product(pc[0], pb[1], cxby1, cxby0);
1721 Two_Two_Diff(bxcy1, bxcy0, cxby1, cxby0, bc[3], bc[2], bc[1], bc[0]);
1722
1723 Two_Product(pc[0], pd[1], cxdy1, cxdy0);
1724 Two_Product(pd[0], pc[1], dxcy1, dxcy0);
1725 Two_Two_Diff(cxdy1, cxdy0, dxcy1, dxcy0, cd[3], cd[2], cd[1], cd[0]);
1726
1727 Two_Product(pd[0], pa[1], dxay1, dxay0);
1728 Two_Product(pa[0], pd[1], axdy1, axdy0);
1729 Two_Two_Diff(dxay1, dxay0, axdy1, axdy0, da[3], da[2], da[1], da[0]);
1730
1731 Two_Product(pa[0], pc[1], axcy1, axcy0);
1732 Two_Product(pc[0], pa[1], cxay1, cxay0);
1733 Two_Two_Diff(axcy1, axcy0, cxay1, cxay0, ac[3], ac[2], ac[1], ac[0]);
1734
1735 Two_Product(pb[0], pd[1], bxdy1, bxdy0);
1736 Two_Product(pd[0], pb[1], dxby1, dxby0);
1737 Two_Two_Diff(bxdy1, bxdy0, dxby1, dxby0, bd[3], bd[2], bd[1], bd[0]);
1738
1739 templen = fast_expansion_sum_zeroelim(4, cd, 4, da, temp8);
1740 cdalen = fast_expansion_sum_zeroelim(templen, temp8, 4, ac, cda);
1741 templen = fast_expansion_sum_zeroelim(4, da, 4, ab, temp8);
1742 dablen = fast_expansion_sum_zeroelim(templen, temp8, 4, bd, dab);
1743 for (i = 0; i < 4; i++) {
1744 bd[i] = -bd[i];
1745 ac[i] = -ac[i];
1746 }
1747 templen = fast_expansion_sum_zeroelim(4, ab, 4, bc, temp8);
1748 abclen = fast_expansion_sum_zeroelim(templen, temp8, 4, ac, abc);
1749 templen = fast_expansion_sum_zeroelim(4, bc, 4, cd, temp8);
1750 bcdlen = fast_expansion_sum_zeroelim(templen, temp8, 4, bd, bcd);
1751
1752 alen = scale_expansion_zeroelim(bcdlen, bcd, pa[2], adet);
1753 blen = scale_expansion_zeroelim(cdalen, cda, -pb[2], bdet);
1754 clen = scale_expansion_zeroelim(dablen, dab, pc[2], cdet);
1755 dlen = scale_expansion_zeroelim(abclen, abc, -pd[2], ddet);
1756
1757 ablen = fast_expansion_sum_zeroelim(alen, adet, blen, bdet, abdet);
1758 cdlen = fast_expansion_sum_zeroelim(clen, cdet, dlen, ddet, cddet);
1759 deterlen = fast_expansion_sum_zeroelim(ablen, abdet, cdlen, cddet, deter);
1760
1761 return deter[deterlen - 1];
1762 }
1763
1764 REAL orient3dslow(REAL *pa, REAL *pb, REAL *pc, REAL *pd)
1765 {
1766 INEXACT REAL adx, ady, adz, bdx, bdy, bdz, cdx, cdy, cdz;
1767 REAL adxtail, adytail, adztail;
1768 REAL bdxtail, bdytail, bdztail;
1769 REAL cdxtail, cdytail, cdztail;
1770 REAL negate, negatetail;
1771 INEXACT REAL axby7, bxcy7, axcy7, bxay7, cxby7, cxay7;
1772 REAL axby[8], bxcy[8], axcy[8], bxay[8], cxby[8], cxay[8];
1773 REAL temp16[16], temp32[32], temp32t[32];
1774 int temp16len, temp32len, temp32tlen;
1775 REAL adet[64], bdet[64], cdet[64];
1776 int alen, blen, clen;
1777 REAL abdet[128];
1778 int ablen;
1779 REAL deter[192];
1780 int deterlen;
1781
1782 INEXACT REAL bvirt;
1783 REAL avirt, bround, around;
1784 INEXACT REAL c;
1785 INEXACT REAL abig;
1786 REAL a0hi, a0lo, a1hi, a1lo, bhi, blo;
1787 REAL err1, err2, err3;
1788 INEXACT REAL _i, _j, _k, _l, _m, _n;
1789 REAL _0, _1, _2;
1790
1791 Two_Diff(pa[0], pd[0], adx, adxtail);
1792 Two_Diff(pa[1], pd[1], ady, adytail);
1793 Two_Diff(pa[2], pd[2], adz, adztail);
1794 Two_Diff(pb[0], pd[0], bdx, bdxtail);
1795 Two_Diff(pb[1], pd[1], bdy, bdytail);
1796 Two_Diff(pb[2], pd[2], bdz, bdztail);
1797 Two_Diff(pc[0], pd[0], cdx, cdxtail);
1798 Two_Diff(pc[1], pd[1], cdy, cdytail);
1799 Two_Diff(pc[2], pd[2], cdz, cdztail);
1800
1801 Two_Two_Product(adx, adxtail, bdy, bdytail,
1802 axby7, axby[6], axby[5], axby[4],
1803 axby[3], axby[2], axby[1], axby[0]);
1804 axby[7] = axby7;
1805 negate = -ady;
1806 negatetail = -adytail;
1807 Two_Two_Product(bdx, bdxtail, negate, negatetail,
1808 bxay7, bxay[6], bxay[5], bxay[4],
1809 bxay[3], bxay[2], bxay[1], bxay[0]);
1810 bxay[7] = bxay7;
1811 Two_Two_Product(bdx, bdxtail, cdy, cdytail,
1812 bxcy7, bxcy[6], bxcy[5], bxcy[4],
1813 bxcy[3], bxcy[2], bxcy[1], bxcy[0]);
1814 bxcy[7] = bxcy7;
1815 negate = -bdy;
1816 negatetail = -bdytail;
1817 Two_Two_Product(cdx, cdxtail, negate, negatetail,
1818 cxby7, cxby[6], cxby[5], cxby[4],
1819 cxby[3], cxby[2], cxby[1], cxby[0]);
1820 cxby[7] = cxby7;
1821 Two_Two_Product(cdx, cdxtail, ady, adytail,
1822 cxay7, cxay[6], cxay[5], cxay[4],
1823 cxay[3], cxay[2], cxay[1], cxay[0]);
1824 cxay[7] = cxay7;
1825 negate = -cdy;
1826 negatetail = -cdytail;
1827 Two_Two_Product(adx, adxtail, negate, negatetail,
1828 axcy7, axcy[6], axcy[5], axcy[4],
1829 axcy[3], axcy[2], axcy[1], axcy[0]);
1830 axcy[7] = axcy7;
1831
1832 temp16len = fast_expansion_sum_zeroelim(8, bxcy, 8, cxby, temp16);
1833 temp32len = scale_expansion_zeroelim(temp16len, temp16, adz, temp32);
1834 temp32tlen = scale_expansion_zeroelim(temp16len, temp16, adztail, temp32t);
1835 alen = fast_expansion_sum_zeroelim(temp32len, temp32, temp32tlen, temp32t,
1836 adet);
1837
1838 temp16len = fast_expansion_sum_zeroelim(8, cxay, 8, axcy, temp16);
1839 temp32len = scale_expansion_zeroelim(temp16len, temp16, bdz, temp32);
1840 temp32tlen = scale_expansion_zeroelim(temp16len, temp16, bdztail, temp32t);
1841 blen = fast_expansion_sum_zeroelim(temp32len, temp32, temp32tlen, temp32t,
1842 bdet);
1843
1844 temp16len = fast_expansion_sum_zeroelim(8, axby, 8, bxay, temp16);
1845 temp32len = scale_expansion_zeroelim(temp16len, temp16, cdz, temp32);
1846 temp32tlen = scale_expansion_zeroelim(temp16len, temp16, cdztail, temp32t);
1847 clen = fast_expansion_sum_zeroelim(temp32len, temp32, temp32tlen, temp32t,
1848 cdet);
1849
1850 ablen = fast_expansion_sum_zeroelim(alen, adet, blen, bdet, abdet);
1851 deterlen = fast_expansion_sum_zeroelim(ablen, abdet, clen, cdet, deter);
1852
1853 return deter[deterlen - 1];
1854 }
1855
1856 REAL orient3dadapt(REAL *pa, REAL *pb, REAL *pc, REAL *pd, REAL permanent)
1857 {
1858 INEXACT REAL adx, bdx, cdx, ady, bdy, cdy, adz, bdz, cdz;
1859 REAL det, errbound;
1860
1861 INEXACT REAL bdxcdy1, cdxbdy1, cdxady1, adxcdy1, adxbdy1, bdxady1;
1862 REAL bdxcdy0, cdxbdy0, cdxady0, adxcdy0, adxbdy0, bdxady0;
1863 REAL bc[4], ca[4], ab[4];
1864 INEXACT REAL bc3, ca3, ab3;
1865 REAL adet[8], bdet[8], cdet[8];
1866 int alen, blen, clen;
1867 REAL abdet[16];
1868 int ablen;
1869 REAL *finnow, *finother, *finswap;
1870 REAL fin1[192], fin2[192];
1871 int finlength;
1872
1873 REAL adxtail, bdxtail, cdxtail;
1874 REAL adytail, bdytail, cdytail;
1875 REAL adztail, bdztail, cdztail;
1876 INEXACT REAL at_blarge, at_clarge;
1877 INEXACT REAL bt_clarge, bt_alarge;
1878 INEXACT REAL ct_alarge, ct_blarge;
1879 REAL at_b[4], at_c[4], bt_c[4], bt_a[4], ct_a[4], ct_b[4];
1880 int at_blen, at_clen, bt_clen, bt_alen, ct_alen, ct_blen;
1881 INEXACT REAL bdxt_cdy1, cdxt_bdy1, cdxt_ady1;
1882 INEXACT REAL adxt_cdy1, adxt_bdy1, bdxt_ady1;
1883 REAL bdxt_cdy0, cdxt_bdy0, cdxt_ady0;
1884 REAL adxt_cdy0, adxt_bdy0, bdxt_ady0;
1885 INEXACT REAL bdyt_cdx1, cdyt_bdx1, cdyt_adx1;
1886 INEXACT REAL adyt_cdx1, adyt_bdx1, bdyt_adx1;
1887 REAL bdyt_cdx0, cdyt_bdx0, cdyt_adx0;
1888 REAL adyt_cdx0, adyt_bdx0, bdyt_adx0;
1889 REAL bct[8], cat[8], abt[8];
1890 int bctlen, catlen, abtlen;
1891 INEXACT REAL bdxt_cdyt1, cdxt_bdyt1, cdxt_adyt1;
1892 INEXACT REAL adxt_cdyt1, adxt_bdyt1, bdxt_adyt1;
1893 REAL bdxt_cdyt0, cdxt_bdyt0, cdxt_adyt0;
1894 REAL adxt_cdyt0, adxt_bdyt0, bdxt_adyt0;
1895 REAL u[4], v[12], w[16];
1896 INEXACT REAL u3;
1897 int vlength, wlength;
1898 REAL negate;
1899
1900 INEXACT REAL bvirt;
1901 REAL avirt, bround, around;
1902 INEXACT REAL c;
1903 INEXACT REAL abig;
1904 REAL ahi, alo, bhi, blo;
1905 REAL err1, err2, err3;
1906 INEXACT REAL _i, _j, _k;
1907 REAL _0;
1908
1909 adx = (REAL) (pa[0] - pd[0]);
1910 bdx = (REAL) (pb[0] - pd[0]);
1911 cdx = (REAL) (pc[0] - pd[0]);
1912 ady = (REAL) (pa[1] - pd[1]);
1913 bdy = (REAL) (pb[1] - pd[1]);
1914 cdy = (REAL) (pc[1] - pd[1]);
1915 adz = (REAL) (pa[2] - pd[2]);
1916 bdz = (REAL) (pb[2] - pd[2]);
1917 cdz = (REAL) (pc[2] - pd[2]);
1918
1919 Two_Product(bdx, cdy, bdxcdy1, bdxcdy0);
1920 Two_Product(cdx, bdy, cdxbdy1, cdxbdy0);
1921 Two_Two_Diff(bdxcdy1, bdxcdy0, cdxbdy1, cdxbdy0, bc3, bc[2], bc[1], bc[0]);
1922 bc[3] = bc3;
1923 alen = scale_expansion_zeroelim(4, bc, adz, adet);
1924
1925 Two_Product(cdx, ady, cdxady1, cdxady0);
1926 Two_Product(adx, cdy, adxcdy1, adxcdy0);
1927 Two_Two_Diff(cdxady1, cdxady0, adxcdy1, adxcdy0, ca3, ca[2], ca[1], ca[0]);
1928 ca[3] = ca3;
1929 blen = scale_expansion_zeroelim(4, ca, bdz, bdet);
1930
1931 Two_Product(adx, bdy, adxbdy1, adxbdy0);
1932 Two_Product(bdx, ady, bdxady1, bdxady0);
1933 Two_Two_Diff(adxbdy1, adxbdy0, bdxady1, bdxady0, ab3, ab[2], ab[1], ab[0]);
1934 ab[3] = ab3;
1935 clen = scale_expansion_zeroelim(4, ab, cdz, cdet);
1936
1937 ablen = fast_expansion_sum_zeroelim(alen, adet, blen, bdet, abdet);
1938 finlength = fast_expansion_sum_zeroelim(ablen, abdet, clen, cdet, fin1);
1939
1940 det = estimate(finlength, fin1);
1941 errbound = o3derrboundB * permanent;
1942 if ((det >= errbound) || (-det >= errbound)) {
1943 return det;
1944 }
1945
1946 Two_Diff_Tail(pa[0], pd[0], adx, adxtail);
1947 Two_Diff_Tail(pb[0], pd[0], bdx, bdxtail);
1948 Two_Diff_Tail(pc[0], pd[0], cdx, cdxtail);
1949 Two_Diff_Tail(pa[1], pd[1], ady, adytail);
1950 Two_Diff_Tail(pb[1], pd[1], bdy, bdytail);
1951 Two_Diff_Tail(pc[1], pd[1], cdy, cdytail);
1952 Two_Diff_Tail(pa[2], pd[2], adz, adztail);
1953 Two_Diff_Tail(pb[2], pd[2], bdz, bdztail);
1954 Two_Diff_Tail(pc[2], pd[2], cdz, cdztail);
1955
1956 if ((adxtail == 0.0) && (bdxtail == 0.0) && (cdxtail == 0.0)
1957 && (adytail == 0.0) && (bdytail == 0.0) && (cdytail == 0.0)
1958 && (adztail == 0.0) && (bdztail == 0.0) && (cdztail == 0.0)) {
1959 return det;
1960 }
1961
1962 errbound = o3derrboundC * permanent + resulterrbound * Absolute(det);
1963 det += (adz * ((bdx * cdytail + cdy * bdxtail)
1964 - (bdy * cdxtail + cdx * bdytail))
1965 + adztail * (bdx * cdy - bdy * cdx))
1966 + (bdz * ((cdx * adytail + ady * cdxtail)
1967 - (cdy * adxtail + adx * cdytail))
1968 + bdztail * (cdx * ady - cdy * adx))
1969 + (cdz * ((adx * bdytail + bdy * adxtail)
1970 - (ady * bdxtail + bdx * adytail))
1971 + cdztail * (adx * bdy - ady * bdx));
1972 if ((det >= errbound) || (-det >= errbound)) {
1973 return det;
1974 }
1975
1976 finnow = fin1;
1977 finother = fin2;
1978
1979 if (adxtail == 0.0) {
1980 if (adytail == 0.0) {
1981 at_b[0] = 0.0;
1982 at_blen = 1;
1983 at_c[0] = 0.0;
1984 at_clen = 1;
1985 } else {
1986 negate = -adytail;
1987 Two_Product(negate, bdx, at_blarge, at_b[0]);
1988 at_b[1] = at_blarge;
1989 at_blen = 2;
1990 Two_Product(adytail, cdx, at_clarge, at_c[0]);
1991 at_c[1] = at_clarge;
1992 at_clen = 2;
1993 }
1994 } else {
1995 if (adytail == 0.0) {
1996 Two_Product(adxtail, bdy, at_blarge, at_b[0]);
1997 at_b[1] = at_blarge;
1998 at_blen = 2;
1999 negate = -adxtail;
2000 Two_Product(negate, cdy, at_clarge, at_c[0]);
2001 at_c[1] = at_clarge;
2002 at_clen = 2;
2003 } else {
2004 Two_Product(adxtail, bdy, adxt_bdy1, adxt_bdy0);
2005 Two_Product(adytail, bdx, adyt_bdx1, adyt_bdx0);
2006 Two_Two_Diff(adxt_bdy1, adxt_bdy0, adyt_bdx1, adyt_bdx0,
2007 at_blarge, at_b[2], at_b[1], at_b[0]);
2008 at_b[3] = at_blarge;
2009 at_blen = 4;
2010 Two_Product(adytail, cdx, adyt_cdx1, adyt_cdx0);
2011 Two_Product(adxtail, cdy, adxt_cdy1, adxt_cdy0);
2012 Two_Two_Diff(adyt_cdx1, adyt_cdx0, adxt_cdy1, adxt_cdy0,
2013 at_clarge, at_c[2], at_c[1], at_c[0]);
2014 at_c[3] = at_clarge;
2015 at_clen = 4;
2016 }
2017 }
2018 if (bdxtail == 0.0) {
2019 if (bdytail == 0.0) {
2020 bt_c[0] = 0.0;
2021 bt_clen = 1;
2022 bt_a[0] = 0.0;
2023 bt_alen = 1;
2024 } else {
2025 negate = -bdytail;
2026 Two_Product(negate, cdx, bt_clarge, bt_c[0]);
2027 bt_c[1] = bt_clarge;
2028 bt_clen = 2;
2029 Two_Product(bdytail, adx, bt_alarge, bt_a[0]);
2030 bt_a[1] = bt_alarge;
2031 bt_alen = 2;
2032 }
2033 } else {
2034 if (bdytail == 0.0) {
2035 Two_Product(bdxtail, cdy, bt_clarge, bt_c[0]);
2036 bt_c[1] = bt_clarge;
2037 bt_clen = 2;
2038 negate = -bdxtail;
2039 Two_Product(negate, ady, bt_alarge, bt_a[0]);
2040 bt_a[1] = bt_alarge;
2041 bt_alen = 2;
2042 } else {
2043 Two_Product(bdxtail, cdy, bdxt_cdy1, bdxt_cdy0);
2044 Two_Product(bdytail, cdx, bdyt_cdx1, bdyt_cdx0);
2045 Two_Two_Diff(bdxt_cdy1, bdxt_cdy0, bdyt_cdx1, bdyt_cdx0,
2046 bt_clarge, bt_c[2], bt_c[1], bt_c[0]);
2047 bt_c[3] = bt_clarge;
2048 bt_clen = 4;
2049 Two_Product(bdytail, adx, bdyt_adx1, bdyt_adx0);
2050 Two_Product(bdxtail, ady, bdxt_ady1, bdxt_ady0);
2051 Two_Two_Diff(bdyt_adx1, bdyt_adx0, bdxt_ady1, bdxt_ady0,
2052 bt_alarge, bt_a[2], bt_a[1], bt_a[0]);
2053 bt_a[3] = bt_alarge;
2054 bt_alen = 4;
2055 }
2056 }
2057 if (cdxtail == 0.0) {
2058 if (cdytail == 0.0) {
2059 ct_a[0] = 0.0;
2060 ct_alen = 1;
2061 ct_b[0] = 0.0;
2062 ct_blen = 1;
2063 } else {
2064 negate = -cdytail;
2065 Two_Product(negate, adx, ct_alarge, ct_a[0]);
2066 ct_a[1] = ct_alarge;
2067 ct_alen = 2;
2068 Two_Product(cdytail, bdx, ct_blarge, ct_b[0]);
2069 ct_b[1] = ct_blarge;
2070 ct_blen = 2;
2071 }
2072 } else {
2073 if (cdytail == 0.0) {
2074 Two_Product(cdxtail, ady, ct_alarge, ct_a[0]);
2075 ct_a[1] = ct_alarge;
2076 ct_alen = 2;
2077 negate = -cdxtail;
2078 Two_Product(negate, bdy, ct_blarge, ct_b[0]);
2079 ct_b[1] = ct_blarge;
2080 ct_blen = 2;
2081 } else {
2082 Two_Product(cdxtail, ady, cdxt_ady1, cdxt_ady0);
2083 Two_Product(cdytail, adx, cdyt_adx1, cdyt_adx0);
2084 Two_Two_Diff(cdxt_ady1, cdxt_ady0, cdyt_adx1, cdyt_adx0,
2085 ct_alarge, ct_a[2], ct_a[1], ct_a[0]);
2086 ct_a[3] = ct_alarge;
2087 ct_alen = 4;
2088 Two_Product(cdytail, bdx, cdyt_bdx1, cdyt_bdx0);
2089 Two_Product(cdxtail, bdy, cdxt_bdy1, cdxt_bdy0);
2090 Two_Two_Diff(cdyt_bdx1, cdyt_bdx0, cdxt_bdy1, cdxt_bdy0,
2091 ct_blarge, ct_b[2], ct_b[1], ct_b[0]);
2092 ct_b[3] = ct_blarge;
2093 ct_blen = 4;
2094 }
2095 }
2096
2097 bctlen = fast_expansion_sum_zeroelim(bt_clen, bt_c, ct_blen, ct_b, bct);
2098 wlength = scale_expansion_zeroelim(bctlen, bct, adz, w);
2099 finlength = fast_expansion_sum_zeroelim(finlength, finnow, wlength, w,
2100 finother);
2101 finswap = finnow;
2102 finnow = finother;
2103 finother = finswap;
2104
2105 catlen = fast_expansion_sum_zeroelim(ct_alen, ct_a, at_clen, at_c, cat);
2106 wlength = scale_expansion_zeroelim(catlen, cat, bdz, w);
2107 finlength = fast_expansion_sum_zeroelim(finlength, finnow, wlength, w,
2108 finother);
2109 finswap = finnow;
2110 finnow = finother;
2111 finother = finswap;
2112
2113 abtlen = fast_expansion_sum_zeroelim(at_blen, at_b, bt_alen, bt_a, abt);
2114 wlength = scale_expansion_zeroelim(abtlen, abt, cdz, w);
2115 finlength = fast_expansion_sum_zeroelim(finlength, finnow, wlength, w,
2116 finother);
2117 finswap = finnow;
2118 finnow = finother;
2119 finother = finswap;
2120
2121 if (adztail != 0.0) {
2122 vlength = scale_expansion_zeroelim(4, bc, adztail, v);
2123 finlength = fast_expansion_sum_zeroelim(finlength, finnow, vlength, v,
2124 finother);
2125 finswap = finnow;
2126 finnow = finother;
2127 finother = finswap;
2128 }
2129 if (bdztail != 0.0) {
2130 vlength = scale_expansion_zeroelim(4, ca, bdztail, v);
2131 finlength = fast_expansion_sum_zeroelim(finlength, finnow, vlength, v,
2132 finother);
2133 finswap = finnow;
2134 finnow = finother;
2135 finother = finswap;
2136 }
2137 if (cdztail != 0.0) {
2138 vlength = scale_expansion_zeroelim(4, ab, cdztail, v);
2139 finlength = fast_expansion_sum_zeroelim(finlength, finnow, vlength, v,
2140 finother);
2141 finswap = finnow;
2142 finnow = finother;
2143 finother = finswap;
2144 }
2145
2146 if (adxtail != 0.0) {
2147 if (bdytail != 0.0) {
2148 Two_Product(adxtail, bdytail, adxt_bdyt1, adxt_bdyt0);
2149 Two_One_Product(adxt_bdyt1, adxt_bdyt0, cdz, u3, u[2], u[1], u[0]);
2150 u[3] = u3;
2151 finlength = fast_expansion_sum_zeroelim(finlength, finnow, 4, u,
2152 finother);
2153 finswap = finnow;
2154 finnow = finother;
2155 finother = finswap;
2156 if (cdztail != 0.0) {
2157 Two_One_Product(adxt_bdyt1, adxt_bdyt0, cdztail, u3, u[2], u[1], u[0]);
2158 u[3] = u3;
2159 finlength = fast_expansion_sum_zeroelim(finlength, finnow, 4, u,
2160 finother);
2161 finswap = finnow;
2162 finnow = finother;
2163 finother = finswap;
2164 }
2165 }
2166 if (cdytail != 0.0) {
2167 negate = -adxtail;
2168 Two_Product(negate, cdytail, adxt_cdyt1, adxt_cdyt0);
2169 Two_One_Product(adxt_cdyt1, adxt_cdyt0, bdz, u3, u[2], u[1], u[0]);
2170 u[3] = u3;
2171 finlength = fast_expansion_sum_zeroelim(finlength, finnow, 4, u,
2172 finother);
2173 finswap = finnow;
2174 finnow = finother;
2175 finother = finswap;
2176 if (bdztail != 0.0) {
2177 Two_One_Product(adxt_cdyt1, adxt_cdyt0, bdztail, u3, u[2], u[1], u[0]);
2178 u[3] = u3;
2179 finlength = fast_expansion_sum_zeroelim(finlength, finnow, 4, u,
2180 finother);
2181 finswap = finnow;
2182 finnow = finother;
2183 finother = finswap;
2184 }
2185 }
2186 }
2187 if (bdxtail != 0.0) {
2188 if (cdytail != 0.0) {
2189 Two_Product(bdxtail, cdytail, bdxt_cdyt1, bdxt_cdyt0);
2190 Two_One_Product(bdxt_cdyt1, bdxt_cdyt0, adz, u3, u[2], u[1], u[0]);
2191 u[3] = u3;
2192 finlength = fast_expansion_sum_zeroelim(finlength, finnow, 4, u,
2193 finother);
2194 finswap = finnow;
2195 finnow = finother;
2196 finother = finswap;
2197 if (adztail != 0.0) {
2198 Two_One_Product(bdxt_cdyt1, bdxt_cdyt0, adztail, u3, u[2], u[1], u[0]);
2199 u[3] = u3;
2200 finlength = fast_expansion_sum_zeroelim(finlength, finnow, 4, u,
2201 finother);
2202 finswap = finnow;
2203 finnow = finother;
2204 finother = finswap;
2205 }
2206 }
2207 if (adytail != 0.0) {
2208 negate = -bdxtail;
2209 Two_Product(negate, adytail, bdxt_adyt1, bdxt_adyt0);
2210 Two_One_Product(bdxt_adyt1, bdxt_adyt0, cdz, u3, u[2], u[1], u[0]);
2211 u[3] = u3;
2212 finlength = fast_expansion_sum_zeroelim(finlength, finnow, 4, u,
2213 finother);
2214 finswap = finnow;
2215 finnow = finother;
2216 finother = finswap;
2217 if (cdztail != 0.0) {
2218 Two_One_Product(bdxt_adyt1, bdxt_adyt0, cdztail, u3, u[2], u[1], u[0]);
2219 u[3] = u3;
2220 finlength = fast_expansion_sum_zeroelim(finlength, finnow, 4, u,
2221 finother);
2222 finswap = finnow;
2223 finnow = finother;
2224 finother = finswap;
2225 }
2226 }
2227 }
2228 if (cdxtail != 0.0) {
2229 if (adytail != 0.0) {
2230 Two_Product(cdxtail, adytail, cdxt_adyt1, cdxt_adyt0);
2231 Two_One_Product(cdxt_adyt1, cdxt_adyt0, bdz, u3, u[2], u[1], u[0]);
2232 u[3] = u3;
2233 finlength = fast_expansion_sum_zeroelim(finlength, finnow, 4, u,
2234 finother);
2235 finswap = finnow;
2236 finnow = finother;
2237 finother = finswap;
2238 if (bdztail != 0.0) {
2239 Two_One_Product(cdxt_adyt1, cdxt_adyt0, bdztail, u3, u[2], u[1], u[0]);
2240 u[3] = u3;
2241 finlength = fast_expansion_sum_zeroelim(finlength, finnow, 4, u,
2242 finother);
2243 finswap = finnow;
2244 finnow = finother;
2245 finother = finswap;
2246 }
2247 }
2248 if (bdytail != 0.0) {
2249 negate = -cdxtail;
2250 Two_Product(negate, bdytail, cdxt_bdyt1, cdxt_bdyt0);
2251 Two_One_Product(cdxt_bdyt1, cdxt_bdyt0, adz, u3, u[2], u[1], u[0]);
2252 u[3] = u3;
2253 finlength = fast_expansion_sum_zeroelim(finlength, finnow, 4, u,
2254 finother);
2255 finswap = finnow;
2256 finnow = finother;
2257 finother = finswap;
2258 if (adztail != 0.0) {
2259 Two_One_Product(cdxt_bdyt1, cdxt_bdyt0, adztail, u3, u[2], u[1], u[0]);
2260 u[3] = u3;
2261 finlength = fast_expansion_sum_zeroelim(finlength, finnow, 4, u,
2262 finother);
2263 finswap = finnow;
2264 finnow = finother;
2265 finother = finswap;
2266 }
2267 }
2268 }
2269
2270 if (adztail != 0.0) {
2271 wlength = scale_expansion_zeroelim(bctlen, bct, adztail, w);
2272 finlength = fast_expansion_sum_zeroelim(finlength, finnow, wlength, w,
2273 finother);
2274 finswap = finnow;
2275 finnow = finother;
2276 finother = finswap;
2277 }
2278 if (bdztail != 0.0) {
2279 wlength = scale_expansion_zeroelim(catlen, cat, bdztail, w);
2280 finlength = fast_expansion_sum_zeroelim(finlength, finnow, wlength, w,
2281 finother);
2282 finswap = finnow;
2283 finnow = finother;
2284 finother = finswap;
2285 }
2286 if (cdztail != 0.0) {
2287 wlength = scale_expansion_zeroelim(abtlen, abt, cdztail, w);
2288 finlength = fast_expansion_sum_zeroelim(finlength, finnow, wlength, w,
2289 finother);
2290 finswap = finnow;
2291 finnow = finother;
2292 finother = finswap;
2293 }
2294
2295 return finnow[finlength - 1];
2296 }
2297
2298 REAL orient3d(REAL *pa, REAL *pb, REAL *pc, REAL *pd)
2299 {
2300 REAL adx, bdx, cdx, ady, bdy, cdy, adz, bdz, cdz;
2301 REAL bdxcdy, cdxbdy, cdxady, adxcdy, adxbdy, bdxady;
2302 REAL det;
2303 REAL permanent, errbound;
2304
2305 adx = pa[0] - pd[0];
2306 bdx = pb[0] - pd[0];
2307 cdx = pc[0] - pd[0];
2308 ady = pa[1] - pd[1];
2309 bdy = pb[1] - pd[1];
2310 cdy = pc[1] - pd[1];
2311 adz = pa[2] - pd[2];
2312 bdz = pb[2] - pd[2];
2313 cdz = pc[2] - pd[2];
2314
2315 bdxcdy = bdx * cdy;
2316 cdxbdy = cdx * bdy;
2317
2318 cdxady = cdx * ady;
2319 adxcdy = adx * cdy;
2320
2321 adxbdy = adx * bdy;
2322 bdxady = bdx * ady;
2323
2324 det = adz * (bdxcdy - cdxbdy)
2325 + bdz * (cdxady - adxcdy)
2326 + cdz * (adxbdy - bdxady);
2327
2328 permanent = (Absolute(bdxcdy) + Absolute(cdxbdy)) * Absolute(adz)
2329 + (Absolute(cdxady) + Absolute(adxcdy)) * Absolute(bdz)
2330 + (Absolute(adxbdy) + Absolute(bdxady)) * Absolute(cdz);
2331 errbound = o3derrboundA * permanent;
2332 if ((det > errbound) || (-det > errbound)) {
2333 return det;
2334 }
2335
2336 return orient3dadapt(pa, pb, pc, pd, permanent);
2337 }
2338
2339 /*****************************************************************************/
2340 /* */
2341 /* incirclefast() Approximate 2D incircle test. Nonrobust. */
2342 /* incircleexact() Exact 2D incircle test. Robust. */
2343 /* incircleslow() Another exact 2D incircle test. Robust. */
2344 /* incircle() Adaptive exact 2D incircle test. Robust. */
2345 /* */
2346 /* Return a positive value if the point pd lies inside the */
2347 /* circle passing through pa, pb, and pc; a negative value if */
2348 /* it lies outside; and zero if the four points are cocircular.*/
2349 /* The points pa, pb, and pc must be in counterclockwise */
2350 /* order, or the sign of the result will be reversed. */
2351 /* */
2352 /* Only the first and last routine should be used; the middle two are for */
2353 /* timings. */
2354 /* */
2355 /* The last three use exact arithmetic to ensure a correct answer. The */
2356 /* result returned is the determinant of a matrix. In incircle() only, */
2357 /* this determinant is computed adaptively, in the sense that exact */
2358 /* arithmetic is used only to the degree it is needed to ensure that the */
2359 /* returned value has the correct sign. Hence, incircle() is usually quite */
2360 /* fast, but will run more slowly when the input points are cocircular or */
2361 /* nearly so. */
2362 /* */
2363 /*****************************************************************************/
2364
2365 REAL incirclefast(REAL *pa, REAL *pb, REAL *pc, REAL *pd)
2366 {
2367 REAL adx, ady, bdx, bdy, cdx, cdy;
2368 REAL abdet, bcdet, cadet;
2369 REAL alift, blift, clift;
2370
2371 adx = pa[0] - pd[0];
2372 ady = pa[1] - pd[1];
2373 bdx = pb[0] - pd[0];
2374 bdy = pb[1] - pd[1];
2375 cdx = pc[0] - pd[0];
2376 cdy = pc[1] - pd[1];
2377
2378 abdet = adx * bdy - bdx * ady;
2379 bcdet = bdx * cdy - cdx * bdy;
2380 cadet = cdx * ady - adx * cdy;
2381 alift = adx * adx + ady * ady;
2382 blift = bdx * bdx + bdy * bdy;
2383 clift = cdx * cdx + cdy * cdy;
2384
2385 return alift * bcdet + blift * cadet + clift * abdet;
2386 }
2387
2388 REAL incircleexact(REAL *pa, REAL *pb, REAL *pc, REAL *pd)
2389 {
2390 INEXACT REAL axby1, bxcy1, cxdy1, dxay1, axcy1, bxdy1;
2391 INEXACT REAL bxay1, cxby1, dxcy1, axdy1, cxay1, dxby1;
2392 REAL axby0, bxcy0, cxdy0, dxay0, axcy0, bxdy0;
2393 REAL bxay0, cxby0, dxcy0, axdy0, cxay0, dxby0;
2394 REAL ab[4], bc[4], cd[4], da[4], ac[4], bd[4];
2395 REAL temp8[8];
2396 int templen;
2397 REAL abc[12], bcd[12], cda[12], dab[12];
2398 int abclen, bcdlen, cdalen, dablen;
2399 REAL det24x[24], det24y[24], det48x[48], det48y[48];
2400 int xlen, ylen;
2401 REAL adet[96], bdet[96], cdet[96], ddet[96];
2402 int alen, blen, clen, dlen;
2403 REAL abdet[192], cddet[192];
2404 int ablen, cdlen;
2405 REAL deter[384];
2406 int deterlen;
2407 int i;
2408
2409 INEXACT REAL bvirt;
2410 REAL avirt, bround, around;
2411 INEXACT REAL c;
2412 INEXACT REAL abig;
2413 REAL ahi, alo, bhi, blo;
2414 REAL err1, err2, err3;
2415 INEXACT REAL _i, _j;
2416 REAL _0;
2417
2418 Two_Product(pa[0], pb[1], axby1, axby0);
2419 Two_Product(pb[0], pa[1], bxay1, bxay0);
2420 Two_Two_Diff(axby1, axby0, bxay1, bxay0, ab[3], ab[2], ab[1], ab[0]);
2421
2422 Two_Product(pb[0], pc[1], bxcy1, bxcy0);
2423 Two_Product(pc[0], pb[1], cxby1, cxby0);
2424 Two_Two_Diff(bxcy1, bxcy0, cxby1, cxby0, bc[3], bc[2], bc[1], bc[0]);
2425
2426 Two_Product(pc[0], pd[1], cxdy1, cxdy0);
2427 Two_Product(pd[0], pc[1], dxcy1, dxcy0);
2428 Two_Two_Diff(cxdy1, cxdy0, dxcy1, dxcy0, cd[3], cd[2], cd[1], cd[0]);
2429
2430 Two_Product(pd[0], pa[1], dxay1, dxay0);
2431 Two_Product(pa[0], pd[1], axdy1, axdy0);
2432 Two_Two_Diff(dxay1, dxay0, axdy1, axdy0, da[3], da[2], da[1], da[0]);
2433
2434 Two_Product(pa[0], pc[1], axcy1, axcy0);
2435 Two_Product(pc[0], pa[1], cxay1, cxay0);
2436 Two_Two_Diff(axcy1, axcy0, cxay1, cxay0, ac[3], ac[2], ac[1], ac[0]);
2437
2438 Two_Product(pb[0], pd[1], bxdy1, bxdy0);
2439 Two_Product(pd[0], pb[1], dxby1, dxby0);
2440 Two_Two_Diff(bxdy1, bxdy0, dxby1, dxby0, bd[3], bd[2], bd[1], bd[0]);
2441
2442 templen = fast_expansion_sum_zeroelim(4, cd, 4, da, temp8);
2443 cdalen = fast_expansion_sum_zeroelim(templen, temp8, 4, ac, cda);
2444 templen = fast_expansion_sum_zeroelim(4, da, 4, ab, temp8);
2445 dablen = fast_expansion_sum_zeroelim(templen, temp8, 4, bd, dab);
2446 for (i = 0; i < 4; i++) {
2447 bd[i] = -bd[i];
2448 ac[i] = -ac[i];
2449 }
2450 templen = fast_expansion_sum_zeroelim(4, ab, 4, bc, temp8);
2451 abclen = fast_expansion_sum_zeroelim(templen, temp8, 4, ac, abc);
2452 templen = fast_expansion_sum_zeroelim(4, bc, 4, cd, temp8);
2453 bcdlen = fast_expansion_sum_zeroelim(templen, temp8, 4, bd, bcd);
2454
2455 xlen = scale_expansion_zeroelim(bcdlen, bcd, pa[0], det24x);
2456 xlen = scale_expansion_zeroelim(xlen, det24x, pa[0], det48x);
2457 ylen = scale_expansion_zeroelim(bcdlen, bcd, pa[1], det24y);
2458 ylen = scale_expansion_zeroelim(ylen, det24y, pa[1], det48y);
2459 alen = fast_expansion_sum_zeroelim(xlen, det48x, ylen, det48y, adet);
2460
2461 xlen = scale_expansion_zeroelim(cdalen, cda, pb[0], det24x);
2462 xlen = scale_expansion_zeroelim(xlen, det24x, -pb[0], det48x);
2463 ylen = scale_expansion_zeroelim(cdalen, cda, pb[1], det24y);
2464 ylen = scale_expansion_zeroelim(ylen, det24y, -pb[1], det48y);
2465 blen = fast_expansion_sum_zeroelim(xlen, det48x, ylen, det48y, bdet);
2466
2467 xlen = scale_expansion_zeroelim(dablen, dab, pc[0], det24x);
2468 xlen = scale_expansion_zeroelim(xlen, det24x, pc[0], det48x);
2469 ylen = scale_expansion_zeroelim(dablen, dab, pc[1], det24y);
2470 ylen = scale_expansion_zeroelim(ylen, det24y, pc[1], det48y);
2471 clen = fast_expansion_sum_zeroelim(xlen, det48x, ylen, det48y, cdet);
2472
2473 xlen = scale_expansion_zeroelim(abclen, abc, pd[0], det24x);
2474 xlen = scale_expansion_zeroelim(xlen, det24x, -pd[0], det48x);
2475 ylen = scale_expansion_zeroelim(abclen, abc, pd[1], det24y);
2476 ylen = scale_expansion_zeroelim(ylen, det24y, -pd[1], det48y);
2477 dlen = fast_expansion_sum_zeroelim(xlen, det48x, ylen, det48y, ddet);
2478
2479 ablen = fast_expansion_sum_zeroelim(alen, adet, blen, bdet, abdet);
2480 cdlen = fast_expansion_sum_zeroelim(clen, cdet, dlen, ddet, cddet);
2481 deterlen = fast_expansion_sum_zeroelim(ablen, abdet, cdlen, cddet, deter);
2482
2483 return deter[deterlen - 1];
2484 }
2485
2486 REAL incircleslow(REAL *pa, REAL *pb, REAL *pc, REAL *pd)
2487 {
2488 INEXACT REAL adx, bdx, cdx, ady, bdy, cdy;
2489 REAL adxtail, bdxtail, cdxtail;
2490 REAL adytail, bdytail, cdytail;
2491 REAL negate, negatetail;
2492 INEXACT REAL axby7, bxcy7, axcy7, bxay7, cxby7, cxay7;
2493 REAL axby[8], bxcy[8], axcy[8], bxay[8], cxby[8], cxay[8];
2494 REAL temp16[16];
2495 int temp16len;
2496 REAL detx[32], detxx[64], detxt[32], detxxt[64], detxtxt[64];
2497 int xlen, xxlen, xtlen, xxtlen, xtxtlen;
2498 REAL x1[128], x2[192];
2499 int x1len, x2len;
2500 REAL dety[32], detyy[64], detyt[32], detyyt[64], detytyt[64];
2501 int ylen, yylen, ytlen, yytlen, ytytlen;
2502 REAL y1[128], y2[192];
2503 int y1len, y2len;
2504 REAL adet[384], bdet[384], cdet[384], abdet[768], deter[1152];
2505 int alen, blen, clen, ablen, deterlen;
2506 int i;
2507
2508 INEXACT REAL bvirt;
2509 REAL avirt, bround, around;
2510 INEXACT REAL c;
2511 INEXACT REAL abig;
2512 REAL a0hi, a0lo, a1hi, a1lo, bhi, blo;
2513 REAL err1, err2, err3;
2514 INEXACT REAL _i, _j, _k, _l, _m, _n;
2515 REAL _0, _1, _2;
2516
2517 Two_Diff(pa[0], pd[0], adx, adxtail);
2518 Two_Diff(pa[1], pd[1], ady, adytail);
2519 Two_Diff(pb[0], pd[0], bdx, bdxtail);
2520 Two_Diff(pb[1], pd[1], bdy, bdytail);
2521 Two_Diff(pc[0], pd[0], cdx, cdxtail);
2522 Two_Diff(pc[1], pd[1], cdy, cdytail);
2523
2524 Two_Two_Product(adx, adxtail, bdy, bdytail,
2525 axby7, axby[6], axby[5], axby[4],
2526 axby[3], axby[2], axby[1], axby[0]);
2527 axby[7] = axby7;
2528 negate = -ady;
2529 negatetail = -adytail;
2530 Two_Two_Product(bdx, bdxtail, negate, negatetail,
2531 bxay7, bxay[6], bxay[5], bxay[4],
2532 bxay[3], bxay[2], bxay[1], bxay[0]);
2533 bxay[7] = bxay7;
2534 Two_Two_Product(bdx, bdxtail, cdy, cdytail,
2535 bxcy7, bxcy[6], bxcy[5], bxcy[4],
2536 bxcy[3], bxcy[2], bxcy[1], bxcy[0]);
2537 bxcy[7] = bxcy7;
2538 negate = -bdy;
2539 negatetail = -bdytail;
2540 Two_Two_Product(cdx, cdxtail, negate, negatetail,
2541 cxby7, cxby[6], cxby[5], cxby[4],
2542 cxby[3], cxby[2], cxby[1], cxby[0]);
2543 cxby[7] = cxby7;
2544 Two_Two_Product(cdx, cdxtail, ady, adytail,
2545 cxay7, cxay[6], cxay[5], cxay[4],
2546 cxay[3], cxay[2], cxay[1], cxay[0]);
2547 cxay[7] = cxay7;
2548 negate = -cdy;
2549 negatetail = -cdytail;
2550 Two_Two_Product(adx, adxtail, negate, negatetail,
2551 axcy7, axcy[6], axcy[5], axcy[4],
2552 axcy[3], axcy[2], axcy[1], axcy[0]);
2553 axcy[7] = axcy7;
2554
2555
2556 temp16len = fast_expansion_sum_zeroelim(8, bxcy, 8, cxby, temp16);
2557
2558 xlen = scale_expansion_zeroelim(temp16len, temp16, adx, detx);
2559 xxlen = scale_expansion_zeroelim(xlen, detx, adx, detxx);
2560 xtlen = scale_expansion_zeroelim(temp16len, temp16, adxtail, detxt);
2561 xxtlen = scale_expansion_zeroelim(xtlen, detxt, adx, detxxt);
2562 for (i = 0; i < xxtlen; i++) {
2563 detxxt[i] *= 2.0;
2564 }
2565 xtxtlen = scale_expansion_zeroelim(xtlen, detxt, adxtail, detxtxt);
2566 x1len = fast_expansion_sum_zeroelim(xxlen, detxx, xxtlen, detxxt, x1);
2567 x2len = fast_expansion_sum_zeroelim(x1len, x1, xtxtlen, detxtxt, x2);
2568
2569 ylen = scale_expansion_zeroelim(temp16len, temp16, ady, dety);
2570 yylen = scale_expansion_zeroelim(ylen, dety, ady, detyy);
2571 ytlen = scale_expansion_zeroelim(temp16len, temp16, adytail, detyt);
2572 yytlen = scale_expansion_zeroelim(ytlen, detyt, ady, detyyt);
2573 for (i = 0; i < yytlen; i++) {
2574 detyyt[i] *= 2.0;
2575 }
2576 ytytlen = scale_expansion_zeroelim(ytlen, detyt, adytail, detytyt);
2577 y1len = fast_expansion_sum_zeroelim(yylen, detyy, yytlen, detyyt, y1);
2578 y2len = fast_expansion_sum_zeroelim(y1len, y1, ytytlen, detytyt, y2);
2579
2580 alen = fast_expansion_sum_zeroelim(x2len, x2, y2len, y2, adet);
2581
2582
2583 temp16len = fast_expansion_sum_zeroelim(8, cxay, 8, axcy, temp16);
2584
2585 xlen = scale_expansion_zeroelim(temp16len, temp16, bdx, detx);
2586 xxlen = scale_expansion_zeroelim(xlen, detx, bdx, detxx);
2587 xtlen = scale_expansion_zeroelim(temp16len, temp16, bdxtail, detxt);
2588 xxtlen = scale_expansion_zeroelim(xtlen, detxt, bdx, detxxt);
2589 for (i = 0; i < xxtlen; i++) {
2590 detxxt[i] *= 2.0;
2591 }
2592 xtxtlen = scale_expansion_zeroelim(xtlen, detxt, bdxtail, detxtxt);
2593 x1len = fast_expansion_sum_zeroelim(xxlen, detxx, xxtlen, detxxt, x1);
2594 x2len = fast_expansion_sum_zeroelim(x1len, x1, xtxtlen, detxtxt, x2);
2595
2596 ylen = scale_expansion_zeroelim(temp16len, temp16, bdy, dety);
2597 yylen = scale_expansion_zeroelim(ylen, dety, bdy, detyy);
2598 ytlen = scale_expansion_zeroelim(temp16len, temp16, bdytail, detyt);
2599 yytlen = scale_expansion_zeroelim(ytlen, detyt, bdy, detyyt);
2600 for (i = 0; i < yytlen; i++) {
2601 detyyt[i] *= 2.0;
2602 }
2603 ytytlen = scale_expansion_zeroelim(ytlen, detyt, bdytail, detytyt);
2604 y1len = fast_expansion_sum_zeroelim(yylen, detyy, yytlen, detyyt, y1);
2605 y2len = fast_expansion_sum_zeroelim(y1len, y1, ytytlen, detytyt, y2);
2606
2607 blen = fast_expansion_sum_zeroelim(x2len, x2, y2len, y2, bdet);
2608
2609
2610 temp16len = fast_expansion_sum_zeroelim(8, axby, 8, bxay, temp16);
2611
2612 xlen = scale_expansion_zeroelim(temp16len, temp16, cdx, detx);
2613 xxlen = scale_expansion_zeroelim(xlen, detx, cdx, detxx);
2614 xtlen = scale_expansion_zeroelim(temp16len, temp16, cdxtail, detxt);
2615 xxtlen = scale_expansion_zeroelim(xtlen, detxt, cdx, detxxt);
2616 for (i = 0; i < xxtlen; i++) {
2617 detxxt[i] *= 2.0;
2618 }
2619 xtxtlen = scale_expansion_zeroelim(xtlen, detxt, cdxtail, detxtxt);
2620 x1len = fast_expansion_sum_zeroelim(xxlen, detxx, xxtlen, detxxt, x1);
2621 x2len = fast_expansion_sum_zeroelim(x1len, x1, xtxtlen, detxtxt, x2);
2622
2623 ylen = scale_expansion_zeroelim(temp16len, temp16, cdy, dety);
2624 yylen = scale_expansion_zeroelim(ylen, dety, cdy, detyy);
2625 ytlen = scale_expansion_zeroelim(temp16len, temp16, cdytail, detyt);
2626 yytlen = scale_expansion_zeroelim(ytlen, detyt, cdy, detyyt);
2627 for (i = 0; i < yytlen; i++) {
2628 detyyt[i] *= 2.0;
2629 }
2630 ytytlen = scale_expansion_zeroelim(ytlen, detyt, cdytail, detytyt);
2631 y1len = fast_expansion_sum_zeroelim(yylen, detyy, yytlen, detyyt, y1);
2632 y2len = fast_expansion_sum_zeroelim(y1len, y1, ytytlen, detytyt, y2);
2633
2634 clen = fast_expansion_sum_zeroelim(x2len, x2, y2len, y2, cdet);
2635
2636 ablen = fast_expansion_sum_zeroelim(alen, adet, blen, bdet, abdet);
2637 deterlen = fast_expansion_sum_zeroelim(ablen, abdet, clen, cdet, deter);
2638
2639 return deter[deterlen - 1];
2640 }
2641
2642 REAL incircleadapt(REAL *pa, REAL *pb, REAL *pc, REAL *pd, REAL permanent)
2643 {
2644 INEXACT REAL adx, bdx, cdx, ady, bdy, cdy;
2645 REAL det, errbound;
2646
2647 INEXACT REAL bdxcdy1, cdxbdy1, cdxady1, adxcdy1, adxbdy1, bdxady1;
2648 REAL bdxcdy0, cdxbdy0, cdxady0, adxcdy0, adxbdy0, bdxady0;
2649 REAL bc[4], ca[4], ab[4];
2650 INEXACT REAL bc3, ca3, ab3;
2651 REAL axbc[8], axxbc[16], aybc[8], ayybc[16], adet[32];
2652 int axbclen, axxbclen, aybclen, ayybclen, alen;
2653 REAL bxca[8], bxxca[16], byca[8], byyca[16], bdet[32];
2654 int bxcalen, bxxcalen, bycalen, byycalen, blen;
2655 REAL cxab[8], cxxab[16], cyab[8], cyyab[16], cdet[32];
2656 int cxablen, cxxablen, cyablen, cyyablen, clen;
2657 REAL abdet[64];
2658 int ablen;
2659 REAL fin1[1152], fin2[1152];
2660 REAL *finnow, *finother, *finswap;
2661 int finlength;
2662
2663 REAL adxtail, bdxtail, cdxtail, adytail, bdytail, cdytail;
2664 INEXACT REAL adxadx1, adyady1, bdxbdx1, bdybdy1, cdxcdx1, cdycdy1;
2665 REAL adxadx0, adyady0, bdxbdx0, bdybdy0, cdxcdx0, cdycdy0;
2666 REAL aa[4], bb[4], cc[4];
2667 INEXACT REAL aa3, bb3, cc3;
2668 INEXACT REAL ti1, tj1;
2669 REAL ti0, tj0;
2670 REAL u[4], v[4];
2671 INEXACT REAL u3, v3;
2672 REAL temp8[8], temp16a[16], temp16b[16], temp16c[16];
2673 REAL temp32a[32], temp32b[32], temp48[48], temp64[64];
2674 int temp8len, temp16alen, temp16blen, temp16clen;
2675 int temp32alen, temp32blen, temp48len, temp64len;
2676 REAL axtbb[8], axtcc[8], aytbb[8], aytcc[8];
2677 int axtbblen, axtcclen, aytbblen, aytcclen;
2678 REAL bxtaa[8], bxtcc[8], bytaa[8], bytcc[8];
2679 int bxtaalen, bxtcclen, bytaalen, bytcclen;
2680 REAL cxtaa[8], cxtbb[8], cytaa[8], cytbb[8];
2681 int cxtaalen, cxtbblen, cytaalen, cytbblen;
2682 REAL axtbc[8], aytbc[8], bxtca[8], bytca[8], cxtab[8], cytab[8];
2683 int axtbclen=0, aytbclen=0, bxtcalen=0, bytcalen=0, cxtablen=0, cytablen=0;
2684 REAL axtbct[16], aytbct[16], bxtcat[16], bytcat[16], cxtabt[16], cytabt[16];
2685 int axtbctlen, aytbctlen, bxtcatlen, bytcatlen, cxtabtlen, cytabtlen;
2686 REAL axtbctt[8], aytbctt[8], bxtcatt[8];
2687 REAL bytcatt[8], cxtabtt[8], cytabtt[8];
2688 int axtbcttlen, aytbcttlen, bxtcattlen, bytcattlen, cxtabttlen, cytabttlen;
2689 REAL abt[8], bct[8], cat[8];
2690 int abtlen, bctlen, catlen;
2691 REAL abtt[4], bctt[4], catt[4];
2692 int abttlen, bcttlen, cattlen;
2693 INEXACT REAL abtt3, bctt3, catt3;
2694 REAL negate;
2695
2696 INEXACT REAL bvirt;
2697 REAL avirt, bround, around;
2698 INEXACT REAL c;
2699 INEXACT REAL abig;
2700 REAL ahi, alo, bhi, blo;
2701 REAL err1, err2, err3;
2702 INEXACT REAL _i, _j;
2703 REAL _0;
2704
2705 adx = (REAL) (pa[0] - pd[0]);
2706 bdx = (REAL) (pb[0] - pd[0]);
2707 cdx = (REAL) (pc[0] - pd[0]);
2708 ady = (REAL) (pa[1] - pd[1]);
2709 bdy = (REAL) (pb[1] - pd[1]);
2710 cdy = (REAL) (pc[1] - pd[1]);
2711
2712 Two_Product(bdx, cdy, bdxcdy1, bdxcdy0);
2713 Two_Product(cdx, bdy, cdxbdy1, cdxbdy0);
2714 Two_Two_Diff(bdxcdy1, bdxcdy0, cdxbdy1, cdxbdy0, bc3, bc[2], bc[1], bc[0]);
2715 bc[3] = bc3;
2716 axbclen = scale_expansion_zeroelim(4, bc, adx, axbc);
2717 axxbclen = scale_expansion_zeroelim(axbclen, axbc, adx, axxbc);
2718 aybclen = scale_expansion_zeroelim(4, bc, ady, aybc);
2719 ayybclen = scale_expansion_zeroelim(aybclen, aybc, ady, ayybc);
2720 alen = fast_expansion_sum_zeroelim(axxbclen, axxbc, ayybclen, ayybc, adet);
2721
2722 Two_Product(cdx, ady, cdxady1, cdxady0);
2723 Two_Product(adx, cdy, adxcdy1, adxcdy0);
2724 Two_Two_Diff(cdxady1, cdxady0, adxcdy1, adxcdy0, ca3, ca[2], ca[1], ca[0]);
2725 ca[3] = ca3;
2726 bxcalen = scale_expansion_zeroelim(4, ca, bdx, bxca);
2727 bxxcalen = scale_expansion_zeroelim(bxcalen, bxca, bdx, bxxca);
2728 bycalen = scale_expansion_zeroelim(4, ca, bdy, byca);
2729 byycalen = scale_expansion_zeroelim(bycalen, byca, bdy, byyca);
2730 blen = fast_expansion_sum_zeroelim(bxxcalen, bxxca, byycalen, byyca, bdet);
2731
2732 Two_Product(adx, bdy, adxbdy1, adxbdy0);
2733 Two_Product(bdx, ady, bdxady1, bdxady0);
2734 Two_Two_Diff(adxbdy1, adxbdy0, bdxady1, bdxady0, ab3, ab[2], ab[1], ab[0]);
2735 ab[3] = ab3;
2736 cxablen = scale_expansion_zeroelim(4, ab, cdx, cxab);
2737 cxxablen = scale_expansion_zeroelim(cxablen, cxab, cdx, cxxab);
2738 cyablen = scale_expansion_zeroelim(4, ab, cdy, cyab);
2739 cyyablen = scale_expansion_zeroelim(cyablen, cyab, cdy, cyyab);
2740 clen = fast_expansion_sum_zeroelim(cxxablen, cxxab, cyyablen, cyyab, cdet);
2741
2742 ablen = fast_expansion_sum_zeroelim(alen, adet, blen, bdet, abdet);
2743 finlength = fast_expansion_sum_zeroelim(ablen, abdet, clen, cdet, fin1);
2744
2745 det = estimate(finlength, fin1);
2746 errbound = iccerrboundB * permanent;
2747 if ((det >= errbound) || (-det >= errbound)) {
2748 return det;
2749 }
2750
2751 Two_Diff_Tail(pa[0], pd[0], adx, adxtail);
2752 Two_Diff_Tail(pa[1], pd[1], ady, adytail);
2753 Two_Diff_Tail(pb[0], pd[0], bdx, bdxtail);
2754 Two_Diff_Tail(pb[1], pd[1], bdy, bdytail);
2755 Two_Diff_Tail(pc[0], pd[0], cdx, cdxtail);
2756 Two_Diff_Tail(pc[1], pd[1], cdy, cdytail);
2757 if ((adxtail == 0.0) && (bdxtail == 0.0) && (cdxtail == 0.0)
2758 && (adytail == 0.0) && (bdytail == 0.0) && (cdytail == 0.0)) {
2759 return det;
2760 }
2761
2762 errbound = iccerrboundC * permanent + resulterrbound * Absolute(det);
2763 det += ((adx * adx + ady * ady) * ((bdx * cdytail + cdy * bdxtail)
2764 - (bdy * cdxtail + cdx * bdytail))
2765 + 2.0 * (adx * adxtail + ady * adytail) * (bdx * cdy - bdy * cdx))
2766 + ((bdx * bdx + bdy * bdy) * ((cdx * adytail + ady * cdxtail)
2767 - (cdy * adxtail + adx * cdytail))
2768 + 2.0 * (bdx * bdxtail + bdy * bdytail) * (cdx * ady - cdy * adx))
2769 + ((cdx * cdx + cdy * cdy) * ((adx * bdytail + bdy * adxtail)
2770 - (ady * bdxtail + bdx * adytail))
2771 + 2.0 * (cdx * cdxtail + cdy * cdytail) * (adx * bdy - ady * bdx));
2772 if ((det >= errbound) || (-det >= errbound)) {
2773 return det;
2774 }
2775
2776 finnow = fin1;
2777 finother = fin2;
2778
2779 if ((bdxtail != 0.0) || (bdytail != 0.0)
2780 || (cdxtail != 0.0) || (cdytail != 0.0)) {
2781 Square(adx, adxadx1, adxadx0);
2782 Square(ady, adyady1, adyady0);
2783 Two_Two_Sum(adxadx1, adxadx0, adyady1, adyady0, aa3, aa[2], aa[1], aa[0]);
2784 aa[3] = aa3;
2785 }
2786 if ((cdxtail != 0.0) || (cdytail != 0.0)
2787 || (adxtail != 0.0) || (adytail != 0.0)) {
2788 Square(bdx, bdxbdx1, bdxbdx0);
2789 Square(bdy, bdybdy1, bdybdy0);
2790 Two_Two_Sum(bdxbdx1, bdxbdx0, bdybdy1, bdybdy0, bb3, bb[2], bb[1], bb[0]);
2791 bb[3] = bb3;
2792 }
2793 if ((adxtail != 0.0) || (adytail != 0.0)
2794 || (bdxtail != 0.0) || (bdytail != 0.0)) {
2795 Square(cdx, cdxcdx1, cdxcdx0);
2796 Square(cdy, cdycdy1, cdycdy0);
2797 Two_Two_Sum(cdxcdx1, cdxcdx0, cdycdy1, cdycdy0, cc3, cc[2], cc[1], cc[0]);
2798 cc[3] = cc3;
2799 }
2800
2801 if (adxtail != 0.0) {
2802 axtbclen = scale_expansion_zeroelim(4, bc, adxtail, axtbc);
2803 temp16alen = scale_expansion_zeroelim(axtbclen, axtbc, 2.0 * adx,
2804 temp16a);
2805
2806 axtcclen = scale_expansion_zeroelim(4, cc, adxtail, axtcc);
2807 temp16blen = scale_expansion_zeroelim(axtcclen, axtcc, bdy, temp16b);
2808
2809 axtbblen = scale_expansion_zeroelim(4, bb, adxtail, axtbb);
2810 temp16clen = scale_expansion_zeroelim(axtbblen, axtbb, -cdy, temp16c);
2811
2812 temp32alen = fast_expansion_sum_zeroelim(temp16alen, temp16a,
2813 temp16blen, temp16b, temp32a);
2814 temp48len = fast_expansion_sum_zeroelim(temp16clen, temp16c,
2815 temp32alen, temp32a, temp48);
2816 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp48len,
2817 temp48, finother);
2818 finswap = finnow;
2819 finnow = finother;
2820 finother = finswap;
2821 }
2822 if (adytail != 0.0) {
2823 aytbclen = scale_expansion_zeroelim(4, bc, adytail, aytbc);
2824 temp16alen = scale_expansion_zeroelim(aytbclen, aytbc, 2.0 * ady,
2825 temp16a);
2826
2827 aytbblen = scale_expansion_zeroelim(4, bb, adytail, aytbb);
2828 temp16blen = scale_expansion_zeroelim(aytbblen, aytbb, cdx, temp16b);
2829
2830 aytcclen = scale_expansion_zeroelim(4, cc, adytail, aytcc);
2831 temp16clen = scale_expansion_zeroelim(aytcclen, aytcc, -bdx, temp16c);
2832
2833 temp32alen = fast_expansion_sum_zeroelim(temp16alen, temp16a,
2834 temp16blen, temp16b, temp32a);
2835 temp48len = fast_expansion_sum_zeroelim(temp16clen, temp16c,
2836 temp32alen, temp32a, temp48);
2837 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp48len,
2838 temp48, finother);
2839 finswap = finnow;
2840 finnow = finother;
2841 finother = finswap;
2842 }
2843 if (bdxtail != 0.0) {
2844 bxtcalen = scale_expansion_zeroelim(4, ca, bdxtail, bxtca);
2845 temp16alen = scale_expansion_zeroelim(bxtcalen, bxtca, 2.0 * bdx,
2846 temp16a);
2847
2848 bxtaalen = scale_expansion_zeroelim(4, aa, bdxtail, bxtaa);
2849 temp16blen = scale_expansion_zeroelim(bxtaalen, bxtaa, cdy, temp16b);
2850
2851 bxtcclen = scale_expansion_zeroelim(4, cc, bdxtail, bxtcc);
2852 temp16clen = scale_expansion_zeroelim(bxtcclen, bxtcc, -ady, temp16c);
2853
2854 temp32alen = fast_expansion_sum_zeroelim(temp16alen, temp16a,
2855 temp16blen, temp16b, temp32a);
2856 temp48len = fast_expansion_sum_zeroelim(temp16clen, temp16c,
2857 temp32alen, temp32a, temp48);
2858 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp48len,
2859 temp48, finother);
2860 finswap = finnow;
2861 finnow = finother;
2862 finother = finswap;
2863 }
2864 if (bdytail != 0.0) {
2865 bytcalen = scale_expansion_zeroelim(4, ca, bdytail, bytca);
2866 temp16alen = scale_expansion_zeroelim(bytcalen, bytca, 2.0 * bdy,
2867 temp16a);
2868
2869 bytcclen = scale_expansion_zeroelim(4, cc, bdytail, bytcc);
2870 temp16blen = scale_expansion_zeroelim(bytcclen, bytcc, adx, temp16b);
2871
2872 bytaalen = scale_expansion_zeroelim(4, aa, bdytail, bytaa);
2873 temp16clen = scale_expansion_zeroelim(bytaalen, bytaa, -cdx, temp16c);
2874
2875 temp32alen = fast_expansion_sum_zeroelim(temp16alen, temp16a,
2876 temp16blen, temp16b, temp32a);
2877 temp48len = fast_expansion_sum_zeroelim(temp16clen, temp16c,
2878 temp32alen, temp32a, temp48);
2879 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp48len,
2880 temp48, finother);
2881 finswap = finnow;
2882 finnow = finother;
2883 finother = finswap;
2884 }
2885 if (cdxtail != 0.0) {
2886 cxtablen = scale_expansion_zeroelim(4, ab, cdxtail, cxtab);
2887 temp16alen = scale_expansion_zeroelim(cxtablen, cxtab, 2.0 * cdx,
2888 temp16a);
2889
2890 cxtbblen = scale_expansion_zeroelim(4, bb, cdxtail, cxtbb);
2891 temp16blen = scale_expansion_zeroelim(cxtbblen, cxtbb, ady, temp16b);
2892
2893 cxtaalen = scale_expansion_zeroelim(4, aa, cdxtail, cxtaa);
2894 temp16clen = scale_expansion_zeroelim(cxtaalen, cxtaa, -bdy, temp16c);
2895
2896 temp32alen = fast_expansion_sum_zeroelim(temp16alen, temp16a,
2897 temp16blen, temp16b, temp32a);
2898 temp48len = fast_expansion_sum_zeroelim(temp16clen, temp16c,
2899 temp32alen, temp32a, temp48);
2900 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp48len,
2901 temp48, finother);
2902 finswap = finnow;
2903 finnow = finother;
2904 finother = finswap;
2905 }
2906 if (cdytail != 0.0) {
2907 cytablen = scale_expansion_zeroelim(4, ab, cdytail, cytab);
2908 temp16alen = scale_expansion_zeroelim(cytablen, cytab, 2.0 * cdy,
2909 temp16a);
2910
2911 cytaalen = scale_expansion_zeroelim(4, aa, cdytail, cytaa);
2912 temp16blen = scale_expansion_zeroelim(cytaalen, cytaa, bdx, temp16b);
2913
2914 cytbblen = scale_expansion_zeroelim(4, bb, cdytail, cytbb);
2915 temp16clen = scale_expansion_zeroelim(cytbblen, cytbb, -adx, temp16c);
2916
2917 temp32alen = fast_expansion_sum_zeroelim(temp16alen, temp16a,
2918 temp16blen, temp16b, temp32a);
2919 temp48len = fast_expansion_sum_zeroelim(temp16clen, temp16c,
2920 temp32alen, temp32a, temp48);
2921 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp48len,
2922 temp48, finother);
2923 finswap = finnow;
2924 finnow = finother;
2925 finother = finswap;
2926 }
2927
2928 if ((adxtail != 0.0) || (adytail != 0.0)) {
2929 if ((bdxtail != 0.0) || (bdytail != 0.0)
2930 || (cdxtail != 0.0) || (cdytail != 0.0)) {
2931 Two_Product(bdxtail, cdy, ti1, ti0);
2932 Two_Product(bdx, cdytail, tj1, tj0);
2933 Two_Two_Sum(ti1, ti0, tj1, tj0, u3, u[2], u[1], u[0]);
2934 u[3] = u3;
2935 negate = -bdy;
2936 Two_Product(cdxtail, negate, ti1, ti0);
2937 negate = -bdytail;
2938 Two_Product(cdx, negate, tj1, tj0);
2939 Two_Two_Sum(ti1, ti0, tj1, tj0, v3, v[2], v[1], v[0]);
2940 v[3] = v3;
2941 bctlen = fast_expansion_sum_zeroelim(4, u, 4, v, bct);
2942
2943 Two_Product(bdxtail, cdytail, ti1, ti0);
2944 Two_Product(cdxtail, bdytail, tj1, tj0);
2945 Two_Two_Diff(ti1, ti0, tj1, tj0, bctt3, bctt[2], bctt[1], bctt[0]);
2946 bctt[3] = bctt3;
2947 bcttlen = 4;
2948 } else {
2949 bct[0] = 0.0;
2950 bctlen = 1;
2951 bctt[0] = 0.0;
2952 bcttlen = 1;
2953 }
2954
2955 if (adxtail != 0.0) {
2956 temp16alen = scale_expansion_zeroelim(axtbclen, axtbc, adxtail, temp16a);
2957 axtbctlen = scale_expansion_zeroelim(bctlen, bct, adxtail, axtbct);
2958 temp32alen = scale_expansion_zeroelim(axtbctlen, axtbct, 2.0 * adx,
2959 temp32a);
2960 temp48len = fast_expansion_sum_zeroelim(temp16alen, temp16a,
2961 temp32alen, temp32a, temp48);
2962 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp48len,
2963 temp48, finother);
2964 finswap = finnow;
2965 finnow = finother;
2966 finother = finswap;
2967 if (bdytail != 0.0) {
2968 temp8len = scale_expansion_zeroelim(4, cc, adxtail, temp8);
2969 temp16alen = scale_expansion_zeroelim(temp8len, temp8, bdytail,
2970 temp16a);
2971 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp16alen,
2972 temp16a, finother);
2973 finswap = finnow;
2974 finnow = finother;
2975 finother = finswap;
2976 }
2977 if (cdytail != 0.0) {
2978 temp8len = scale_expansion_zeroelim(4, bb, -adxtail, temp8);
2979 temp16alen = scale_expansion_zeroelim(temp8len, temp8, cdytail,
2980 temp16a);
2981 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp16alen,
2982 temp16a, finother);
2983 finswap = finnow;
2984 finnow = finother;
2985 finother = finswap;
2986 }
2987
2988 temp32alen = scale_expansion_zeroelim(axtbctlen, axtbct, adxtail,
2989 temp32a);
2990 axtbcttlen = scale_expansion_zeroelim(bcttlen, bctt, adxtail, axtbctt);
2991 temp16alen = scale_expansion_zeroelim(axtbcttlen, axtbctt, 2.0 * adx,
2992 temp16a);
2993 temp16blen = scale_expansion_zeroelim(axtbcttlen, axtbctt, adxtail,
2994 temp16b);
2995 temp32blen = fast_expansion_sum_zeroelim(temp16alen, temp16a,
2996 temp16blen, temp16b, temp32b);
2997 temp64len = fast_expansion_sum_zeroelim(temp32alen, temp32a,
2998 temp32blen, temp32b, temp64);
2999 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp64len,
3000 temp64, finother);
3001 finswap = finnow;
3002 finnow = finother;
3003 finother = finswap;
3004 }
3005 if (adytail != 0.0) {
3006 temp16alen = scale_expansion_zeroelim(aytbclen, aytbc, adytail, temp16a);
3007 aytbctlen = scale_expansion_zeroelim(bctlen, bct, adytail, aytbct);
3008 temp32alen = scale_expansion_zeroelim(aytbctlen, aytbct, 2.0 * ady,
3009 temp32a);
3010 temp48len = fast_expansion_sum_zeroelim(temp16alen, temp16a,
3011 temp32alen, temp32a, temp48);
3012 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp48len,
3013 temp48, finother);
3014 finswap = finnow;
3015 finnow = finother;
3016 finother = finswap;
3017
3018
3019 temp32alen = scale_expansion_zeroelim(aytbctlen, aytbct, adytail,
3020 temp32a);
3021 aytbcttlen = scale_expansion_zeroelim(bcttlen, bctt, adytail, aytbctt);
3022 temp16alen = scale_expansion_zeroelim(aytbcttlen, aytbctt, 2.0 * ady,
3023 temp16a);
3024 temp16blen = scale_expansion_zeroelim(aytbcttlen, aytbctt, adytail,
3025 temp16b);
3026 temp32blen = fast_expansion_sum_zeroelim(temp16alen, temp16a,
3027 temp16blen, temp16b, temp32b);
3028 temp64len = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3029 temp32blen, temp32b, temp64);
3030 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp64len,
3031 temp64, finother);
3032 finswap = finnow;
3033 finnow = finother;
3034 finother = finswap;
3035 }
3036 }
3037 if ((bdxtail != 0.0) || (bdytail != 0.0)) {
3038 if ((cdxtail != 0.0) || (cdytail != 0.0)
3039 || (adxtail != 0.0) || (adytail != 0.0)) {
3040 Two_Product(cdxtail, ady, ti1, ti0);
3041 Two_Product(cdx, adytail, tj1, tj0);
3042 Two_Two_Sum(ti1, ti0, tj1, tj0, u3, u[2], u[1], u[0]);
3043 u[3] = u3;
3044 negate = -cdy;
3045 Two_Product(adxtail, negate, ti1, ti0);
3046 negate = -cdytail;
3047 Two_Product(adx, negate, tj1, tj0);
3048 Two_Two_Sum(ti1, ti0, tj1, tj0, v3, v[2], v[1], v[0]);
3049 v[3] = v3;
3050 catlen = fast_expansion_sum_zeroelim(4, u, 4, v, cat);
3051
3052 Two_Product(cdxtail, adytail, ti1, ti0);
3053 Two_Product(adxtail, cdytail, tj1, tj0);
3054 Two_Two_Diff(ti1, ti0, tj1, tj0, catt3, catt[2], catt[1], catt[0]);
3055 catt[3] = catt3;
3056 cattlen = 4;
3057 } else {
3058 cat[0] = 0.0;
3059 catlen = 1;
3060 catt[0] = 0.0;
3061 cattlen = 1;
3062 }
3063
3064 if (bdxtail != 0.0) {
3065 temp16alen = scale_expansion_zeroelim(bxtcalen, bxtca, bdxtail, temp16a);
3066 bxtcatlen = scale_expansion_zeroelim(catlen, cat, bdxtail, bxtcat);
3067 temp32alen = scale_expansion_zeroelim(bxtcatlen, bxtcat, 2.0 * bdx,
3068 temp32a);
3069 temp48len = fast_expansion_sum_zeroelim(temp16alen, temp16a,
3070 temp32alen, temp32a, temp48);
3071 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp48len,
3072 temp48, finother);
3073 finswap = finnow;
3074 finnow = finother;
3075 finother = finswap;
3076 if (cdytail != 0.0) {
3077 temp8len = scale_expansion_zeroelim(4, aa, bdxtail, temp8);
3078 temp16alen = scale_expansion_zeroelim(temp8len, temp8, cdytail,
3079 temp16a);
3080 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp16alen,
3081 temp16a, finother);
3082 finswap = finnow;
3083 finnow = finother;
3084 finother = finswap;
3085 }
3086 if (adytail != 0.0) {
3087 temp8len = scale_expansion_zeroelim(4, cc, -bdxtail, temp8);
3088 temp16alen = scale_expansion_zeroelim(temp8len, temp8, adytail,
3089 temp16a);
3090 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp16alen,
3091 temp16a, finother);
3092 finswap = finnow;
3093 finnow = finother;
3094 finother = finswap;
3095 }
3096
3097 temp32alen = scale_expansion_zeroelim(bxtcatlen, bxtcat, bdxtail,
3098 temp32a);
3099 bxtcattlen = scale_expansion_zeroelim(cattlen, catt, bdxtail, bxtcatt);
3100 temp16alen = scale_expansion_zeroelim(bxtcattlen, bxtcatt, 2.0 * bdx,
3101 temp16a);
3102 temp16blen = scale_expansion_zeroelim(bxtcattlen, bxtcatt, bdxtail,
3103 temp16b);
3104 temp32blen = fast_expansion_sum_zeroelim(temp16alen, temp16a,
3105 temp16blen, temp16b, temp32b);
3106 temp64len = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3107 temp32blen, temp32b, temp64);
3108 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp64len,
3109 temp64, finother);
3110 finswap = finnow;
3111 finnow = finother;
3112 finother = finswap;
3113 }
3114 if (bdytail != 0.0) {
3115 temp16alen = scale_expansion_zeroelim(bytcalen, bytca, bdytail, temp16a);
3116 bytcatlen = scale_expansion_zeroelim(catlen, cat, bdytail, bytcat);
3117 temp32alen = scale_expansion_zeroelim(bytcatlen, bytcat, 2.0 * bdy,
3118 temp32a);
3119 temp48len = fast_expansion_sum_zeroelim(temp16alen, temp16a,
3120 temp32alen, temp32a, temp48);
3121 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp48len,
3122 temp48, finother);
3123 finswap = finnow;
3124 finnow = finother;
3125 finother = finswap;
3126
3127
3128 temp32alen = scale_expansion_zeroelim(bytcatlen, bytcat, bdytail,
3129 temp32a);
3130 bytcattlen = scale_expansion_zeroelim(cattlen, catt, bdytail, bytcatt);
3131 temp16alen = scale_expansion_zeroelim(bytcattlen, bytcatt, 2.0 * bdy,
3132 temp16a);
3133 temp16blen = scale_expansion_zeroelim(bytcattlen, bytcatt, bdytail,
3134 temp16b);
3135 temp32blen = fast_expansion_sum_zeroelim(temp16alen, temp16a,
3136 temp16blen, temp16b, temp32b);
3137 temp64len = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3138 temp32blen, temp32b, temp64);
3139 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp64len,
3140 temp64, finother);
3141 finswap = finnow;
3142 finnow = finother;
3143 finother = finswap;
3144 }
3145 }
3146 if ((cdxtail != 0.0) || (cdytail != 0.0)) {
3147 if ((adxtail != 0.0) || (adytail != 0.0)
3148 || (bdxtail != 0.0) || (bdytail != 0.0)) {
3149 Two_Product(adxtail, bdy, ti1, ti0);
3150 Two_Product(adx, bdytail, tj1, tj0);
3151 Two_Two_Sum(ti1, ti0, tj1, tj0, u3, u[2], u[1], u[0]);
3152 u[3] = u3;
3153 negate = -ady;
3154 Two_Product(bdxtail, negate, ti1, ti0);
3155 negate = -adytail;
3156 Two_Product(bdx, negate, tj1, tj0);
3157 Two_Two_Sum(ti1, ti0, tj1, tj0, v3, v[2], v[1], v[0]);
3158 v[3] = v3;
3159 abtlen = fast_expansion_sum_zeroelim(4, u, 4, v, abt);
3160
3161 Two_Product(adxtail, bdytail, ti1, ti0);
3162 Two_Product(bdxtail, adytail, tj1, tj0);
3163 Two_Two_Diff(ti1, ti0, tj1, tj0, abtt3, abtt[2], abtt[1], abtt[0]);
3164 abtt[3] = abtt3;
3165 abttlen = 4;
3166 } else {
3167 abt[0] = 0.0;
3168 abtlen = 1;
3169 abtt[0] = 0.0;
3170 abttlen = 1;
3171 }
3172
3173 if (cdxtail != 0.0) {
3174 temp16alen = scale_expansion_zeroelim(cxtablen, cxtab, cdxtail, temp16a);
3175 cxtabtlen = scale_expansion_zeroelim(abtlen, abt, cdxtail, cxtabt);
3176 temp32alen = scale_expansion_zeroelim(cxtabtlen, cxtabt, 2.0 * cdx,
3177 temp32a);
3178 temp48len = fast_expansion_sum_zeroelim(temp16alen, temp16a,
3179 temp32alen, temp32a, temp48);
3180 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp48len,
3181 temp48, finother);
3182 finswap = finnow;
3183 finnow = finother;
3184 finother = finswap;
3185 if (adytail != 0.0) {
3186 temp8len = scale_expansion_zeroelim(4, bb, cdxtail, temp8);
3187 temp16alen = scale_expansion_zeroelim(temp8len, temp8, adytail,
3188 temp16a);
3189 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp16alen,
3190 temp16a, finother);
3191 finswap = finnow;
3192 finnow = finother;
3193 finother = finswap;
3194 }
3195 if (bdytail != 0.0) {
3196 temp8len = scale_expansion_zeroelim(4, aa, -cdxtail, temp8);
3197 temp16alen = scale_expansion_zeroelim(temp8len, temp8, bdytail,
3198 temp16a);
3199 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp16alen,
3200 temp16a, finother);
3201 finswap = finnow;
3202 finnow = finother;
3203 finother = finswap;
3204 }
3205
3206 temp32alen = scale_expansion_zeroelim(cxtabtlen, cxtabt, cdxtail,
3207 temp32a);
3208 cxtabttlen = scale_expansion_zeroelim(abttlen, abtt, cdxtail, cxtabtt);
3209 temp16alen = scale_expansion_zeroelim(cxtabttlen, cxtabtt, 2.0 * cdx,
3210 temp16a);
3211 temp16blen = scale_expansion_zeroelim(cxtabttlen, cxtabtt, cdxtail,
3212 temp16b);
3213 temp32blen = fast_expansion_sum_zeroelim(temp16alen, temp16a,
3214 temp16blen, temp16b, temp32b);
3215 temp64len = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3216 temp32blen, temp32b, temp64);
3217 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp64len,
3218 temp64, finother);
3219 finswap = finnow;
3220 finnow = finother;
3221 finother = finswap;
3222 }
3223 if (cdytail != 0.0) {
3224 temp16alen = scale_expansion_zeroelim(cytablen, cytab, cdytail, temp16a);
3225 cytabtlen = scale_expansion_zeroelim(abtlen, abt, cdytail, cytabt);
3226 temp32alen = scale_expansion_zeroelim(cytabtlen, cytabt, 2.0 * cdy,
3227 temp32a);
3228 temp48len = fast_expansion_sum_zeroelim(temp16alen, temp16a,
3229 temp32alen, temp32a, temp48);
3230 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp48len,
3231 temp48, finother);
3232 finswap = finnow;
3233 finnow = finother;
3234 finother = finswap;
3235
3236
3237 temp32alen = scale_expansion_zeroelim(cytabtlen, cytabt, cdytail,
3238 temp32a);
3239 cytabttlen = scale_expansion_zeroelim(abttlen, abtt, cdytail, cytabtt);
3240 temp16alen = scale_expansion_zeroelim(cytabttlen, cytabtt, 2.0 * cdy,
3241 temp16a);
3242 temp16blen = scale_expansion_zeroelim(cytabttlen, cytabtt, cdytail,
3243 temp16b);
3244 temp32blen = fast_expansion_sum_zeroelim(temp16alen, temp16a,
3245 temp16blen, temp16b, temp32b);
3246 temp64len = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3247 temp32blen, temp32b, temp64);
3248 finlength = fast_expansion_sum_zeroelim(finlength, finnow, temp64len,
3249 temp64, finother);
3250 finswap = finnow;
3251 finnow = finother;
3252 finother = finswap;
3253 }
3254 }
3255
3256 return finnow[finlength - 1];
3257 }
3258
3259 REAL incircle(REAL *pa, REAL *pb, REAL *pc, REAL *pd)
3260 {
3261 REAL adx, bdx, cdx, ady, bdy, cdy;
3262 REAL bdxcdy, cdxbdy, cdxady, adxcdy, adxbdy, bdxady;
3263 REAL alift, blift, clift;
3264 REAL det;
3265 REAL permanent, errbound;
3266
3267 adx = pa[0] - pd[0];
3268 bdx = pb[0] - pd[0];
3269 cdx = pc[0] - pd[0];
3270 ady = pa[1] - pd[1];
3271 bdy = pb[1] - pd[1];
3272 cdy = pc[1] - pd[1];
3273
3274 bdxcdy = bdx * cdy;
3275 cdxbdy = cdx * bdy;
3276 alift = adx * adx + ady * ady;
3277
3278 cdxady = cdx * ady;
3279 adxcdy = adx * cdy;
3280 blift = bdx * bdx + bdy * bdy;
3281
3282 adxbdy = adx * bdy;
3283 bdxady = bdx * ady;
3284 clift = cdx * cdx + cdy * cdy;
3285
3286 det = alift * (bdxcdy - cdxbdy)
3287 + blift * (cdxady - adxcdy)
3288 + clift * (adxbdy - bdxady);
3289
3290 permanent = (Absolute(bdxcdy) + Absolute(cdxbdy)) * alift
3291 + (Absolute(cdxady) + Absolute(adxcdy)) * blift
3292 + (Absolute(adxbdy) + Absolute(bdxady)) * clift;
3293 errbound = iccerrboundA * permanent;
3294 if ((det > errbound) || (-det > errbound)) {
3295 return det;
3296 }
3297
3298 return incircleadapt(pa, pb, pc, pd, permanent);
3299 }
3300
3301 /*****************************************************************************/
3302 /* */
3303 /* inspherefast() Approximate 3D insphere test. Nonrobust. */
3304 /* insphereexact() Exact 3D insphere test. Robust. */
3305 /* insphereslow() Another exact 3D insphere test. Robust. */
3306 /* insphere() Adaptive exact 3D insphere test. Robust. */
3307 /* */
3308 /* Return a positive value if the point pe lies inside the */
3309 /* sphere passing through pa, pb, pc, and pd; a negative value */
3310 /* if it lies outside; and zero if the five points are */
3311 /* cospherical. The points pa, pb, pc, and pd must be ordered */
3312 /* so that they have a positive orientation (as defined by */
3313 /* orient3d()), or the sign of the result will be reversed. */
3314 /* */
3315 /* Only the first and last routine should be used; the middle two are for */
3316 /* timings. */
3317 /* */
3318 /* The last three use exact arithmetic to ensure a correct answer. The */
3319 /* result returned is the determinant of a matrix. In insphere() only, */
3320 /* this determinant is computed adaptively, in the sense that exact */
3321 /* arithmetic is used only to the degree it is needed to ensure that the */
3322 /* returned value has the correct sign. Hence, insphere() is usually quite */
3323 /* fast, but will run more slowly when the input points are cospherical or */
3324 /* nearly so. */
3325 /* */
3326 /*****************************************************************************/
3327
3328 REAL inspherefast(REAL *pa, REAL *pb, REAL *pc, REAL *pd, REAL *pe)
3329 {
3330 REAL aex, bex, cex, dex;
3331 REAL aey, bey, cey, dey;
3332 REAL aez, bez, cez, dez;
3333 REAL alift, blift, clift, dlift;
3334 REAL ab, bc, cd, da, ac, bd;
3335 REAL abc, bcd, cda, dab;
3336
3337 aex = pa[0] - pe[0];
3338 bex = pb[0] - pe[0];
3339 cex = pc[0] - pe[0];
3340 dex = pd[0] - pe[0];
3341 aey = pa[1] - pe[1];
3342 bey = pb[1] - pe[1];
3343 cey = pc[1] - pe[1];
3344 dey = pd[1] - pe[1];
3345 aez = pa[2] - pe[2];
3346 bez = pb[2] - pe[2];
3347 cez = pc[2] - pe[2];
3348 dez = pd[2] - pe[2];
3349
3350 ab = aex * bey - bex * aey;
3351 bc = bex * cey - cex * bey;
3352 cd = cex * dey - dex * cey;
3353 da = dex * aey - aex * dey;
3354
3355 ac = aex * cey - cex * aey;
3356 bd = bex * dey - dex * bey;
3357
3358 abc = aez * bc - bez * ac + cez * ab;
3359 bcd = bez * cd - cez * bd + dez * bc;
3360 cda = cez * da + dez * ac + aez * cd;
3361 dab = dez * ab + aez * bd + bez * da;
3362
3363 alift = aex * aex + aey * aey + aez * aez;
3364 blift = bex * bex + bey * bey + bez * bez;
3365 clift = cex * cex + cey * cey + cez * cez;
3366 dlift = dex * dex + dey * dey + dez * dez;
3367
3368 return (dlift * abc - clift * dab) + (blift * cda - alift * bcd);
3369 }
3370
3371 REAL insphereexact(REAL *pa, REAL *pb, REAL *pc, REAL *pd, REAL *pe)
3372 {
3373 INEXACT REAL axby1, bxcy1, cxdy1, dxey1, exay1;
3374 INEXACT REAL bxay1, cxby1, dxcy1, exdy1, axey1;
3375 INEXACT REAL axcy1, bxdy1, cxey1, dxay1, exby1;
3376 INEXACT REAL cxay1, dxby1, excy1, axdy1, bxey1;
3377 REAL axby0, bxcy0, cxdy0, dxey0, exay0;
3378 REAL bxay0, cxby0, dxcy0, exdy0, axey0;
3379 REAL axcy0, bxdy0, cxey0, dxay0, exby0;
3380 REAL cxay0, dxby0, excy0, axdy0, bxey0;
3381 REAL ab[4], bc[4], cd[4], de[4], ea[4];
3382 REAL ac[4], bd[4], ce[4], da[4], eb[4];
3383 REAL temp8a[8], temp8b[8], temp16[16];
3384 int temp8alen, temp8blen, temp16len;
3385 REAL abc[24], bcd[24], cde[24], dea[24], eab[24];
3386 REAL abd[24], bce[24], cda[24], deb[24], eac[24];
3387 int abclen, bcdlen, cdelen, dealen, eablen;
3388 int abdlen, bcelen, cdalen, deblen, eaclen;
3389 REAL temp48a[48], temp48b[48];
3390 int temp48alen, temp48blen;
3391 REAL abcd[96], bcde[96], cdea[96], deab[96], eabc[96];
3392 int abcdlen, bcdelen, cdealen, deablen, eabclen;
3393 REAL temp192[192];
3394 REAL det384x[384], det384y[384], det384z[384];
3395 int xlen, ylen, zlen;
3396 REAL detxy[768];
3397 int xylen;
3398 REAL adet[1152], bdet[1152], cdet[1152], ddet[1152], edet[1152];
3399 int alen, blen, clen, dlen, elen;
3400 REAL abdet[2304], cddet[2304], cdedet[3456];
3401 int ablen, cdlen;
3402 REAL deter[5760];
3403 int deterlen;
3404 int i;
3405
3406 INEXACT REAL bvirt;
3407 REAL avirt, bround, around;
3408 INEXACT REAL c;
3409 INEXACT REAL abig;
3410 REAL ahi, alo, bhi, blo;
3411 REAL err1, err2, err3;
3412 INEXACT REAL _i, _j;
3413 REAL _0;
3414
3415 Two_Product(pa[0], pb[1], axby1, axby0);
3416 Two_Product(pb[0], pa[1], bxay1, bxay0);
3417 Two_Two_Diff(axby1, axby0, bxay1, bxay0, ab[3], ab[2], ab[1], ab[0]);
3418
3419 Two_Product(pb[0], pc[1], bxcy1, bxcy0);
3420 Two_Product(pc[0], pb[1], cxby1, cxby0);
3421 Two_Two_Diff(bxcy1, bxcy0, cxby1, cxby0, bc[3], bc[2], bc[1], bc[0]);
3422
3423 Two_Product(pc[0], pd[1], cxdy1, cxdy0);
3424 Two_Product(pd[0], pc[1], dxcy1, dxcy0);
3425 Two_Two_Diff(cxdy1, cxdy0, dxcy1, dxcy0, cd[3], cd[2], cd[1], cd[0]);
3426
3427 Two_Product(pd[0], pe[1], dxey1, dxey0);
3428 Two_Product(pe[0], pd[1], exdy1, exdy0);
3429 Two_Two_Diff(dxey1, dxey0, exdy1, exdy0, de[3], de[2], de[1], de[0]);
3430
3431 Two_Product(pe[0], pa[1], exay1, exay0);
3432 Two_Product(pa[0], pe[1], axey1, axey0);
3433 Two_Two_Diff(exay1, exay0, axey1, axey0, ea[3], ea[2], ea[1], ea[0]);
3434
3435 Two_Product(pa[0], pc[1], axcy1, axcy0);
3436 Two_Product(pc[0], pa[1], cxay1, cxay0);
3437 Two_Two_Diff(axcy1, axcy0, cxay1, cxay0, ac[3], ac[2], ac[1], ac[0]);
3438
3439 Two_Product(pb[0], pd[1], bxdy1, bxdy0);
3440 Two_Product(pd[0], pb[1], dxby1, dxby0);
3441 Two_Two_Diff(bxdy1, bxdy0, dxby1, dxby0, bd[3], bd[2], bd[1], bd[0]);
3442
3443 Two_Product(pc[0], pe[1], cxey1, cxey0);
3444 Two_Product(pe[0], pc[1], excy1, excy0);
3445 Two_Two_Diff(cxey1, cxey0, excy1, excy0, ce[3], ce[2], ce[1], ce[0]);
3446
3447 Two_Product(pd[0], pa[1], dxay1, dxay0);
3448 Two_Product(pa[0], pd[1], axdy1, axdy0);
3449 Two_Two_Diff(dxay1, dxay0, axdy1, axdy0, da[3], da[2], da[1], da[0]);
3450
3451 Two_Product(pe[0], pb[1], exby1, exby0);
3452 Two_Product(pb[0], pe[1], bxey1, bxey0);
3453 Two_Two_Diff(exby1, exby0, bxey1, bxey0, eb[3], eb[2], eb[1], eb[0]);
3454
3455 temp8alen = scale_expansion_zeroelim(4, bc, pa[2], temp8a);
3456 temp8blen = scale_expansion_zeroelim(4, ac, -pb[2], temp8b);
3457 temp16len = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp8blen, temp8b,
3458 temp16);
3459 temp8alen = scale_expansion_zeroelim(4, ab, pc[2], temp8a);
3460 abclen = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp16len, temp16,
3461 abc);
3462
3463 temp8alen = scale_expansion_zeroelim(4, cd, pb[2], temp8a);
3464 temp8blen = scale_expansion_zeroelim(4, bd, -pc[2], temp8b);
3465 temp16len = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp8blen, temp8b,
3466 temp16);
3467 temp8alen = scale_expansion_zeroelim(4, bc, pd[2], temp8a);
3468 bcdlen = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp16len, temp16,
3469 bcd);
3470
3471 temp8alen = scale_expansion_zeroelim(4, de, pc[2], temp8a);
3472 temp8blen = scale_expansion_zeroelim(4, ce, -pd[2], temp8b);
3473 temp16len = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp8blen, temp8b,
3474 temp16);
3475 temp8alen = scale_expansion_zeroelim(4, cd, pe[2], temp8a);
3476 cdelen = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp16len, temp16,
3477 cde);
3478
3479 temp8alen = scale_expansion_zeroelim(4, ea, pd[2], temp8a);
3480 temp8blen = scale_expansion_zeroelim(4, da, -pe[2], temp8b);
3481 temp16len = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp8blen, temp8b,
3482 temp16);
3483 temp8alen = scale_expansion_zeroelim(4, de, pa[2], temp8a);
3484 dealen = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp16len, temp16,
3485 dea);
3486
3487 temp8alen = scale_expansion_zeroelim(4, ab, pe[2], temp8a);
3488 temp8blen = scale_expansion_zeroelim(4, eb, -pa[2], temp8b);
3489 temp16len = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp8blen, temp8b,
3490 temp16);
3491 temp8alen = scale_expansion_zeroelim(4, ea, pb[2], temp8a);
3492 eablen = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp16len, temp16,
3493 eab);
3494
3495 temp8alen = scale_expansion_zeroelim(4, bd, pa[2], temp8a);
3496 temp8blen = scale_expansion_zeroelim(4, da, pb[2], temp8b);
3497 temp16len = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp8blen, temp8b,
3498 temp16);
3499 temp8alen = scale_expansion_zeroelim(4, ab, pd[2], temp8a);
3500 abdlen = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp16len, temp16,
3501 abd);
3502
3503 temp8alen = scale_expansion_zeroelim(4, ce, pb[2], temp8a);
3504 temp8blen = scale_expansion_zeroelim(4, eb, pc[2], temp8b);
3505 temp16len = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp8blen, temp8b,
3506 temp16);
3507 temp8alen = scale_expansion_zeroelim(4, bc, pe[2], temp8a);
3508 bcelen = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp16len, temp16,
3509 bce);
3510
3511 temp8alen = scale_expansion_zeroelim(4, da, pc[2], temp8a);
3512 temp8blen = scale_expansion_zeroelim(4, ac, pd[2], temp8b);
3513 temp16len = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp8blen, temp8b,
3514 temp16);
3515 temp8alen = scale_expansion_zeroelim(4, cd, pa[2], temp8a);
3516 cdalen = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp16len, temp16,
3517 cda);
3518
3519 temp8alen = scale_expansion_zeroelim(4, eb, pd[2], temp8a);
3520 temp8blen = scale_expansion_zeroelim(4, bd, pe[2], temp8b);
3521 temp16len = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp8blen, temp8b,
3522 temp16);
3523 temp8alen = scale_expansion_zeroelim(4, de, pb[2], temp8a);
3524 deblen = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp16len, temp16,
3525 deb);
3526
3527 temp8alen = scale_expansion_zeroelim(4, ac, pe[2], temp8a);
3528 temp8blen = scale_expansion_zeroelim(4, ce, pa[2], temp8b);
3529 temp16len = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp8blen, temp8b,
3530 temp16);
3531 temp8alen = scale_expansion_zeroelim(4, ea, pc[2], temp8a);
3532 eaclen = fast_expansion_sum_zeroelim(temp8alen, temp8a, temp16len, temp16,
3533 eac);
3534
3535 temp48alen = fast_expansion_sum_zeroelim(cdelen, cde, bcelen, bce, temp48a);
3536 temp48blen = fast_expansion_sum_zeroelim(deblen, deb, bcdlen, bcd, temp48b);
3537 for (i = 0; i < temp48blen; i++) {
3538 temp48b[i] = -temp48b[i];
3539 }
3540 bcdelen = fast_expansion_sum_zeroelim(temp48alen, temp48a,
3541 temp48blen, temp48b, bcde);
3542 xlen = scale_expansion_zeroelim(bcdelen, bcde, pa[0], temp192);
3543 xlen = scale_expansion_zeroelim(xlen, temp192, pa[0], det384x);
3544 ylen = scale_expansion_zeroelim(bcdelen, bcde, pa[1], temp192);
3545 ylen = scale_expansion_zeroelim(ylen, temp192, pa[1], det384y);
3546 zlen = scale_expansion_zeroelim(bcdelen, bcde, pa[2], temp192);
3547 zlen = scale_expansion_zeroelim(zlen, temp192, pa[2], det384z);
3548 xylen = fast_expansion_sum_zeroelim(xlen, det384x, ylen, det384y, detxy);
3549 alen = fast_expansion_sum_zeroelim(xylen, detxy, zlen, det384z, adet);
3550
3551 temp48alen = fast_expansion_sum_zeroelim(dealen, dea, cdalen, cda, temp48a);
3552 temp48blen = fast_expansion_sum_zeroelim(eaclen, eac, cdelen, cde, temp48b);
3553 for (i = 0; i < temp48blen; i++) {
3554 temp48b[i] = -temp48b[i];
3555 }
3556 cdealen = fast_expansion_sum_zeroelim(temp48alen, temp48a,
3557 temp48blen, temp48b, cdea);
3558 xlen = scale_expansion_zeroelim(cdealen, cdea, pb[0], temp192);
3559 xlen = scale_expansion_zeroelim(xlen, temp192, pb[0], det384x);
3560 ylen = scale_expansion_zeroelim(cdealen, cdea, pb[1], temp192);
3561 ylen = scale_expansion_zeroelim(ylen, temp192, pb[1], det384y);
3562 zlen = scale_expansion_zeroelim(cdealen, cdea, pb[2], temp192);
3563 zlen = scale_expansion_zeroelim(zlen, temp192, pb[2], det384z);
3564 xylen = fast_expansion_sum_zeroelim(xlen, det384x, ylen, det384y, detxy);
3565 blen = fast_expansion_sum_zeroelim(xylen, detxy, zlen, det384z, bdet);
3566
3567 temp48alen = fast_expansion_sum_zeroelim(eablen, eab, deblen, deb, temp48a);
3568 temp48blen = fast_expansion_sum_zeroelim(abdlen, abd, dealen, dea, temp48b);
3569 for (i = 0; i < temp48blen; i++) {
3570 temp48b[i] = -temp48b[i];
3571 }
3572 deablen = fast_expansion_sum_zeroelim(temp48alen, temp48a,
3573 temp48blen, temp48b, deab);
3574 xlen = scale_expansion_zeroelim(deablen, deab, pc[0], temp192);
3575 xlen = scale_expansion_zeroelim(xlen, temp192, pc[0], det384x);
3576 ylen = scale_expansion_zeroelim(deablen, deab, pc[1], temp192);
3577 ylen = scale_expansion_zeroelim(ylen, temp192, pc[1], det384y);
3578 zlen = scale_expansion_zeroelim(deablen, deab, pc[2], temp192);
3579 zlen = scale_expansion_zeroelim(zlen, temp192, pc[2], det384z);
3580 xylen = fast_expansion_sum_zeroelim(xlen, det384x, ylen, det384y, detxy);
3581 clen = fast_expansion_sum_zeroelim(xylen, detxy, zlen, det384z, cdet);
3582
3583 temp48alen = fast_expansion_sum_zeroelim(abclen, abc, eaclen, eac, temp48a);
3584 temp48blen = fast_expansion_sum_zeroelim(bcelen, bce, eablen, eab, temp48b);
3585 for (i = 0; i < temp48blen; i++) {
3586 temp48b[i] = -temp48b[i];
3587 }
3588 eabclen = fast_expansion_sum_zeroelim(temp48alen, temp48a,
3589 temp48blen, temp48b, eabc);
3590 xlen = scale_expansion_zeroelim(eabclen, eabc, pd[0], temp192);
3591 xlen = scale_expansion_zeroelim(xlen, temp192, pd[0], det384x);
3592 ylen = scale_expansion_zeroelim(eabclen, eabc, pd[1], temp192);
3593 ylen = scale_expansion_zeroelim(ylen, temp192, pd[1], det384y);
3594 zlen = scale_expansion_zeroelim(eabclen, eabc, pd[2], temp192);
3595 zlen = scale_expansion_zeroelim(zlen, temp192, pd[2], det384z);
3596 xylen = fast_expansion_sum_zeroelim(xlen, det384x, ylen, det384y, detxy);
3597 dlen = fast_expansion_sum_zeroelim(xylen, detxy, zlen, det384z, ddet);
3598
3599 temp48alen = fast_expansion_sum_zeroelim(bcdlen, bcd, abdlen, abd, temp48a);
3600 temp48blen = fast_expansion_sum_zeroelim(cdalen, cda, abclen, abc, temp48b);
3601 for (i = 0; i < temp48blen; i++) {
3602 temp48b[i] = -temp48b[i];
3603 }
3604 abcdlen = fast_expansion_sum_zeroelim(temp48alen, temp48a,
3605 temp48blen, temp48b, abcd);
3606 xlen = scale_expansion_zeroelim(abcdlen, abcd, pe[0], temp192);
3607 xlen = scale_expansion_zeroelim(xlen, temp192, pe[0], det384x);
3608 ylen = scale_expansion_zeroelim(abcdlen, abcd, pe[1], temp192);
3609 ylen = scale_expansion_zeroelim(ylen, temp192, pe[1], det384y);
3610 zlen = scale_expansion_zeroelim(abcdlen, abcd, pe[2], temp192);
3611 zlen = scale_expansion_zeroelim(zlen, temp192, pe[2], det384z);
3612 xylen = fast_expansion_sum_zeroelim(xlen, det384x, ylen, det384y, detxy);
3613 elen = fast_expansion_sum_zeroelim(xylen, detxy, zlen, det384z, edet);
3614
3615 ablen = fast_expansion_sum_zeroelim(alen, adet, blen, bdet, abdet);
3616 cdlen = fast_expansion_sum_zeroelim(clen, cdet, dlen, ddet, cddet);
3617 cdelen = fast_expansion_sum_zeroelim(cdlen, cddet, elen, edet, cdedet);
3618 deterlen = fast_expansion_sum_zeroelim(ablen, abdet, cdelen, cdedet, deter);
3619
3620 return deter[deterlen - 1];
3621 }
3622
3623 REAL insphereslow(REAL *pa, REAL *pb, REAL *pc, REAL *pd, REAL *pe)
3624 {
3625 INEXACT REAL aex, bex, cex, dex, aey, bey, cey, dey, aez, bez, cez, dez;
3626 REAL aextail, bextail, cextail, dextail;
3627 REAL aeytail, beytail, ceytail, deytail;
3628 REAL aeztail, beztail, ceztail, deztail;
3629 REAL negate, negatetail;
3630 INEXACT REAL axby7, bxcy7, cxdy7, dxay7, axcy7, bxdy7;
3631 INEXACT REAL bxay7, cxby7, dxcy7, axdy7, cxay7, dxby7;
3632 REAL axby[8], bxcy[8], cxdy[8], dxay[8], axcy[8], bxdy[8];
3633 REAL bxay[8], cxby[8], dxcy[8], axdy[8], cxay[8], dxby[8];
3634 REAL ab[16], bc[16], cd[16], da[16], ac[16], bd[16];
3635 int ablen, bclen, cdlen, dalen, aclen, bdlen;
3636 REAL temp32a[32], temp32b[32], temp64a[64], temp64b[64], temp64c[64];
3637 int temp32alen, temp32blen, temp64alen, temp64blen, temp64clen;
3638 REAL temp128[128], temp192[192];
3639 int temp128len, temp192len;
3640 REAL detx[384], detxx[768], detxt[384], detxxt[768], detxtxt[768];
3641 int xlen, xxlen, xtlen, xxtlen, xtxtlen;
3642 REAL x1[1536], x2[2304];
3643 int x1len, x2len;
3644 REAL dety[384], detyy[768], detyt[384], detyyt[768], detytyt[768];
3645 int ylen, yylen, ytlen, yytlen, ytytlen;
3646 REAL y1[1536], y2[2304];
3647 int y1len, y2len;
3648 REAL detz[384], detzz[768], detzt[384], detzzt[768], detztzt[768];
3649 int zlen, zzlen, ztlen, zztlen, ztztlen;
3650 REAL z1[1536], z2[2304];
3651 int z1len, z2len;
3652 REAL detxy[4608];
3653 int xylen;
3654 REAL adet[6912], bdet[6912], cdet[6912], ddet[6912];
3655 int alen, blen, clen, dlen;
3656 REAL abdet[13824], cddet[13824], deter[27648];
3657 int deterlen;
3658 int i;
3659
3660 INEXACT REAL bvirt;
3661 REAL avirt, bround, around;
3662 INEXACT REAL c;
3663 INEXACT REAL abig;
3664 REAL a0hi, a0lo, a1hi, a1lo, bhi, blo;
3665 REAL err1, err2, err3;
3666 INEXACT REAL _i, _j, _k, _l, _m, _n;
3667 REAL _0, _1, _2;
3668
3669 Two_Diff(pa[0], pe[0], aex, aextail);
3670 Two_Diff(pa[1], pe[1], aey, aeytail);
3671 Two_Diff(pa[2], pe[2], aez, aeztail);
3672 Two_Diff(pb[0], pe[0], bex, bextail);
3673 Two_Diff(pb[1], pe[1], bey, beytail);
3674 Two_Diff(pb[2], pe[2], bez, beztail);
3675 Two_Diff(pc[0], pe[0], cex, cextail);
3676 Two_Diff(pc[1], pe[1], cey, ceytail);
3677 Two_Diff(pc[2], pe[2], cez, ceztail);
3678 Two_Diff(pd[0], pe[0], dex, dextail);
3679 Two_Diff(pd[1], pe[1], dey, deytail);
3680 Two_Diff(pd[2], pe[2], dez, deztail);
3681
3682 Two_Two_Product(aex, aextail, bey, beytail,
3683 axby7, axby[6], axby[5], axby[4],
3684 axby[3], axby[2], axby[1], axby[0]);
3685 axby[7] = axby7;
3686 negate = -aey;
3687 negatetail = -aeytail;
3688 Two_Two_Product(bex, bextail, negate, negatetail,
3689 bxay7, bxay[6], bxay[5], bxay[4],
3690 bxay[3], bxay[2], bxay[1], bxay[0]);
3691 bxay[7] = bxay7;
3692 ablen = fast_expansion_sum_zeroelim(8, axby, 8, bxay, ab);
3693 Two_Two_Product(bex, bextail, cey, ceytail,
3694 bxcy7, bxcy[6], bxcy[5], bxcy[4],
3695 bxcy[3], bxcy[2], bxcy[1], bxcy[0]);
3696 bxcy[7] = bxcy7;
3697 negate = -bey;
3698 negatetail = -beytail;
3699 Two_Two_Product(cex, cextail, negate, negatetail,
3700 cxby7, cxby[6], cxby[5], cxby[4],
3701 cxby[3], cxby[2], cxby[1], cxby[0]);
3702 cxby[7] = cxby7;
3703 bclen = fast_expansion_sum_zeroelim(8, bxcy, 8, cxby, bc);
3704 Two_Two_Product(cex, cextail, dey, deytail,
3705 cxdy7, cxdy[6], cxdy[5], cxdy[4],
3706 cxdy[3], cxdy[2], cxdy[1], cxdy[0]);
3707 cxdy[7] = cxdy7;
3708 negate = -cey;
3709 negatetail = -ceytail;
3710 Two_Two_Product(dex, dextail, negate, negatetail,
3711 dxcy7, dxcy[6], dxcy[5], dxcy[4],
3712 dxcy[3], dxcy[2], dxcy[1], dxcy[0]);
3713 dxcy[7] = dxcy7;
3714 cdlen = fast_expansion_sum_zeroelim(8, cxdy, 8, dxcy, cd);
3715 Two_Two_Product(dex, dextail, aey, aeytail,
3716 dxay7, dxay[6], dxay[5], dxay[4],
3717 dxay[3], dxay[2], dxay[1], dxay[0]);
3718 dxay[7] = dxay7;
3719 negate = -dey;
3720 negatetail = -deytail;
3721 Two_Two_Product(aex, aextail, negate, negatetail,
3722 axdy7, axdy[6], axdy[5], axdy[4],
3723 axdy[3], axdy[2], axdy[1], axdy[0]);
3724 axdy[7] = axdy7;
3725 dalen = fast_expansion_sum_zeroelim(8, dxay, 8, axdy, da);
3726 Two_Two_Product(aex, aextail, cey, ceytail,
3727 axcy7, axcy[6], axcy[5], axcy[4],
3728 axcy[3], axcy[2], axcy[1], axcy[0]);
3729 axcy[7] = axcy7;
3730 negate = -aey;
3731 negatetail = -aeytail;
3732 Two_Two_Product(cex, cextail, negate, negatetail,
3733 cxay7, cxay[6], cxay[5], cxay[4],
3734 cxay[3], cxay[2], cxay[1], cxay[0]);
3735 cxay[7] = cxay7;
3736 aclen = fast_expansion_sum_zeroelim(8, axcy, 8, cxay, ac);
3737 Two_Two_Product(bex, bextail, dey, deytail,
3738 bxdy7, bxdy[6], bxdy[5], bxdy[4],
3739 bxdy[3], bxdy[2], bxdy[1], bxdy[0]);
3740 bxdy[7] = bxdy7;
3741 negate = -bey;
3742 negatetail = -beytail;
3743 Two_Two_Product(dex, dextail, negate, negatetail,
3744 dxby7, dxby[6], dxby[5], dxby[4],
3745 dxby[3], dxby[2], dxby[1], dxby[0]);
3746 dxby[7] = dxby7;
3747 bdlen = fast_expansion_sum_zeroelim(8, bxdy, 8, dxby, bd);
3748
3749 temp32alen = scale_expansion_zeroelim(cdlen, cd, -bez, temp32a);
3750 temp32blen = scale_expansion_zeroelim(cdlen, cd, -beztail, temp32b);
3751 temp64alen = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3752 temp32blen, temp32b, temp64a);
3753 temp32alen = scale_expansion_zeroelim(bdlen, bd, cez, temp32a);
3754 temp32blen = scale_expansion_zeroelim(bdlen, bd, ceztail, temp32b);
3755 temp64blen = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3756 temp32blen, temp32b, temp64b);
3757 temp32alen = scale_expansion_zeroelim(bclen, bc, -dez, temp32a);
3758 temp32blen = scale_expansion_zeroelim(bclen, bc, -deztail, temp32b);
3759 temp64clen = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3760 temp32blen, temp32b, temp64c);
3761 temp128len = fast_expansion_sum_zeroelim(temp64alen, temp64a,
3762 temp64blen, temp64b, temp128);
3763 temp192len = fast_expansion_sum_zeroelim(temp64clen, temp64c,
3764 temp128len, temp128, temp192);
3765 xlen = scale_expansion_zeroelim(temp192len, temp192, aex, detx);
3766 xxlen = scale_expansion_zeroelim(xlen, detx, aex, detxx);
3767 xtlen = scale_expansion_zeroelim(temp192len, temp192, aextail, detxt);
3768 xxtlen = scale_expansion_zeroelim(xtlen, detxt, aex, detxxt);
3769 for (i = 0; i < xxtlen; i++) {
3770 detxxt[i] *= 2.0;
3771 }
3772 xtxtlen = scale_expansion_zeroelim(xtlen, detxt, aextail, detxtxt);
3773 x1len = fast_expansion_sum_zeroelim(xxlen, detxx, xxtlen, detxxt, x1);
3774 x2len = fast_expansion_sum_zeroelim(x1len, x1, xtxtlen, detxtxt, x2);
3775 ylen = scale_expansion_zeroelim(temp192len, temp192, aey, dety);
3776 yylen = scale_expansion_zeroelim(ylen, dety, aey, detyy);
3777 ytlen = scale_expansion_zeroelim(temp192len, temp192, aeytail, detyt);
3778 yytlen = scale_expansion_zeroelim(ytlen, detyt, aey, detyyt);
3779 for (i = 0; i < yytlen; i++) {
3780 detyyt[i] *= 2.0;
3781 }
3782 ytytlen = scale_expansion_zeroelim(ytlen, detyt, aeytail, detytyt);
3783 y1len = fast_expansion_sum_zeroelim(yylen, detyy, yytlen, detyyt, y1);
3784 y2len = fast_expansion_sum_zeroelim(y1len, y1, ytytlen, detytyt, y2);
3785 zlen = scale_expansion_zeroelim(temp192len, temp192, aez, detz);
3786 zzlen = scale_expansion_zeroelim(zlen, detz, aez, detzz);
3787 ztlen = scale_expansion_zeroelim(temp192len, temp192, aeztail, detzt);
3788 zztlen = scale_expansion_zeroelim(ztlen, detzt, aez, detzzt);
3789 for (i = 0; i < zztlen; i++) {
3790 detzzt[i] *= 2.0;
3791 }
3792 ztztlen = scale_expansion_zeroelim(ztlen, detzt, aeztail, detztzt);
3793 z1len = fast_expansion_sum_zeroelim(zzlen, detzz, zztlen, detzzt, z1);
3794 z2len = fast_expansion_sum_zeroelim(z1len, z1, ztztlen, detztzt, z2);
3795 xylen = fast_expansion_sum_zeroelim(x2len, x2, y2len, y2, detxy);
3796 alen = fast_expansion_sum_zeroelim(z2len, z2, xylen, detxy, adet);
3797
3798 temp32alen = scale_expansion_zeroelim(dalen, da, cez, temp32a);
3799 temp32blen = scale_expansion_zeroelim(dalen, da, ceztail, temp32b);
3800 temp64alen = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3801 temp32blen, temp32b, temp64a);
3802 temp32alen = scale_expansion_zeroelim(aclen, ac, dez, temp32a);
3803 temp32blen = scale_expansion_zeroelim(aclen, ac, deztail, temp32b);
3804 temp64blen = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3805 temp32blen, temp32b, temp64b);
3806 temp32alen = scale_expansion_zeroelim(cdlen, cd, aez, temp32a);
3807 temp32blen = scale_expansion_zeroelim(cdlen, cd, aeztail, temp32b);
3808 temp64clen = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3809 temp32blen, temp32b, temp64c);
3810 temp128len = fast_expansion_sum_zeroelim(temp64alen, temp64a,
3811 temp64blen, temp64b, temp128);
3812 temp192len = fast_expansion_sum_zeroelim(temp64clen, temp64c,
3813 temp128len, temp128, temp192);
3814 xlen = scale_expansion_zeroelim(temp192len, temp192, bex, detx);
3815 xxlen = scale_expansion_zeroelim(xlen, detx, bex, detxx);
3816 xtlen = scale_expansion_zeroelim(temp192len, temp192, bextail, detxt);
3817 xxtlen = scale_expansion_zeroelim(xtlen, detxt, bex, detxxt);
3818 for (i = 0; i < xxtlen; i++) {
3819 detxxt[i] *= 2.0;
3820 }
3821 xtxtlen = scale_expansion_zeroelim(xtlen, detxt, bextail, detxtxt);
3822 x1len = fast_expansion_sum_zeroelim(xxlen, detxx, xxtlen, detxxt, x1);
3823 x2len = fast_expansion_sum_zeroelim(x1len, x1, xtxtlen, detxtxt, x2);
3824 ylen = scale_expansion_zeroelim(temp192len, temp192, bey, dety);
3825 yylen = scale_expansion_zeroelim(ylen, dety, bey, detyy);
3826 ytlen = scale_expansion_zeroelim(temp192len, temp192, beytail, detyt);
3827 yytlen = scale_expansion_zeroelim(ytlen, detyt, bey, detyyt);
3828 for (i = 0; i < yytlen; i++) {
3829 detyyt[i] *= 2.0;
3830 }
3831 ytytlen = scale_expansion_zeroelim(ytlen, detyt, beytail, detytyt);
3832 y1len = fast_expansion_sum_zeroelim(yylen, detyy, yytlen, detyyt, y1);
3833 y2len = fast_expansion_sum_zeroelim(y1len, y1, ytytlen, detytyt, y2);
3834 zlen = scale_expansion_zeroelim(temp192len, temp192, bez, detz);
3835 zzlen = scale_expansion_zeroelim(zlen, detz, bez, detzz);
3836 ztlen = scale_expansion_zeroelim(temp192len, temp192, beztail, detzt);
3837 zztlen = scale_expansion_zeroelim(ztlen, detzt, bez, detzzt);
3838 for (i = 0; i < zztlen; i++) {
3839 detzzt[i] *= 2.0;
3840 }
3841 ztztlen = scale_expansion_zeroelim(ztlen, detzt, beztail, detztzt);
3842 z1len = fast_expansion_sum_zeroelim(zzlen, detzz, zztlen, detzzt, z1);
3843 z2len = fast_expansion_sum_zeroelim(z1len, z1, ztztlen, detztzt, z2);
3844 xylen = fast_expansion_sum_zeroelim(x2len, x2, y2len, y2, detxy);
3845 blen = fast_expansion_sum_zeroelim(z2len, z2, xylen, detxy, bdet);
3846
3847 temp32alen = scale_expansion_zeroelim(ablen, ab, -dez, temp32a);
3848 temp32blen = scale_expansion_zeroelim(ablen, ab, -deztail, temp32b);
3849 temp64alen = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3850 temp32blen, temp32b, temp64a);
3851 temp32alen = scale_expansion_zeroelim(bdlen, bd, -aez, temp32a);
3852 temp32blen = scale_expansion_zeroelim(bdlen, bd, -aeztail, temp32b);
3853 temp64blen = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3854 temp32blen, temp32b, temp64b);
3855 temp32alen = scale_expansion_zeroelim(dalen, da, -bez, temp32a);
3856 temp32blen = scale_expansion_zeroelim(dalen, da, -beztail, temp32b);
3857 temp64clen = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3858 temp32blen, temp32b, temp64c);
3859 temp128len = fast_expansion_sum_zeroelim(temp64alen, temp64a,
3860 temp64blen, temp64b, temp128);
3861 temp192len = fast_expansion_sum_zeroelim(temp64clen, temp64c,
3862 temp128len, temp128, temp192);
3863 xlen = scale_expansion_zeroelim(temp192len, temp192, cex, detx);
3864 xxlen = scale_expansion_zeroelim(xlen, detx, cex, detxx);
3865 xtlen = scale_expansion_zeroelim(temp192len, temp192, cextail, detxt);
3866 xxtlen = scale_expansion_zeroelim(xtlen, detxt, cex, detxxt);
3867 for (i = 0; i < xxtlen; i++) {
3868 detxxt[i] *= 2.0;
3869 }
3870 xtxtlen = scale_expansion_zeroelim(xtlen, detxt, cextail, detxtxt);
3871 x1len = fast_expansion_sum_zeroelim(xxlen, detxx, xxtlen, detxxt, x1);
3872 x2len = fast_expansion_sum_zeroelim(x1len, x1, xtxtlen, detxtxt, x2);
3873 ylen = scale_expansion_zeroelim(temp192len, temp192, cey, dety);
3874 yylen = scale_expansion_zeroelim(ylen, dety, cey, detyy);
3875 ytlen = scale_expansion_zeroelim(temp192len, temp192, ceytail, detyt);
3876 yytlen = scale_expansion_zeroelim(ytlen, detyt, cey, detyyt);
3877 for (i = 0; i < yytlen; i++) {
3878 detyyt[i] *= 2.0;
3879 }
3880 ytytlen = scale_expansion_zeroelim(ytlen, detyt, ceytail, detytyt);
3881 y1len = fast_expansion_sum_zeroelim(yylen, detyy, yytlen, detyyt, y1);
3882 y2len = fast_expansion_sum_zeroelim(y1len, y1, ytytlen, detytyt, y2);
3883 zlen = scale_expansion_zeroelim(temp192len, temp192, cez, detz);
3884 zzlen = scale_expansion_zeroelim(zlen, detz, cez, detzz);
3885 ztlen = scale_expansion_zeroelim(temp192len, temp192, ceztail, detzt);
3886 zztlen = scale_expansion_zeroelim(ztlen, detzt, cez, detzzt);
3887 for (i = 0; i < zztlen; i++) {
3888 detzzt[i] *= 2.0;
3889 }
3890 ztztlen = scale_expansion_zeroelim(ztlen, detzt, ceztail, detztzt);
3891 z1len = fast_expansion_sum_zeroelim(zzlen, detzz, zztlen, detzzt, z1);
3892 z2len = fast_expansion_sum_zeroelim(z1len, z1, ztztlen, detztzt, z2);
3893 xylen = fast_expansion_sum_zeroelim(x2len, x2, y2len, y2, detxy);
3894 clen = fast_expansion_sum_zeroelim(z2len, z2, xylen, detxy, cdet);
3895
3896 temp32alen = scale_expansion_zeroelim(bclen, bc, aez, temp32a);
3897 temp32blen = scale_expansion_zeroelim(bclen, bc, aeztail, temp32b);
3898 temp64alen = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3899 temp32blen, temp32b, temp64a);
3900 temp32alen = scale_expansion_zeroelim(aclen, ac, -bez, temp32a);
3901 temp32blen = scale_expansion_zeroelim(aclen, ac, -beztail, temp32b);
3902 temp64blen = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3903 temp32blen, temp32b, temp64b);
3904 temp32alen = scale_expansion_zeroelim(ablen, ab, cez, temp32a);
3905 temp32blen = scale_expansion_zeroelim(ablen, ab, ceztail, temp32b);
3906 temp64clen = fast_expansion_sum_zeroelim(temp32alen, temp32a,
3907 temp32blen, temp32b, temp64c);
3908 temp128len = fast_expansion_sum_zeroelim(temp64alen, temp64a,
3909 temp64blen, temp64b, temp128);
3910 temp192len = fast_expansion_sum_zeroelim(temp64clen, temp64c,
3911 temp128len, temp128, temp192);
3912 xlen = scale_expansion_zeroelim(temp192len, temp192, dex, detx);
3913 xxlen = scale_expansion_zeroelim(xlen, detx, dex, detxx);
3914 xtlen = scale_expansion_zeroelim(temp192len, temp192, dextail, detxt);
3915 xxtlen = scale_expansion_zeroelim(xtlen, detxt, dex, detxxt);
3916 for (i = 0; i < xxtlen; i++) {
3917 detxxt[i] *= 2.0;
3918 }
3919 xtxtlen = scale_expansion_zeroelim(xtlen, detxt, dextail, detxtxt);
3920 x1len = fast_expansion_sum_zeroelim(xxlen, detxx, xxtlen, detxxt, x1);
3921 x2len = fast_expansion_sum_zeroelim(x1len, x1, xtxtlen, detxtxt, x2);
3922 ylen = scale_expansion_zeroelim(temp192len, temp192, dey, dety);
3923 yylen = scale_expansion_zeroelim(ylen, dety, dey, detyy);
3924 ytlen = scale_expansion_zeroelim(temp192len, temp192, deytail, detyt);
3925 yytlen = scale_expansion_zeroelim(ytlen, detyt, dey, detyyt);
3926 for (i = 0; i < yytlen; i++) {
3927 detyyt[i] *= 2.0;
3928 }
3929 ytytlen = scale_expansion_zeroelim(ytlen, detyt, deytail, detytyt);
3930 y1len = fast_expansion_sum_zeroelim(yylen, detyy, yytlen, detyyt, y1);
3931 y2len = fast_expansion_sum_zeroelim(y1len, y1, ytytlen, detytyt, y2);
3932 zlen = scale_expansion_zeroelim(temp192len, temp192, dez, detz);
3933 zzlen = scale_expansion_zeroelim(zlen, detz, dez, detzz);
3934 ztlen = scale_expansion_zeroelim(temp192len, temp192, deztail, detzt);
3935 zztlen = scale_expansion_zeroelim(ztlen, detzt, dez, detzzt);
3936 for (i = 0; i < zztlen; i++) {
3937 detzzt[i] *= 2.0;
3938 }
3939 ztztlen = scale_expansion_zeroelim(ztlen, detzt, deztail, detztzt);
3940 z1len = fast_expansion_sum_zeroelim(zzlen, detzz, zztlen, detzzt, z1);
3941 z2len = fast_expansion_sum_zeroelim(z1len, z1, ztztlen, detztzt, z2);
3942 xylen = fast_expansion_sum_zeroelim(x2len, x2, y2len, y2, detxy);
3943 dlen = fast_expansion_sum_zeroelim(z2len, z2, xylen, detxy, ddet);
3944
3945 ablen = fast_expansion_sum_zeroelim(alen, adet, blen, bdet, abdet);
3946 cdlen = fast_expansion_sum_zeroelim(clen, cdet, dlen, ddet, cddet);
3947 deterlen = fast_expansion_sum_zeroelim(ablen, abdet, cdlen, cddet, deter);
3948
3949 return deter[deterlen - 1];
3950 }
3951
3952 REAL insphereadapt(REAL *pa, REAL *pb, REAL *pc, REAL *pd, REAL *pe,
3953 REAL permanent)
3954 {
3955 INEXACT REAL aex, bex, cex, dex, aey, bey, cey, dey, aez, bez, cez, dez;
3956 REAL det, errbound;
3957
3958 INEXACT REAL aexbey1, bexaey1, bexcey1, cexbey1;
3959 INEXACT REAL cexdey1, dexcey1, dexaey1, aexdey1;
3960 INEXACT REAL aexcey1, cexaey1, bexdey1, dexbey1;
3961 REAL aexbey0, bexaey0, bexcey0, cexbey0;
3962 REAL cexdey0, dexcey0, dexaey0, aexdey0;
3963 REAL aexcey0, cexaey0, bexdey0, dexbey0;
3964 REAL ab[4], bc[4], cd[4], da[4], ac[4], bd[4];
3965 INEXACT REAL ab3, bc3, cd3, da3, ac3, bd3;
3966 REAL abeps, bceps, cdeps, daeps, aceps, bdeps;
3967 REAL temp8a[8], temp8b[8], temp8c[8], temp16[16], temp24[24], temp48[48];
3968 int temp8alen, temp8blen, temp8clen, temp16len, temp24len, temp48len;
3969 REAL xdet[96], ydet[96], zdet[96], xydet[192];
3970 int xlen, ylen, zlen, xylen;
3971 REAL adet[288], bdet[288], cdet[288], ddet[288];
3972 int alen, blen, clen, dlen;
3973 REAL abdet[576], cddet[576];
3974 int ablen, cdlen;
3975 REAL fin1[1152];
3976 int finlength;
3977
3978 REAL aextail, bextail, cextail, dextail;
3979 REAL aeytail, beytail, ceytail, deytail;
3980 REAL aeztail, beztail, ceztail, deztail;
3981
3982 INEXACT REAL bvirt;
3983 REAL avirt, bround, around;
3984 INEXACT REAL c;
3985 INEXACT REAL abig;
3986 REAL ahi, alo, bhi, blo;
3987 REAL err1, err2, err3;
3988 INEXACT REAL _i, _j;
3989 REAL _0;
3990
3991 aex = (REAL) (pa[0] - pe[0]);
3992 bex = (REAL) (pb[0] - pe[0]);
3993 cex = (REAL) (pc[0] - pe[0]);
3994 dex = (REAL) (pd[0] - pe[0]);
3995 aey = (REAL) (pa[1] - pe[1]);
3996 bey = (REAL) (pb[1] - pe[1]);
3997 cey = (REAL) (pc[1] - pe[1]);
3998 dey = (REAL) (pd[1] - pe[1]);
3999 aez = (REAL) (pa[2] - pe[2]);
4000 bez = (REAL) (pb[2] - pe[2]);
4001 cez = (REAL) (pc[2] - pe[2]);
4002 dez = (REAL) (pd[2] - pe[2]);
4003
4004 Two_Product(aex, bey, aexbey1, aexbey0);
4005 Two_Product(bex, aey, bexaey1, bexaey0);
4006 Two_Two_Diff(aexbey1, aexbey0, bexaey1, bexaey0, ab3, ab[2], ab[1], ab[0]);
4007 ab[3] = ab3;
4008
4009 Two_Product(bex, cey, bexcey1, bexcey0);
4010 Two_Product(cex, bey, cexbey1, cexbey0);
4011 Two_Two_Diff(bexcey1, bexcey0, cexbey1, cexbey0, bc3, bc[2], bc[1], bc[0]);
4012 bc[3] = bc3;
4013
4014 Two_Product(cex, dey, cexdey1, cexdey0);
4015 Two_Product(dex, cey, dexcey1, dexcey0);
4016 Two_Two_Diff(cexdey1, cexdey0, dexcey1, dexcey0, cd3, cd[2], cd[1], cd[0]);
4017 cd[3] = cd3;
4018
4019 Two_Product(dex, aey, dexaey1, dexaey0);
4020 Two_Product(aex, dey, aexdey1, aexdey0);
4021 Two_Two_Diff(dexaey1, dexaey0, aexdey1, aexdey0, da3, da[2], da[1], da[0]);
4022 da[3] = da3;
4023
4024 Two_Product(aex, cey, aexcey1, aexcey0);
4025 Two_Product(cex, aey, cexaey1, cexaey0);
4026 Two_Two_Diff(aexcey1, aexcey0, cexaey1, cexaey0, ac3, ac[2], ac[1], ac[0]);
4027 ac[3] = ac3;
4028
4029 Two_Product(bex, dey, bexdey1, bexdey0);
4030 Two_Product(dex, bey, dexbey1, dexbey0);
4031 Two_Two_Diff(bexdey1, bexdey0, dexbey1, dexbey0, bd3, bd[2], bd[1], bd[0]);
4032 bd[3] = bd3;
4033
4034 temp8alen = scale_expansion_zeroelim(4, cd, bez, temp8a);
4035 temp8blen = scale_expansion_zeroelim(4, bd, -cez, temp8b);
4036 temp8clen = scale_expansion_zeroelim(4, bc, dez, temp8c);
4037 temp16len = fast_expansion_sum_zeroelim(temp8alen, temp8a,
4038 temp8blen, temp8b, temp16);
4039 temp24len = fast_expansion_sum_zeroelim(temp8clen, temp8c,
4040 temp16len, temp16, temp24);
4041 temp48len = scale_expansion_zeroelim(temp24len, temp24, aex, temp48);
4042 xlen = scale_expansion_zeroelim(temp48len, temp48, -aex, xdet);
4043 temp48len = scale_expansion_zeroelim(temp24len, temp24, aey, temp48);
4044 ylen = scale_expansion_zeroelim(temp48len, temp48, -aey, ydet);
4045 temp48len = scale_expansion_zeroelim(temp24len, temp24, aez, temp48);
4046 zlen = scale_expansion_zeroelim(temp48len, temp48, -aez, zdet);
4047 xylen = fast_expansion_sum_zeroelim(xlen, xdet, ylen, ydet, xydet);
4048 alen = fast_expansion_sum_zeroelim(xylen, xydet, zlen, zdet, adet);
4049
4050 temp8alen = scale_expansion_zeroelim(4, da, cez, temp8a);
4051 temp8blen = scale_expansion_zeroelim(4, ac, dez, temp8b);
4052 temp8clen = scale_expansion_zeroelim(4, cd, aez, temp8c);
4053 temp16len = fast_expansion_sum_zeroelim(temp8alen, temp8a,
4054 temp8blen, temp8b, temp16);
4055 temp24len = fast_expansion_sum_zeroelim(temp8clen, temp8c,
4056 temp16len, temp16, temp24);
4057 temp48len = scale_expansion_zeroelim(temp24len, temp24, bex, temp48);
4058 xlen = scale_expansion_zeroelim(temp48len, temp48, bex, xdet);
4059 temp48len = scale_expansion_zeroelim(temp24len, temp24, bey, temp48);
4060 ylen = scale_expansion_zeroelim(temp48len, temp48, bey, ydet);
4061 temp48len = scale_expansion_zeroelim(temp24len, temp24, bez, temp48);
4062 zlen = scale_expansion_zeroelim(temp48len, temp48, bez, zdet);
4063 xylen = fast_expansion_sum_zeroelim(xlen, xdet, ylen, ydet, xydet);
4064 blen = fast_expansion_sum_zeroelim(xylen, xydet, zlen, zdet, bdet);
4065
4066 temp8alen = scale_expansion_zeroelim(4, ab, dez, temp8a);
4067 temp8blen = scale_expansion_zeroelim(4, bd, aez, temp8b);
4068 temp8clen = scale_expansion_zeroelim(4, da, bez, temp8c);
4069 temp16len = fast_expansion_sum_zeroelim(temp8alen, temp8a,
4070 temp8blen, temp8b, temp16);
4071 temp24len = fast_expansion_sum_zeroelim(temp8clen, temp8c,
4072 temp16len, temp16, temp24);
4073 temp48len = scale_expansion_zeroelim(temp24len, temp24, cex, temp48);
4074 xlen = scale_expansion_zeroelim(temp48len, temp48, -cex, xdet);
4075 temp48len = scale_expansion_zeroelim(temp24len, temp24, cey, temp48);
4076 ylen = scale_expansion_zeroelim(temp48len, temp48, -cey, ydet);
4077 temp48len = scale_expansion_zeroelim(temp24len, temp24, cez, temp48);
4078 zlen = scale_expansion_zeroelim(temp48len, temp48, -cez, zdet);
4079 xylen = fast_expansion_sum_zeroelim(xlen, xdet, ylen, ydet, xydet);
4080 clen = fast_expansion_sum_zeroelim(xylen, xydet, zlen, zdet, cdet);
4081
4082 temp8alen = scale_expansion_zeroelim(4, bc, aez, temp8a);
4083 temp8blen = scale_expansion_zeroelim(4, ac, -bez, temp8b);
4084 temp8clen = scale_expansion_zeroelim(4, ab, cez, temp8c);
4085 temp16len = fast_expansion_sum_zeroelim(temp8alen, temp8a,
4086 temp8blen, temp8b, temp16);
4087 temp24len = fast_expansion_sum_zeroelim(temp8clen, temp8c,
4088 temp16len, temp16, temp24);
4089 temp48len = scale_expansion_zeroelim(temp24len, temp24, dex, temp48);
4090 xlen = scale_expansion_zeroelim(temp48len, temp48, dex, xdet);
4091 temp48len = scale_expansion_zeroelim(temp24len, temp24, dey, temp48);
4092 ylen = scale_expansion_zeroelim(temp48len, temp48, dey, ydet);
4093 temp48len = scale_expansion_zeroelim(temp24len, temp24, dez, temp48);
4094 zlen = scale_expansion_zeroelim(temp48len, temp48, dez, zdet);
4095 xylen = fast_expansion_sum_zeroelim(xlen, xdet, ylen, ydet, xydet);
4096 dlen = fast_expansion_sum_zeroelim(xylen, xydet, zlen, zdet, ddet);
4097
4098 ablen = fast_expansion_sum_zeroelim(alen, adet, blen, bdet, abdet);
4099 cdlen = fast_expansion_sum_zeroelim(clen, cdet, dlen, ddet, cddet);
4100 finlength = fast_expansion_sum_zeroelim(ablen, abdet, cdlen, cddet, fin1);
4101
4102 det = estimate(finlength, fin1);
4103 errbound = isperrboundB * permanent;
4104 if ((det >= errbound) || (-det >= errbound)) {
4105 return det;
4106 }
4107
4108 Two_Diff_Tail(pa[0], pe[0], aex, aextail);
4109 Two_Diff_Tail(pa[1], pe[1], aey, aeytail);
4110 Two_Diff_Tail(pa[2], pe[2], aez, aeztail);
4111 Two_Diff_Tail(pb[0], pe[0], bex, bextail);
4112 Two_Diff_Tail(pb[1], pe[1], bey, beytail);
4113 Two_Diff_Tail(pb[2], pe[2], bez, beztail);
4114 Two_Diff_Tail(pc[0], pe[0], cex, cextail);
4115 Two_Diff_Tail(pc[1], pe[1], cey, ceytail);
4116 Two_Diff_Tail(pc[2], pe[2], cez, ceztail);
4117 Two_Diff_Tail(pd[0], pe[0], dex, dextail);
4118 Two_Diff_Tail(pd[1], pe[1], dey, deytail);
4119 Two_Diff_Tail(pd[2], pe[2], dez, deztail);
4120 if ((aextail == 0.0) && (aeytail == 0.0) && (aeztail == 0.0)
4121 && (bextail == 0.0) && (beytail == 0.0) && (beztail == 0.0)
4122 && (cextail == 0.0) && (ceytail == 0.0) && (ceztail == 0.0)
4123 && (dextail == 0.0) && (deytail == 0.0) && (deztail == 0.0)) {
4124 return det;
4125 }
4126
4127 errbound = isperrboundC * permanent + resulterrbound * Absolute(det);
4128 abeps = (aex * beytail + bey * aextail)
4129 - (aey * bextail + bex * aeytail);
4130 bceps = (bex * ceytail + cey * bextail)
4131 - (bey * cextail + cex * beytail);
4132 cdeps = (cex * deytail + dey * cextail)
4133 - (cey * dextail + dex * ceytail);
4134 daeps = (dex * aeytail + aey * dextail)
4135 - (dey * aextail + aex * deytail);
4136 aceps = (aex * ceytail + cey * aextail)
4137 - (aey * cextail + cex * aeytail);
4138 bdeps = (bex * deytail + dey * bextail)
4139 - (bey * dextail + dex * beytail);
4140 det += (((bex * bex + bey * bey + bez * bez)
4141 * ((cez * daeps + dez * aceps + aez * cdeps)
4142 + (ceztail * da3 + deztail * ac3 + aeztail * cd3))
4143 + (dex * dex + dey * dey + dez * dez)
4144 * ((aez * bceps - bez * aceps + cez * abeps)
4145 + (aeztail * bc3 - beztail * ac3 + ceztail * ab3)))
4146 - ((aex * aex + aey * aey + aez * aez)
4147 * ((bez * cdeps - cez * bdeps + dez * bceps)
4148 + (beztail * cd3 - ceztail * bd3 + deztail * bc3))
4149 + (cex * cex + cey * cey + cez * cez)
4150 * ((dez * abeps + aez * bdeps + bez * daeps)
4151 + (deztail * ab3 + aeztail * bd3 + beztail * da3))))
4152 + 2.0 * (((bex * bextail + bey * beytail + bez * beztail)
4153 * (cez * da3 + dez * ac3 + aez * cd3)
4154 + (dex * dextail + dey * deytail + dez * deztail)
4155 * (aez * bc3 - bez * ac3 + cez * ab3))
4156 - ((aex * aextail + aey * aeytail + aez * aeztail)
4157 * (bez * cd3 - cez * bd3 + dez * bc3)
4158 + (cex * cextail + cey * ceytail + cez * ceztail)
4159 * (dez * ab3 + aez * bd3 + bez * da3)));
4160 if ((det >= errbound) || (-det >= errbound)) {
4161 return det;
4162 }
4163
4164 return insphereexact(pa, pb, pc, pd, pe);
4165 }
4166
4167 REAL insphere(REAL *pa, REAL *pb, REAL *pc, REAL *pd, REAL *pe)
4168 {
4169 REAL aex, bex, cex, dex;
4170 REAL aey, bey, cey, dey;
4171 REAL aez, bez, cez, dez;
4172 REAL aexbey, bexaey, bexcey, cexbey, cexdey, dexcey, dexaey, aexdey;
4173 REAL aexcey, cexaey, bexdey, dexbey;
4174 REAL alift, blift, clift, dlift;
4175 REAL ab, bc, cd, da, ac, bd;
4176 REAL abc, bcd, cda, dab;
4177 REAL aezplus, bezplus, cezplus, dezplus;
4178 REAL aexbeyplus, bexaeyplus, bexceyplus, cexbeyplus;
4179 REAL cexdeyplus, dexceyplus, dexaeyplus, aexdeyplus;
4180 REAL aexceyplus, cexaeyplus, bexdeyplus, dexbeyplus;
4181 REAL det;
4182 REAL permanent, errbound;
4183
4184 aex = pa[0] - pe[0];
4185 bex = pb[0] - pe[0];
4186 cex = pc[0] - pe[0];
4187 dex = pd[0] - pe[0];
4188 aey = pa[1] - pe[1];
4189 bey = pb[1] - pe[1];
4190 cey = pc[1] - pe[1];
4191 dey = pd[1] - pe[1];
4192 aez = pa[2] - pe[2];
4193 bez = pb[2] - pe[2];
4194 cez = pc[2] - pe[2];
4195 dez = pd[2] - pe[2];
4196
4197 aexbey = aex * bey;
4198 bexaey = bex * aey;
4199 ab = aexbey - bexaey;
4200 bexcey = bex * cey;
4201 cexbey = cex * bey;
4202 bc = bexcey - cexbey;
4203 cexdey = cex * dey;
4204 dexcey = dex * cey;
4205 cd = cexdey - dexcey;
4206 dexaey = dex * aey;
4207 aexdey = aex * dey;
4208 da = dexaey - aexdey;
4209
4210 aexcey = aex * cey;
4211 cexaey = cex * aey;
4212 ac = aexcey - cexaey;
4213 bexdey = bex * dey;
4214 dexbey = dex * bey;
4215 bd = bexdey - dexbey;
4216
4217 abc = aez * bc - bez * ac + cez * ab;
4218 bcd = bez * cd - cez * bd + dez * bc;
4219 cda = cez * da + dez * ac + aez * cd;
4220 dab = dez * ab + aez * bd + bez * da;
4221
4222 alift = aex * aex + aey * aey + aez * aez;
4223 blift = bex * bex + bey * bey + bez * bez;
4224 clift = cex * cex + cey * cey + cez * cez;
4225 dlift = dex * dex + dey * dey + dez * dez;
4226
4227 det = (dlift * abc - clift * dab) + (blift * cda - alift * bcd);
4228
4229 aezplus = Absolute(aez);
4230 bezplus = Absolute(bez);
4231 cezplus = Absolute(cez);
4232 dezplus = Absolute(dez);
4233 aexbeyplus = Absolute(aexbey);
4234 bexaeyplus = Absolute(bexaey);
4235 bexceyplus = Absolute(bexcey);
4236 cexbeyplus = Absolute(cexbey);
4237 cexdeyplus = Absolute(cexdey);
4238 dexceyplus = Absolute(dexcey);
4239 dexaeyplus = Absolute(dexaey);
4240 aexdeyplus = Absolute(aexdey);
4241 aexceyplus = Absolute(aexcey);
4242 cexaeyplus = Absolute(cexaey);
4243 bexdeyplus = Absolute(bexdey);
4244 dexbeyplus = Absolute(dexbey);
4245 permanent = ((cexdeyplus + dexceyplus) * bezplus
4246 + (dexbeyplus + bexdeyplus) * cezplus
4247 + (bexceyplus + cexbeyplus) * dezplus)
4248 * alift
4249 + ((dexaeyplus + aexdeyplus) * cezplus
4250 + (aexceyplus + cexaeyplus) * dezplus
4251 + (cexdeyplus + dexceyplus) * aezplus)
4252 * blift
4253 + ((aexbeyplus + bexaeyplus) * dezplus
4254 + (bexdeyplus + dexbeyplus) * aezplus
4255 + (dexaeyplus + aexdeyplus) * bezplus)
4256 * clift
4257 + ((bexceyplus + cexbeyplus) * aezplus
4258 + (cexaeyplus + aexceyplus) * bezplus
4259 + (aexbeyplus + bexaeyplus) * cezplus)
4260 * dlift;
4261 errbound = isperrboundA * permanent;
4262 if ((det > errbound) || (-det > errbound)) {
4263 return det;
4264 }
4265
4266 return insphereadapt(pa, pb, pc, pd, pe, permanent);
4267 }
4268
4269 } // namespace robustPredicates