xref: /freebsd/contrib/bc/src/num.c (revision 7cc42f6d)
1 /*
2  * *****************************************************************************
3  *
4  * SPDX-License-Identifier: BSD-2-Clause
5  *
6  * Copyright (c) 2018-2020 Gavin D. Howard and contributors.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * * Redistributions of source code must retain the above copyright notice, this
12  *   list of conditions and the following disclaimer.
13  *
14  * * Redistributions in binary form must reproduce the above copyright notice,
15  *   this list of conditions and the following disclaimer in the documentation
16  *   and/or other materials provided with the distribution.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
19  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
22  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
23  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
24  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
25  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
26  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
27  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
28  * POSSIBILITY OF SUCH DAMAGE.
29  *
30  * *****************************************************************************
31  *
32  * Code for the number type.
33  *
34  */
35 
36 #include <assert.h>
37 #include <ctype.h>
38 #include <stdbool.h>
39 #include <stdlib.h>
40 #include <string.h>
41 #include <setjmp.h>
42 #include <limits.h>
43 
44 #include <status.h>
45 #include <num.h>
46 #include <rand.h>
47 #include <vm.h>
48 
49 static void bc_num_m(BcNum *a, BcNum *b, BcNum *restrict c, size_t scale);
50 
51 static inline ssize_t bc_num_neg(size_t n, bool neg) {
52 	return (((ssize_t) n) ^ -((ssize_t) neg)) + neg;
53 }
54 
55 ssize_t bc_num_cmpZero(const BcNum *n) {
56 	return bc_num_neg((n)->len != 0, (n)->neg);
57 }
58 
59 static inline size_t bc_num_int(const BcNum *n) {
60 	return n->len ? n->len - n->rdx : 0;
61 }
62 
63 static void bc_num_expand(BcNum *restrict n, size_t req) {
64 
65 	assert(n != NULL);
66 
67 	req = req >= BC_NUM_DEF_SIZE ? req : BC_NUM_DEF_SIZE;
68 
69 	if (req > n->cap) {
70 
71 		BC_SIG_LOCK;
72 
73 		n->num = bc_vm_realloc(n->num, BC_NUM_SIZE(req));
74 		n->cap = req;
75 
76 		BC_SIG_UNLOCK;
77 	}
78 }
79 
80 static void bc_num_setToZero(BcNum *restrict n, size_t scale) {
81 	assert(n != NULL);
82 	n->scale = scale;
83 	n->len = n->rdx = 0;
84 	n->neg = false;
85 }
86 
87 static inline void bc_num_zero(BcNum *restrict n) {
88 	bc_num_setToZero(n, 0);
89 }
90 
91 void bc_num_one(BcNum *restrict n) {
92 	bc_num_zero(n);
93 	n->len = 1;
94 	n->num[0] = 1;
95 }
96 
97 static void bc_num_clean(BcNum *restrict n) {
98 
99 	while (BC_NUM_NONZERO(n) && !n->num[n->len - 1]) n->len -= 1;
100 
101 	if (BC_NUM_ZERO(n)) {
102 		n->neg = false;
103 		n->rdx = 0;
104 	}
105 	else if (n->len < n->rdx) n->len = n->rdx;
106 }
107 
108 static size_t bc_num_log10(size_t i) {
109 	size_t len;
110 	for (len = 1; i; i /= BC_BASE, ++len);
111 	assert(len - 1 <= BC_BASE_DIGS + 1);
112 	return len - 1;
113 }
114 
115 static inline size_t bc_num_zeroDigits(const BcDig *n) {
116 	assert(*n >= 0);
117 	assert(((size_t) *n) < BC_BASE_POW);
118 	return BC_BASE_DIGS - bc_num_log10((size_t) *n);
119 }
120 
121 static size_t bc_num_intDigits(const BcNum *n) {
122 	size_t digits = bc_num_int(n) * BC_BASE_DIGS;
123 	if (digits > 0) digits -= bc_num_zeroDigits(n->num + n->len - 1);
124 	return digits;
125 }
126 
127 static size_t bc_num_nonzeroLen(const BcNum *restrict n) {
128 	size_t i, len = n->len;
129 	assert(len == n->rdx);
130 	for (i = len - 1; i < len && !n->num[i]; --i);
131 	assert(i + 1 > 0);
132 	return i + 1;
133 }
134 
135 static BcDig bc_num_addDigits(BcDig a, BcDig b, bool *carry) {
136 
137 	assert(((BcBigDig) BC_BASE_POW) * 2 == ((BcDig) BC_BASE_POW) * 2);
138 	assert(a < BC_BASE_POW);
139 	assert(b < BC_BASE_POW);
140 
141 	a += b + *carry;
142 	*carry = (a >= BC_BASE_POW);
143 	if (*carry) a -= BC_BASE_POW;
144 
145 	assert(a >= 0);
146 	assert(a < BC_BASE_POW);
147 
148 	return a;
149 }
150 
151 static BcDig bc_num_subDigits(BcDig a, BcDig b, bool *carry) {
152 
153 	assert(a < BC_BASE_POW);
154 	assert(b < BC_BASE_POW);
155 
156 	b += *carry;
157 	*carry = (a < b);
158 	if (*carry) a += BC_BASE_POW;
159 
160 	assert(a - b >= 0);
161 	assert(a - b < BC_BASE_POW);
162 
163 	return a - b;
164 }
165 
166 static void bc_num_addArrays(BcDig *restrict a, const BcDig *restrict b,
167                              size_t len)
168 {
169 	size_t i;
170 	bool carry = false;
171 
172 	for (i = 0; i < len; ++i) a[i] = bc_num_addDigits(a[i], b[i], &carry);
173 
174 	for (; carry; ++i) a[i] = bc_num_addDigits(a[i], 0, &carry);
175 }
176 
177 static void bc_num_subArrays(BcDig *restrict a, const BcDig *restrict b,
178                              size_t len)
179 {
180 	size_t i;
181 	bool carry = false;
182 
183 	for (i = 0; i < len; ++i) a[i] = bc_num_subDigits(a[i], b[i], &carry);
184 
185 	for (; carry; ++i) a[i] = bc_num_subDigits(a[i], 0, &carry);
186 }
187 
188 static void bc_num_mulArray(const BcNum *restrict a, BcBigDig b,
189                             BcNum *restrict c)
190 {
191 	size_t i;
192 	BcBigDig carry = 0;
193 
194 	assert(b <= BC_BASE_POW);
195 
196 	if (a->len + 1 > c->cap) bc_num_expand(c, a->len + 1);
197 
198 	memset(c->num, 0, BC_NUM_SIZE(c->cap));
199 
200 	for (i = 0; i < a->len; ++i) {
201 		BcBigDig in = ((BcBigDig) a->num[i]) * b + carry;
202 		c->num[i] = in % BC_BASE_POW;
203 		carry = in / BC_BASE_POW;
204 	}
205 
206 	assert(carry < BC_BASE_POW);
207 	c->num[i] = (BcDig) carry;
208 	c->len = a->len;
209 	c->len += (carry != 0);
210 
211 	bc_num_clean(c);
212 
213 	assert(!c->neg || BC_NUM_NONZERO(c));
214 	assert(c->rdx <= c->len || !c->len);
215 	assert(!c->len || c->num[c->len - 1] || c->rdx == c->len);
216 }
217 
218 static void bc_num_divArray(const BcNum *restrict a, BcBigDig b,
219                             BcNum *restrict c, BcBigDig *rem)
220 {
221 	size_t i;
222 	BcBigDig carry = 0;
223 
224 	assert(c->cap >= a->len);
225 
226 	for (i = a->len - 1; i < a->len; --i) {
227 		BcBigDig in = ((BcBigDig) a->num[i]) + carry * BC_BASE_POW;
228 		assert(in / b < BC_BASE_POW);
229 		c->num[i] = (BcDig) (in / b);
230 		carry = in % b;
231 	}
232 
233 	c->len = a->len;
234 	bc_num_clean(c);
235 	*rem = carry;
236 
237 	assert(!c->neg || BC_NUM_NONZERO(c));
238 	assert(c->rdx <= c->len || !c->len);
239 	assert(!c->len || c->num[c->len - 1] || c->rdx == c->len);
240 }
241 
242 static ssize_t bc_num_compare(const BcDig *restrict a, const BcDig *restrict b,
243                               size_t len)
244 {
245 	size_t i;
246 	BcDig c = 0;
247 	for (i = len - 1; i < len && !(c = a[i] - b[i]); --i);
248 	return bc_num_neg(i + 1, c < 0);
249 }
250 
251 ssize_t bc_num_cmp(const BcNum *a, const BcNum *b) {
252 
253 	size_t i, min, a_int, b_int, diff;
254 	BcDig *max_num, *min_num;
255 	bool a_max, neg = false;
256 	ssize_t cmp;
257 
258 	assert(a != NULL && b != NULL);
259 
260 	if (a == b) return 0;
261 	if (BC_NUM_ZERO(a)) return bc_num_neg(b->len != 0, !b->neg);
262 	if (BC_NUM_ZERO(b)) return bc_num_cmpZero(a);
263 	if (a->neg) {
264 		if (b->neg) neg = true;
265 		else return -1;
266 	}
267 	else if (b->neg) return 1;
268 
269 	a_int = bc_num_int(a);
270 	b_int = bc_num_int(b);
271 	a_int -= b_int;
272 
273 	if (a_int) return neg ? -((ssize_t) a_int) : (ssize_t) a_int;
274 
275 	a_max = (a->rdx > b->rdx);
276 
277 	if (a_max) {
278 		min = b->rdx;
279 		diff = a->rdx - b->rdx;
280 		max_num = a->num + diff;
281 		min_num = b->num;
282 	}
283 	else {
284 		min = a->rdx;
285 		diff = b->rdx - a->rdx;
286 		max_num = b->num + diff;
287 		min_num = a->num;
288 	}
289 
290 	cmp = bc_num_compare(max_num, min_num, b_int + min);
291 
292 	if (cmp) return bc_num_neg((size_t) cmp, !a_max == !neg);
293 
294 	for (max_num -= diff, i = diff - 1; i < diff; --i) {
295 		if (max_num[i]) return bc_num_neg(1, !a_max == !neg);
296 	}
297 
298 	return 0;
299 }
300 
301 void bc_num_truncate(BcNum *restrict n, size_t places) {
302 
303 	size_t places_rdx;
304 
305 	if (!places) return;
306 
307 	places_rdx = n->rdx ? n->rdx - BC_NUM_RDX(n->scale - places) : 0;
308 	assert(places <= n->scale && (BC_NUM_ZERO(n) || places_rdx <= n->len));
309 
310 	n->scale -= places;
311 	n->rdx -= places_rdx;
312 
313 	if (BC_NUM_NONZERO(n)) {
314 
315 		size_t pow;
316 
317 		pow = n->scale % BC_BASE_DIGS;
318 		pow = pow ? BC_BASE_DIGS - pow : 0;
319 		pow = bc_num_pow10[pow];
320 
321 		n->len -= places_rdx;
322 		memmove(n->num, n->num + places_rdx, BC_NUM_SIZE(n->len));
323 
324 		// Clear the lower part of the last digit.
325 		if (BC_NUM_NONZERO(n)) n->num[0] -= n->num[0] % (BcDig) pow;
326 
327 		bc_num_clean(n);
328 	}
329 }
330 
331 static void bc_num_extend(BcNum *restrict n, size_t places) {
332 
333 	size_t places_rdx;
334 
335 	if (!places) return;
336 	if (BC_NUM_ZERO(n)) {
337 		n->scale += places;
338 		return;
339 	}
340 
341 	places_rdx = BC_NUM_RDX(places + n->scale) - n->rdx;
342 
343 	if (places_rdx) {
344 		bc_num_expand(n, bc_vm_growSize(n->len, places_rdx));
345 		memmove(n->num + places_rdx, n->num, BC_NUM_SIZE(n->len));
346 		memset(n->num, 0, BC_NUM_SIZE(places_rdx));
347 	}
348 
349 	n->rdx += places_rdx;
350 	n->scale += places;
351 	n->len += places_rdx;
352 
353 	assert(n->rdx == BC_NUM_RDX(n->scale));
354 }
355 
356 static void bc_num_retireMul(BcNum *restrict n, size_t scale,
357                              bool neg1, bool neg2)
358 {
359 	if (n->scale < scale) bc_num_extend(n, scale - n->scale);
360 	else bc_num_truncate(n, n->scale - scale);
361 
362 	bc_num_clean(n);
363 	if (BC_NUM_NONZERO(n)) n->neg = (!neg1 != !neg2);
364 }
365 
366 static void bc_num_split(const BcNum *restrict n, size_t idx,
367                          BcNum *restrict a, BcNum *restrict b)
368 {
369 	assert(BC_NUM_ZERO(a));
370 	assert(BC_NUM_ZERO(b));
371 
372 	if (idx < n->len) {
373 
374 		b->len = n->len - idx;
375 		a->len = idx;
376 		a->scale = a->rdx = b->scale = b->rdx = 0;
377 
378 		assert(a->cap >= a->len);
379 		assert(b->cap >= b->len);
380 
381 		memcpy(b->num, n->num + idx, BC_NUM_SIZE(b->len));
382 		memcpy(a->num, n->num, BC_NUM_SIZE(idx));
383 
384 		bc_num_clean(b);
385 	}
386 	else bc_num_copy(a, n);
387 
388 	bc_num_clean(a);
389 }
390 
391 static size_t bc_num_shiftZero(BcNum *restrict n) {
392 
393 	size_t i;
394 
395 	assert(!n->rdx || BC_NUM_ZERO(n));
396 
397 	for (i = 0; i < n->len && !n->num[i]; ++i);
398 
399 	n->len -= i;
400 	n->num += i;
401 
402 	return i;
403 }
404 
405 static void bc_num_unshiftZero(BcNum *restrict n, size_t places_rdx) {
406 	n->len += places_rdx;
407 	n->num -= places_rdx;
408 }
409 
410 static void bc_num_shift(BcNum *restrict n, BcBigDig dig) {
411 
412 	size_t i, len = n->len;
413 	BcBigDig carry = 0, pow;
414 	BcDig *ptr = n->num;
415 
416 	assert(dig < BC_BASE_DIGS);
417 
418 	pow = bc_num_pow10[dig];
419 	dig = bc_num_pow10[BC_BASE_DIGS - dig];
420 
421 	for (i = len - 1; i < len; --i) {
422 		BcBigDig in, temp;
423 		in = ((BcBigDig) ptr[i]);
424 		temp = carry * dig;
425 		carry = in % pow;
426 		ptr[i] = ((BcDig) (in / pow)) + (BcDig) temp;
427 	}
428 
429 	assert(!carry);
430 }
431 
432 static void bc_num_shiftLeft(BcNum *restrict n, size_t places) {
433 
434 	BcBigDig dig;
435 	size_t places_rdx;
436 	bool shift;
437 
438 	if (!places) return;
439 	if (places > n->scale) {
440 		size_t size = bc_vm_growSize(BC_NUM_RDX(places - n->scale), n->len);
441 		if (size > SIZE_MAX - 1) bc_vm_err(BC_ERROR_MATH_OVERFLOW);
442 	}
443 	if (BC_NUM_ZERO(n)) {
444 		if (n->scale >= places) n->scale -= places;
445 		else n->scale = 0;
446 		return;
447 	}
448 
449 	dig = (BcBigDig) (places % BC_BASE_DIGS);
450 	shift = (dig != 0);
451 	places_rdx = BC_NUM_RDX(places);
452 
453 	if (n->scale) {
454 
455 		if (n->rdx >= places_rdx) {
456 
457 			size_t mod = n->scale % BC_BASE_DIGS, revdig;
458 
459 			mod = mod ? mod : BC_BASE_DIGS;
460 			revdig = dig ? BC_BASE_DIGS - dig : 0;
461 
462 			if (mod + revdig > BC_BASE_DIGS) places_rdx = 1;
463 			else places_rdx = 0;
464 		}
465 		else places_rdx -= n->rdx;
466 	}
467 
468 	if (places_rdx) {
469 		bc_num_expand(n, bc_vm_growSize(n->len, places_rdx));
470 		memmove(n->num + places_rdx, n->num, BC_NUM_SIZE(n->len));
471 		memset(n->num, 0, BC_NUM_SIZE(places_rdx));
472 		n->len += places_rdx;
473 	}
474 
475 	if (places > n->scale) n->scale = n->rdx = 0;
476 	else {
477 		n->scale -= places;
478 		n->rdx = BC_NUM_RDX(n->scale);
479 	}
480 
481 	if (shift) bc_num_shift(n, BC_BASE_DIGS - dig);
482 
483 	bc_num_clean(n);
484 }
485 
486 static void bc_num_shiftRight(BcNum *restrict n, size_t places) {
487 
488 	BcBigDig dig;
489 	size_t places_rdx, scale, scale_mod, int_len, expand;
490 	bool shift;
491 
492 	if (!places) return;
493 	if (BC_NUM_ZERO(n)) {
494 		n->scale += places;
495 		bc_num_expand(n, BC_NUM_RDX(n->scale));
496 		return;
497 	}
498 
499 	dig = (BcBigDig) (places % BC_BASE_DIGS);
500 	shift = (dig != 0);
501 	scale = n->scale;
502 	scale_mod = scale % BC_BASE_DIGS;
503 	scale_mod = scale_mod ? scale_mod : BC_BASE_DIGS;
504 	int_len = bc_num_int(n);
505 	places_rdx = BC_NUM_RDX(places);
506 
507 	if (scale_mod + dig > BC_BASE_DIGS) {
508 		expand = places_rdx - 1;
509 		places_rdx = 1;
510 	}
511 	else {
512 		expand = places_rdx;
513 		places_rdx = 0;
514 	}
515 
516 	if (expand > int_len) expand -= int_len;
517 	else expand = 0;
518 
519 	bc_num_extend(n, places_rdx * BC_BASE_DIGS);
520 	bc_num_expand(n, bc_vm_growSize(expand, n->len));
521 	memset(n->num + n->len, 0, BC_NUM_SIZE(expand));
522 	n->len += expand;
523 	n->scale = n->rdx = 0;
524 
525 	if (shift) bc_num_shift(n, dig);
526 
527 	n->scale = scale + places;
528 	n->rdx = BC_NUM_RDX(n->scale);
529 
530 	bc_num_clean(n);
531 
532 	assert(n->rdx <= n->len && n->len <= n->cap);
533 	assert(n->rdx == BC_NUM_RDX(n->scale));
534 }
535 
536 static void bc_num_inv(BcNum *a, BcNum *b, size_t scale) {
537 
538 	BcNum one;
539 	BcDig num[2];
540 
541 	assert(BC_NUM_NONZERO(a));
542 
543 	bc_num_setup(&one, num, sizeof(num) / sizeof(BcDig));
544 	bc_num_one(&one);
545 
546 	bc_num_div(&one, a, b, scale);
547 }
548 
549 #if BC_ENABLE_EXTRA_MATH
550 static void bc_num_intop(const BcNum *a, const BcNum *b, BcNum *restrict c,
551                          BcBigDig *v)
552 {
553 	if (BC_ERR(b->rdx)) bc_vm_err(BC_ERROR_MATH_NON_INTEGER);
554 	bc_num_copy(c, a);
555 	bc_num_bigdig(b, v);
556 }
557 #endif // BC_ENABLE_EXTRA_MATH
558 
559 static void bc_num_as(BcNum *a, BcNum *b, BcNum *restrict c, size_t sub) {
560 
561 	BcDig *ptr_c, *ptr_l, *ptr_r;
562 	size_t i, min_rdx, max_rdx, diff, a_int, b_int, min_len, max_len, max_int;
563 	size_t len_l, len_r;
564 	bool b_neg, do_sub, do_rev_sub, carry;
565 
566 	// Because this function doesn't need to use scale (per the bc spec),
567 	// I am hijacking it to say whether it's doing an add or a subtract.
568 	// Convert substraction to addition of negative second operand.
569 
570 	if (BC_NUM_ZERO(b)) {
571 		bc_num_copy(c, a);
572 		return;
573 	}
574 	if (BC_NUM_ZERO(a)) {
575 		bc_num_copy(c, b);
576 		c->neg = (b->neg != sub);
577 		return;
578 	}
579 
580 	// Invert sign of b if it is to be subtracted. This operation must
581 	// preced the tests for any of the operands being zero.
582 	b_neg = (b->neg != sub);
583 
584 	// Actually add the numbers if their signs are equal, else subtract.
585 	do_sub = (a->neg != b_neg);
586 
587 	a_int = bc_num_int(a);
588 	b_int = bc_num_int(b);
589 	max_int = BC_MAX(a_int, b_int);
590 
591 	min_rdx = BC_MIN(a->rdx, b->rdx);
592 	max_rdx = BC_MAX(a->rdx, b->rdx);
593 	diff = max_rdx - min_rdx;
594 
595 	max_len = max_int + max_rdx;
596 
597 	if (do_sub) {
598 
599 		// Check whether b has to be subtracted from a or a from b.
600 		if (a_int != b_int) do_rev_sub = (a_int < b_int);
601 		else if (a->rdx > b->rdx)
602 			do_rev_sub = (bc_num_compare(a->num + diff, b->num, b->len) < 0);
603 		else
604 			do_rev_sub = (bc_num_compare(a->num, b->num + diff, a->len) <= 0);
605 	}
606 	else {
607 
608 		// The result array of the addition might come out one element
609 		// longer than the bigger of the operand arrays.
610 		max_len += 1;
611 		do_rev_sub = (a_int < b_int);
612 	}
613 
614 	assert(max_len <= c->cap);
615 
616 	if (do_rev_sub) {
617 		ptr_l = b->num;
618 		ptr_r = a->num;
619 		len_l = b->len;
620 		len_r = a->len;
621 	}
622 	else {
623 		ptr_l = a->num;
624 		ptr_r = b->num;
625 		len_l = a->len;
626 		len_r = b->len;
627 	}
628 
629 	ptr_c = c->num;
630 	carry = false;
631 
632 	if (diff) {
633 
634 		// If the rdx values of the operands do not match, the result will
635 		// have low end elements that are the positive or negative trailing
636 		// elements of the operand with higher rdx value.
637 		if ((a->rdx > b->rdx) != do_rev_sub) {
638 
639 			// !do_rev_sub && a->rdx > b->rdx || do_rev_sub && b->rdx > a->rdx
640 			// The left operand has BcDig values that need to be copied,
641 			// either from a or from b (in case of a reversed subtraction).
642 			memcpy(ptr_c, ptr_l, BC_NUM_SIZE(diff));
643 			ptr_l += diff;
644 			len_l -= diff;
645 		}
646 		else {
647 
648 			// The right operand has BcDig values that need to be copied
649 			// or subtracted from zero (in case of a subtraction).
650 			if (do_sub) {
651 
652 				// do_sub (do_rev_sub && a->rdx > b->rdx ||
653 				// !do_rev_sub && b->rdx > a->rdx)
654 				for (i = 0; i < diff; i++)
655 					ptr_c[i] = bc_num_subDigits(0, ptr_r[i], &carry);
656 			}
657 			else {
658 
659 				// !do_sub && b->rdx > a->rdx
660 				memcpy(ptr_c, ptr_r, BC_NUM_SIZE(diff));
661 			}
662 
663 			ptr_r += diff;
664 			len_r -= diff;
665 		}
666 
667 		ptr_c += diff;
668 	}
669 
670 	min_len = BC_MIN(len_l, len_r);
671 
672 	// After dealing with possible low array elements that depend on only one
673 	// operand, the actual add or subtract can be performed as if the rdx of
674 	// both operands was the same.
675 	// Inlining takes care of eliminating constant zero arguments to
676 	// addDigit/subDigit (checked in disassembly of resulting bc binary
677 	// compiled with gcc and clang).
678 	if (do_sub) {
679 		for (i = 0; i < min_len; ++i)
680 			ptr_c[i] = bc_num_subDigits(ptr_l[i], ptr_r[i], &carry);
681 		for (; i < len_l; ++i) ptr_c[i] = bc_num_subDigits(ptr_l[i], 0, &carry);
682 	}
683 	else {
684 		for (i = 0; i < min_len; ++i)
685 			ptr_c[i] = bc_num_addDigits(ptr_l[i], ptr_r[i], &carry);
686 		for (; i < len_l; ++i) ptr_c[i] = bc_num_addDigits(ptr_l[i], 0, &carry);
687 		ptr_c[i] = bc_num_addDigits(0, 0, &carry);
688 	}
689 
690 	assert(carry == false);
691 
692 	// The result has the same sign as a, unless the operation was a
693 	// reverse subtraction (b - a).
694 	c->neg = (a->neg != (do_sub && do_rev_sub));
695 	c->len = max_len;
696 	c->rdx = max_rdx;
697 	c->scale = BC_MAX(a->scale, b->scale);
698 
699 	bc_num_clean(c);
700 }
701 
702 static void bc_num_m_simp(const BcNum *a, const BcNum *b, BcNum *restrict c)
703 {
704 	size_t i, alen = a->len, blen = b->len, clen;
705 	BcDig *ptr_a = a->num, *ptr_b = b->num, *ptr_c;
706 	BcBigDig sum = 0, carry = 0;
707 
708 	assert(sizeof(sum) >= sizeof(BcDig) * 2);
709 	assert(!a->rdx && !b->rdx);
710 
711 	clen = bc_vm_growSize(alen, blen);
712 	bc_num_expand(c, bc_vm_growSize(clen, 1));
713 
714 	ptr_c = c->num;
715 	memset(ptr_c, 0, BC_NUM_SIZE(c->cap));
716 
717 	for (i = 0; i < clen; ++i) {
718 
719 		ssize_t sidx = (ssize_t) (i - blen + 1);
720 		size_t j = (size_t) BC_MAX(0, sidx), k = BC_MIN(i, blen - 1);
721 
722 		for (; j < alen && k < blen; ++j, --k) {
723 
724 			sum += ((BcBigDig) ptr_a[j]) * ((BcBigDig) ptr_b[k]);
725 
726 			if (sum >= ((BcBigDig) BC_BASE_POW) * BC_BASE_POW) {
727 				carry += sum / BC_BASE_POW;
728 				sum %= BC_BASE_POW;
729 			}
730 		}
731 
732 		if (sum >= BC_BASE_POW) {
733 			carry += sum / BC_BASE_POW;
734 			sum %= BC_BASE_POW;
735 		}
736 
737 		ptr_c[i] = (BcDig) sum;
738 		assert(ptr_c[i] < BC_BASE_POW);
739 		sum = carry;
740 		carry = 0;
741 	}
742 
743 	// This should always be true because there should be no carry on the last
744 	// digit; multiplication never goes above the sum of both lengths.
745 	assert(!sum);
746 
747 	c->len = clen;
748 }
749 
750 static void bc_num_shiftAddSub(BcNum *restrict n, const BcNum *restrict a,
751                                size_t shift, BcNumShiftAddOp op)
752 {
753 	assert(n->len >= shift + a->len);
754 	assert(!n->rdx && !a->rdx);
755 	op(n->num + shift, a->num, a->len);
756 }
757 
758 static void bc_num_k(BcNum *a, BcNum *b, BcNum *restrict c) {
759 
760 	size_t max, max2, total;
761 	BcNum l1, h1, l2, h2, m2, m1, z0, z1, z2, temp;
762 	BcDig *digs, *dig_ptr;
763 	BcNumShiftAddOp op;
764 	bool aone = BC_NUM_ONE(a);
765 
766 	assert(BC_NUM_ZERO(c));
767 
768 	if (BC_NUM_ZERO(a) || BC_NUM_ZERO(b)) return;
769 	if (aone || BC_NUM_ONE(b)) {
770 		bc_num_copy(c, aone ? b : a);
771 		if ((aone && a->neg) || b->neg) c->neg = !c->neg;
772 		return;
773 	}
774 	if (a->len < BC_NUM_KARATSUBA_LEN || b->len < BC_NUM_KARATSUBA_LEN) {
775 		bc_num_m_simp(a, b, c);
776 		return;
777 	}
778 
779 	max = BC_MAX(a->len, b->len);
780 	max = BC_MAX(max, BC_NUM_DEF_SIZE);
781 	max2 = (max + 1) / 2;
782 
783 	total = bc_vm_arraySize(BC_NUM_KARATSUBA_ALLOCS, max);
784 
785 	BC_SIG_LOCK;
786 
787 	digs = dig_ptr = bc_vm_malloc(BC_NUM_SIZE(total));
788 
789 	bc_num_setup(&l1, dig_ptr, max);
790 	dig_ptr += max;
791 	bc_num_setup(&h1, dig_ptr, max);
792 	dig_ptr += max;
793 	bc_num_setup(&l2, dig_ptr, max);
794 	dig_ptr += max;
795 	bc_num_setup(&h2, dig_ptr, max);
796 	dig_ptr += max;
797 	bc_num_setup(&m1, dig_ptr, max);
798 	dig_ptr += max;
799 	bc_num_setup(&m2, dig_ptr, max);
800 	max = bc_vm_growSize(max, 1);
801 	bc_num_init(&z0, max);
802 	bc_num_init(&z1, max);
803 	bc_num_init(&z2, max);
804 	max = bc_vm_growSize(max, max) + 1;
805 	bc_num_init(&temp, max);
806 
807 	BC_SETJMP_LOCKED(err);
808 
809 	BC_SIG_UNLOCK;
810 
811 	bc_num_split(a, max2, &l1, &h1);
812 	bc_num_split(b, max2, &l2, &h2);
813 
814 	bc_num_expand(c, max);
815 	c->len = max;
816 	memset(c->num, 0, BC_NUM_SIZE(c->len));
817 
818 	bc_num_sub(&h1, &l1, &m1, 0);
819 	bc_num_sub(&l2, &h2, &m2, 0);
820 
821 	if (BC_NUM_NONZERO(&h1) && BC_NUM_NONZERO(&h2)) {
822 
823 		bc_num_m(&h1, &h2, &z2, 0);
824 		bc_num_clean(&z2);
825 
826 		bc_num_shiftAddSub(c, &z2, max2 * 2, bc_num_addArrays);
827 		bc_num_shiftAddSub(c, &z2, max2, bc_num_addArrays);
828 	}
829 
830 	if (BC_NUM_NONZERO(&l1) && BC_NUM_NONZERO(&l2)) {
831 
832 		bc_num_m(&l1, &l2, &z0, 0);
833 		bc_num_clean(&z0);
834 
835 		bc_num_shiftAddSub(c, &z0, max2, bc_num_addArrays);
836 		bc_num_shiftAddSub(c, &z0, 0, bc_num_addArrays);
837 	}
838 
839 	if (BC_NUM_NONZERO(&m1) && BC_NUM_NONZERO(&m2)) {
840 
841 		bc_num_m(&m1, &m2, &z1, 0);
842 		bc_num_clean(&z1);
843 
844 		op = (m1.neg != m2.neg) ? bc_num_subArrays : bc_num_addArrays;
845 		bc_num_shiftAddSub(c, &z1, max2, op);
846 	}
847 
848 err:
849 	BC_SIG_MAYLOCK;
850 	free(digs);
851 	bc_num_free(&temp);
852 	bc_num_free(&z2);
853 	bc_num_free(&z1);
854 	bc_num_free(&z0);
855 	BC_LONGJMP_CONT;
856 }
857 
858 static void bc_num_m(BcNum *a, BcNum *b, BcNum *restrict c, size_t scale) {
859 
860 	BcNum cpa, cpb;
861 	size_t ascale, bscale, ardx, brdx, azero = 0, bzero = 0, zero, len, rscale;
862 
863 	bc_num_zero(c);
864 	ascale = a->scale;
865 	bscale = b->scale;
866 	scale = BC_MAX(scale, ascale);
867 	scale = BC_MAX(scale, bscale);
868 
869 	rscale = ascale + bscale;
870 	scale = BC_MIN(rscale, scale);
871 
872 	if ((a->len == 1 || b->len == 1) && !a->rdx && !b->rdx) {
873 
874 		BcNum *operand;
875 		BcBigDig dig;
876 
877 		if (a->len == 1) {
878 			dig = (BcBigDig) a->num[0];
879 			operand = b;
880 		}
881 		else {
882 			dig = (BcBigDig) b->num[0];
883 			operand = a;
884 		}
885 
886 		bc_num_mulArray(operand, dig, c);
887 
888 		if (BC_NUM_NONZERO(c)) c->neg = (a->neg != b->neg);
889 
890 		return;
891 	}
892 
893 	BC_SIG_LOCK;
894 
895 	bc_num_init(&cpa, a->len + a->rdx);
896 	bc_num_init(&cpb, b->len + b->rdx);
897 
898 	BC_SETJMP_LOCKED(err);
899 
900 	BC_SIG_UNLOCK;
901 
902 	bc_num_copy(&cpa, a);
903 	bc_num_copy(&cpb, b);
904 
905 	cpa.neg = cpb.neg = false;
906 
907 	ardx = cpa.rdx * BC_BASE_DIGS;
908 	bc_num_shiftLeft(&cpa, ardx);
909 
910 	brdx = cpb.rdx * BC_BASE_DIGS;
911 	bc_num_shiftLeft(&cpb, brdx);
912 
913 	// We need to reset the jump here because azero and bzero are used in the
914 	// cleanup, and local variables are not guaranteed to be the same after a
915 	// jump.
916 	BC_SIG_LOCK;
917 
918 	BC_UNSETJMP;
919 
920 	azero = bc_num_shiftZero(&cpa);
921 	bzero = bc_num_shiftZero(&cpb);
922 
923 	BC_SETJMP_LOCKED(err);
924 
925 	BC_SIG_UNLOCK;
926 
927 	bc_num_clean(&cpa);
928 	bc_num_clean(&cpb);
929 
930 	bc_num_k(&cpa, &cpb, c);
931 
932 	zero = bc_vm_growSize(azero, bzero);
933 	len = bc_vm_growSize(c->len, zero);
934 
935 	bc_num_expand(c, len);
936 	bc_num_shiftLeft(c, (len - c->len) * BC_BASE_DIGS);
937 	bc_num_shiftRight(c, ardx + brdx);
938 
939 	bc_num_retireMul(c, scale, a->neg, b->neg);
940 
941 err:
942 	BC_SIG_MAYLOCK;
943 	bc_num_unshiftZero(&cpb, bzero);
944 	bc_num_free(&cpb);
945 	bc_num_unshiftZero(&cpa, azero);
946 	bc_num_free(&cpa);
947 	BC_LONGJMP_CONT;
948 }
949 
950 static bool bc_num_nonZeroDig(BcDig *restrict a, size_t len) {
951 	size_t i;
952 	bool nonzero = false;
953 	for (i = len - 1; !nonzero && i < len; --i) nonzero = (a[i] != 0);
954 	return nonzero;
955 }
956 
957 static ssize_t bc_num_divCmp(const BcDig *a, const BcNum *b, size_t len) {
958 
959 	ssize_t cmp;
960 
961 	if (b->len > len && a[len]) cmp = bc_num_compare(a, b->num, len + 1);
962 	else if (b->len <= len) {
963 		if (a[len]) cmp = 1;
964 		else cmp = bc_num_compare(a, b->num, len);
965 	}
966 	else cmp = -1;
967 
968 	return cmp;
969 }
970 
971 static void bc_num_divExtend(BcNum *restrict a, BcNum *restrict b,
972                              BcBigDig divisor)
973 {
974 	size_t pow;
975 
976 	assert(divisor < BC_BASE_POW);
977 
978 	pow = BC_BASE_DIGS - bc_num_log10((size_t) divisor);
979 
980 	bc_num_shiftLeft(a, pow);
981 	bc_num_shiftLeft(b, pow);
982 }
983 
984 static void bc_num_d_long(BcNum *restrict a, BcNum *restrict b,
985                           BcNum *restrict c, size_t scale)
986 {
987 	BcBigDig divisor;
988 	size_t len, end, i, rdx;
989 	BcNum cpb;
990 	bool nonzero = false;
991 
992 	assert(b->len < a->len);
993 	len = b->len;
994 	end = a->len - len;
995 	assert(len >= 1);
996 
997 	bc_num_expand(c, a->len);
998 	memset(c->num, 0, c->cap * sizeof(BcDig));
999 
1000 	c->rdx = a->rdx;
1001 	c->scale = a->scale;
1002 	c->len = a->len;
1003 
1004 	divisor = (BcBigDig) b->num[len - 1];
1005 
1006 	if (len > 1 && bc_num_nonZeroDig(b->num, len - 1)) {
1007 
1008 		nonzero = (divisor > 1 << ((10 * BC_BASE_DIGS) / 6 + 1));
1009 
1010 		if (!nonzero) {
1011 
1012 			bc_num_divExtend(a, b, divisor);
1013 
1014 			len = BC_MAX(a->len, b->len);
1015 			bc_num_expand(a, len + 1);
1016 
1017 			if (len + 1 > a->len) a->len = len + 1;
1018 
1019 			len = b->len;
1020 			end = a->len - len;
1021 			divisor = (BcBigDig) b->num[len - 1];
1022 
1023 			nonzero = bc_num_nonZeroDig(b->num, len - 1);
1024 		}
1025 	}
1026 
1027 	divisor += nonzero;
1028 
1029 	bc_num_expand(c, a->len);
1030 	memset(c->num, 0, BC_NUM_SIZE(c->cap));
1031 
1032 	assert(c->scale >= scale);
1033 	rdx = c->rdx - BC_NUM_RDX(scale);
1034 
1035 	BC_SIG_LOCK;
1036 
1037 	bc_num_init(&cpb, len + 1);
1038 
1039 	BC_SETJMP_LOCKED(err);
1040 
1041 	BC_SIG_UNLOCK;
1042 
1043 	i = end - 1;
1044 
1045 	for (; i < end && i >= rdx && BC_NUM_NONZERO(a); --i) {
1046 
1047 		ssize_t cmp;
1048 		BcDig *n;
1049 		BcBigDig result;
1050 
1051 		n = a->num + i;
1052 		assert(n >= a->num);
1053 		result = 0;
1054 
1055 		cmp = bc_num_divCmp(n, b, len);
1056 
1057 		while (cmp >= 0) {
1058 
1059 			BcBigDig n1, dividend, q;
1060 
1061 			n1 = (BcBigDig) n[len];
1062 			dividend = n1 * BC_BASE_POW + (BcBigDig) n[len - 1];
1063 			q = (dividend / divisor);
1064 
1065 			if (q <= 1) {
1066 				q = 1;
1067 				bc_num_subArrays(n, b->num, len);
1068 			}
1069 			else {
1070 
1071 				assert(q <= BC_BASE_POW);
1072 
1073 				bc_num_mulArray(b, (BcBigDig) q, &cpb);
1074 				bc_num_subArrays(n, cpb.num, cpb.len);
1075 			}
1076 
1077 			result += q;
1078 			assert(result <= BC_BASE_POW);
1079 
1080 			if (nonzero) cmp = bc_num_divCmp(n, b, len);
1081 			else cmp = -1;
1082 		}
1083 
1084 		assert(result < BC_BASE_POW);
1085 
1086 		c->num[i] = (BcDig) result;
1087 	}
1088 
1089 err:
1090 	BC_SIG_MAYLOCK;
1091 	bc_num_free(&cpb);
1092 	BC_LONGJMP_CONT;
1093 }
1094 
1095 static void bc_num_d(BcNum *a, BcNum *b, BcNum *restrict c, size_t scale) {
1096 
1097 	size_t len;
1098 	BcNum cpa, cpb;
1099 
1100 	if (BC_NUM_ZERO(b)) bc_vm_err(BC_ERROR_MATH_DIVIDE_BY_ZERO);
1101 	if (BC_NUM_ZERO(a)) {
1102 		bc_num_setToZero(c, scale);
1103 		return;
1104 	}
1105 	if (BC_NUM_ONE(b)) {
1106 		bc_num_copy(c, a);
1107 		bc_num_retireMul(c, scale, a->neg, b->neg);
1108 		return;
1109 	}
1110 	if (!a->rdx && !b->rdx && b->len == 1 && !scale) {
1111 		BcBigDig rem;
1112 		bc_num_divArray(a, (BcBigDig) b->num[0], c, &rem);
1113 		bc_num_retireMul(c, scale, a->neg, b->neg);
1114 		return;
1115 	}
1116 
1117 	len = bc_num_mulReq(a, b, scale);
1118 
1119 	BC_SIG_LOCK;
1120 
1121 	bc_num_init(&cpa, len);
1122 	bc_num_copy(&cpa, a);
1123 	bc_num_createCopy(&cpb, b);
1124 
1125 	BC_SETJMP_LOCKED(err);
1126 
1127 	BC_SIG_UNLOCK;
1128 
1129 	len = b->len;
1130 
1131 	if (len > cpa.len) {
1132 		bc_num_expand(&cpa, bc_vm_growSize(len, 2));
1133 		bc_num_extend(&cpa, (len - cpa.len) * BC_BASE_DIGS);
1134 	}
1135 
1136 	cpa.scale = cpa.rdx * BC_BASE_DIGS;
1137 
1138 	bc_num_extend(&cpa, b->scale);
1139 	cpa.rdx -= BC_NUM_RDX(b->scale);
1140 	cpa.scale = cpa.rdx * BC_BASE_DIGS;
1141 
1142 	if (scale > cpa.scale) {
1143 		bc_num_extend(&cpa, scale);
1144 		cpa.scale = cpa.rdx * BC_BASE_DIGS;
1145 	}
1146 
1147 	if (cpa.cap == cpa.len) bc_num_expand(&cpa, bc_vm_growSize(cpa.len, 1));
1148 
1149 	// We want an extra zero in front to make things simpler.
1150 	cpa.num[cpa.len++] = 0;
1151 
1152 	if (cpa.rdx == cpa.len) cpa.len = bc_num_nonzeroLen(&cpa);
1153 	if (cpb.rdx == cpb.len) cpb.len = bc_num_nonzeroLen(&cpb);
1154 	cpb.scale = cpb.rdx = 0;
1155 
1156 	bc_num_d_long(&cpa, &cpb, c, scale);
1157 
1158 	bc_num_retireMul(c, scale, a->neg, b->neg);
1159 
1160 err:
1161 	BC_SIG_MAYLOCK;
1162 	bc_num_free(&cpb);
1163 	bc_num_free(&cpa);
1164 	BC_LONGJMP_CONT;
1165 }
1166 
1167 static void bc_num_r(BcNum *a, BcNum *b, BcNum *restrict c,
1168                      BcNum *restrict d, size_t scale, size_t ts)
1169 {
1170 	BcNum temp;
1171 	bool neg;
1172 
1173 	if (BC_NUM_ZERO(b)) bc_vm_err(BC_ERROR_MATH_DIVIDE_BY_ZERO);
1174 	if (BC_NUM_ZERO(a)) {
1175 		bc_num_setToZero(c, ts);
1176 		bc_num_setToZero(d, ts);
1177 		return;
1178 	}
1179 
1180 	BC_SIG_LOCK;
1181 
1182 	bc_num_init(&temp, d->cap);
1183 
1184 	BC_SETJMP_LOCKED(err);
1185 
1186 	BC_SIG_UNLOCK;
1187 
1188 	bc_num_d(a, b, c, scale);
1189 
1190 	if (scale) scale = ts + 1;
1191 
1192 	bc_num_m(c, b, &temp, scale);
1193 	bc_num_sub(a, &temp, d, scale);
1194 
1195 	if (ts > d->scale && BC_NUM_NONZERO(d)) bc_num_extend(d, ts - d->scale);
1196 
1197 	neg = d->neg;
1198 	bc_num_retireMul(d, ts, a->neg, b->neg);
1199 	d->neg = BC_NUM_NONZERO(d) ? neg : false;
1200 
1201 err:
1202 	BC_SIG_MAYLOCK;
1203 	bc_num_free(&temp);
1204 	BC_LONGJMP_CONT;
1205 }
1206 
1207 static void bc_num_rem(BcNum *a, BcNum *b, BcNum *restrict c, size_t scale) {
1208 
1209 	BcNum c1;
1210 	size_t ts;
1211 
1212 	ts = bc_vm_growSize(scale, b->scale);
1213 	ts = BC_MAX(ts, a->scale);
1214 
1215 	BC_SIG_LOCK;
1216 
1217 	bc_num_init(&c1, bc_num_mulReq(a, b, ts));
1218 
1219 	BC_SETJMP_LOCKED(err);
1220 
1221 	BC_SIG_UNLOCK;
1222 
1223 	bc_num_r(a, b, &c1, c, scale, ts);
1224 
1225 err:
1226 	BC_SIG_MAYLOCK;
1227 	bc_num_free(&c1);
1228 	BC_LONGJMP_CONT;
1229 }
1230 
1231 static void bc_num_p(BcNum *a, BcNum *b, BcNum *restrict c, size_t scale) {
1232 
1233 	BcNum copy;
1234 	BcBigDig pow = 0;
1235 	size_t i, powrdx, resrdx;
1236 	bool neg, zero;
1237 
1238 	if (BC_ERR(b->rdx)) bc_vm_err(BC_ERROR_MATH_NON_INTEGER);
1239 
1240 	if (BC_NUM_ZERO(b)) {
1241 		bc_num_one(c);
1242 		return;
1243 	}
1244 	if (BC_NUM_ZERO(a)) {
1245 		if (b->neg) bc_vm_err(BC_ERROR_MATH_DIVIDE_BY_ZERO);
1246 		bc_num_setToZero(c, scale);
1247 		return;
1248 	}
1249 	if (BC_NUM_ONE(b)) {
1250 		if (!b->neg) bc_num_copy(c, a);
1251 		else bc_num_inv(a, c, scale);
1252 		return;
1253 	}
1254 
1255 	BC_SIG_LOCK;
1256 
1257 	neg = b->neg;
1258 	b->neg = false;
1259 	bc_num_bigdig(b, &pow);
1260 	b->neg = neg;
1261 
1262 	bc_num_createCopy(&copy, a);
1263 
1264 	BC_SETJMP_LOCKED(err);
1265 
1266 	BC_SIG_UNLOCK;
1267 
1268 	if (!neg) {
1269 		size_t max = BC_MAX(scale, a->scale), scalepow = a->scale * pow;
1270 		scale = BC_MIN(scalepow, max);
1271 	}
1272 
1273 	for (powrdx = a->scale; !(pow & 1); pow >>= 1) {
1274 		powrdx <<= 1;
1275 		bc_num_mul(&copy, &copy, &copy, powrdx);
1276 	}
1277 
1278 	bc_num_copy(c, &copy);
1279 	resrdx = powrdx;
1280 
1281 	while (pow >>= 1) {
1282 
1283 		powrdx <<= 1;
1284 		bc_num_mul(&copy, &copy, &copy, powrdx);
1285 
1286 		if (pow & 1) {
1287 			resrdx += powrdx;
1288 			bc_num_mul(c, &copy, c, resrdx);
1289 		}
1290 	}
1291 
1292 	if (neg) bc_num_inv(c, c, scale);
1293 
1294 	if (c->scale > scale) bc_num_truncate(c, c->scale - scale);
1295 
1296 	// We can't use bc_num_clean() here.
1297 	for (zero = true, i = 0; zero && i < c->len; ++i) zero = !c->num[i];
1298 	if (zero) bc_num_setToZero(c, scale);
1299 
1300 err:
1301 	BC_SIG_MAYLOCK;
1302 	bc_num_free(&copy);
1303 	BC_LONGJMP_CONT;
1304 }
1305 
1306 #if BC_ENABLE_EXTRA_MATH
1307 static void bc_num_place(BcNum *a, BcNum *b, BcNum *restrict c, size_t scale) {
1308 
1309 	BcBigDig val = 0;
1310 
1311 	BC_UNUSED(scale);
1312 
1313 	bc_num_intop(a, b, c, &val);
1314 
1315 	if (val < c->scale) bc_num_truncate(c, c->scale - val);
1316 	else if (val > c->scale) bc_num_extend(c, val - c->scale);
1317 }
1318 
1319 static void bc_num_left(BcNum *a, BcNum *b, BcNum *restrict c, size_t scale) {
1320 
1321 	BcBigDig val = 0;
1322 
1323 	BC_UNUSED(scale);
1324 
1325 	bc_num_intop(a, b, c, &val);
1326 
1327 	bc_num_shiftLeft(c, (size_t) val);
1328 }
1329 
1330 static void bc_num_right(BcNum *a, BcNum *b, BcNum *restrict c, size_t scale) {
1331 
1332 	BcBigDig val = 0;
1333 
1334 	BC_UNUSED(scale);
1335 
1336 	bc_num_intop(a, b, c, &val);
1337 
1338 	if (BC_NUM_ZERO(c)) return;
1339 
1340 	bc_num_shiftRight(c, (size_t) val);
1341 }
1342 #endif // BC_ENABLE_EXTRA_MATH
1343 
1344 static void bc_num_binary(BcNum *a, BcNum *b, BcNum *c, size_t scale,
1345                           BcNumBinaryOp op, size_t req)
1346 {
1347 	BcNum num2, *ptr_a, *ptr_b;
1348 	bool init = false;
1349 
1350 	assert(a != NULL && b != NULL && c != NULL && op != NULL);
1351 
1352 	BC_SIG_LOCK;
1353 
1354 	if (c == a) {
1355 
1356 		ptr_a = &num2;
1357 
1358 		memcpy(ptr_a, c, sizeof(BcNum));
1359 		init = true;
1360 	}
1361 	else ptr_a = a;
1362 
1363 	if (c == b) {
1364 
1365 		ptr_b = &num2;
1366 
1367 		if (c != a) {
1368 			memcpy(ptr_b, c, sizeof(BcNum));
1369 			init = true;
1370 		}
1371 	}
1372 	else ptr_b = b;
1373 
1374 	if (init) {
1375 
1376 		bc_num_init(c, req);
1377 
1378 		BC_SETJMP_LOCKED(err);
1379 		BC_SIG_UNLOCK;
1380 	}
1381 	else {
1382 		BC_SIG_UNLOCK;
1383 		bc_num_expand(c, req);
1384 	}
1385 
1386 	op(ptr_a, ptr_b, c, scale);
1387 
1388 	assert(!c->neg || BC_NUM_NONZERO(c));
1389 	assert(c->rdx <= c->len || !c->len);
1390 	assert(!c->len || c->num[c->len - 1] || c->rdx == c->len);
1391 
1392 err:
1393 	if (init) {
1394 		BC_SIG_MAYLOCK;
1395 		bc_num_free(&num2);
1396 		BC_LONGJMP_CONT;
1397 	}
1398 }
1399 
1400 #ifndef NDEBUG
1401 static bool bc_num_strValid(const char *val) {
1402 
1403 	bool radix = false;
1404 	size_t i, len = strlen(val);
1405 
1406 	if (!len) return true;
1407 
1408 	for (i = 0; i < len; ++i) {
1409 
1410 		BcDig c = val[i];
1411 
1412 		if (c == '.') {
1413 
1414 			if (radix) return false;
1415 
1416 			radix = true;
1417 			continue;
1418 		}
1419 
1420 		if (!(isdigit(c) || isupper(c))) return false;
1421 	}
1422 
1423 	return true;
1424 }
1425 #endif // NDEBUG
1426 
1427 static BcBigDig bc_num_parseChar(char c, size_t base_t) {
1428 
1429 	if (isupper(c)) {
1430 		c = BC_NUM_NUM_LETTER(c);
1431 		c = ((size_t) c) >= base_t ? (char) base_t - 1 : c;
1432 	}
1433 	else c -= '0';
1434 
1435 	return (BcBigDig) (uchar) c;
1436 }
1437 
1438 static void bc_num_parseDecimal(BcNum *restrict n, const char *restrict val) {
1439 
1440 	size_t len, i, temp, mod;
1441 	const char *ptr;
1442 	bool zero = true, rdx;
1443 
1444 	for (i = 0; val[i] == '0'; ++i);
1445 
1446 	val += i;
1447 	assert(!val[0] || isalnum(val[0]) || val[0] == '.');
1448 
1449 	// All 0's. We can just return, since this
1450 	// procedure expects a virgin (already 0) BcNum.
1451 	if (!val[0]) return;
1452 
1453 	len = strlen(val);
1454 
1455 	ptr = strchr(val, '.');
1456 	rdx = (ptr != NULL);
1457 
1458 	for (i = 0; i < len && (zero = (val[i] == '0' || val[i] == '.')); ++i);
1459 
1460 	n->scale = (size_t) (rdx * ((val + len) - (ptr + 1)));
1461 	n->rdx = BC_NUM_RDX(n->scale);
1462 
1463 	i = len - (ptr == val ? 0 : i) - rdx;
1464 	temp = BC_NUM_ROUND_POW(i);
1465 	mod = n->scale % BC_BASE_DIGS;
1466 	i = mod ? BC_BASE_DIGS - mod : 0;
1467 	n->len = ((temp + i) / BC_BASE_DIGS);
1468 
1469 	bc_num_expand(n, n->len);
1470 	memset(n->num, 0, BC_NUM_SIZE(n->len));
1471 
1472 	if (zero) n->len = n->rdx = 0;
1473 	else {
1474 
1475 		BcBigDig exp, pow;
1476 
1477 		assert(i <= BC_NUM_BIGDIG_MAX);
1478 
1479 		exp = (BcBigDig) i;
1480 		pow = bc_num_pow10[exp];
1481 
1482 		for (i = len - 1; i < len; --i, ++exp) {
1483 
1484 			char c = val[i];
1485 
1486 			if (c == '.') exp -= 1;
1487 			else {
1488 
1489 				size_t idx = exp / BC_BASE_DIGS;
1490 
1491 				if (isupper(c)) c = '9';
1492 				n->num[idx] += (((BcBigDig) c) - '0') * pow;
1493 
1494 				if ((exp + 1) % BC_BASE_DIGS == 0) pow = 1;
1495 				else pow *= BC_BASE;
1496 			}
1497 		}
1498 	}
1499 }
1500 
1501 static void bc_num_parseBase(BcNum *restrict n, const char *restrict val,
1502                              BcBigDig base)
1503 {
1504 	BcNum temp, mult1, mult2, result1, result2, *m1, *m2, *ptr;
1505 	char c = 0;
1506 	bool zero = true;
1507 	BcBigDig v;
1508 	size_t i, digs, len = strlen(val);
1509 
1510 	for (i = 0; zero && i < len; ++i) zero = (val[i] == '.' || val[i] == '0');
1511 	if (zero) return;
1512 
1513 	BC_SIG_LOCK;
1514 
1515 	bc_num_init(&temp, BC_NUM_BIGDIG_LOG10);
1516 	bc_num_init(&mult1, BC_NUM_BIGDIG_LOG10);
1517 
1518 	BC_SETJMP_LOCKED(int_err);
1519 
1520 	BC_SIG_UNLOCK;
1521 
1522 	for (i = 0; i < len && (c = val[i]) && c != '.'; ++i) {
1523 
1524 		v = bc_num_parseChar(c, base);
1525 
1526 		bc_num_mulArray(n, base, &mult1);
1527 		bc_num_bigdig2num(&temp, v);
1528 		bc_num_add(&mult1, &temp, n, 0);
1529 	}
1530 
1531 	if (i == len && !(c = val[i])) goto int_err;
1532 
1533 	assert(c == '.');
1534 
1535 	BC_SIG_LOCK;
1536 
1537 	BC_UNSETJMP;
1538 
1539 	bc_num_init(&mult2, BC_NUM_BIGDIG_LOG10);
1540 	bc_num_init(&result1, BC_NUM_DEF_SIZE);
1541 	bc_num_init(&result2, BC_NUM_DEF_SIZE);
1542 	bc_num_one(&mult1);
1543 
1544 	BC_SETJMP_LOCKED(err);
1545 
1546 	BC_SIG_UNLOCK;
1547 
1548 	m1 = &mult1;
1549 	m2 = &mult2;
1550 
1551 	for (i += 1, digs = 0; i < len && (c = val[i]); ++i, ++digs) {
1552 
1553 		v = bc_num_parseChar(c, base);
1554 
1555 		bc_num_mulArray(&result1, base, &result2);
1556 
1557 		bc_num_bigdig2num(&temp, v);
1558 		bc_num_add(&result2, &temp, &result1, 0);
1559 		bc_num_mulArray(m1, base, m2);
1560 
1561 		if (m2->len < m2->rdx) m2->len = m2->rdx;
1562 
1563 		ptr = m1;
1564 		m1 = m2;
1565 		m2 = ptr;
1566 	}
1567 
1568 	// This one cannot be a divide by 0 because mult starts out at 1, then is
1569 	// multiplied by base, and base cannot be 0, so mult cannot be 0.
1570 	bc_num_div(&result1, m1, &result2, digs * 2);
1571 	bc_num_truncate(&result2, digs);
1572 	bc_num_add(n, &result2, n, digs);
1573 
1574 	if (BC_NUM_NONZERO(n)) {
1575 		if (n->scale < digs) bc_num_extend(n, digs - n->scale);
1576 	}
1577 	else bc_num_zero(n);
1578 
1579 err:
1580 	BC_SIG_MAYLOCK;
1581 	bc_num_free(&result2);
1582 	bc_num_free(&result1);
1583 	bc_num_free(&mult2);
1584 int_err:
1585 	BC_SIG_MAYLOCK;
1586 	bc_num_free(&mult1);
1587 	bc_num_free(&temp);
1588 	BC_LONGJMP_CONT;
1589 }
1590 
1591 static void bc_num_printNewline(void) {
1592 	if (vm.nchars >= vm.line_len - 1) {
1593 		bc_vm_putchar('\\');
1594 		bc_vm_putchar('\n');
1595 	}
1596 }
1597 
1598 static void bc_num_putchar(int c) {
1599 	if (c != '\n') bc_num_printNewline();
1600 	bc_vm_putchar(c);
1601 }
1602 
1603 #if DC_ENABLED
1604 static void bc_num_printChar(size_t n, size_t len, bool rdx) {
1605 	BC_UNUSED(rdx);
1606 	BC_UNUSED(len);
1607 	assert(len == 1);
1608 	bc_vm_putchar((uchar) n);
1609 }
1610 #endif // DC_ENABLED
1611 
1612 static void bc_num_printDigits(size_t n, size_t len, bool rdx) {
1613 
1614 	size_t exp, pow;
1615 
1616 	bc_num_putchar(rdx ? '.' : ' ');
1617 
1618 	for (exp = 0, pow = 1; exp < len - 1; ++exp, pow *= BC_BASE);
1619 
1620 	for (exp = 0; exp < len; pow /= BC_BASE, ++exp) {
1621 		size_t dig = n / pow;
1622 		n -= dig * pow;
1623 		bc_num_putchar(((uchar) dig) + '0');
1624 	}
1625 }
1626 
1627 static void bc_num_printHex(size_t n, size_t len, bool rdx) {
1628 
1629 	BC_UNUSED(len);
1630 
1631 	assert(len == 1);
1632 
1633 	if (rdx) bc_num_putchar('.');
1634 
1635 	bc_num_putchar(bc_num_hex_digits[n]);
1636 }
1637 
1638 static void bc_num_printDecimal(const BcNum *restrict n) {
1639 
1640 	size_t i, j, rdx = n->rdx;
1641 	bool zero = true;
1642 	size_t buffer[BC_BASE_DIGS];
1643 
1644 	if (n->neg) bc_num_putchar('-');
1645 
1646 	for (i = n->len - 1; i < n->len; --i) {
1647 
1648 		BcDig n9 = n->num[i];
1649 		size_t temp;
1650 		bool irdx = (i == rdx - 1);
1651 
1652 		zero = (zero & !irdx);
1653 		temp = n->scale % BC_BASE_DIGS;
1654 		temp = i || !temp ? 0 : BC_BASE_DIGS - temp;
1655 
1656 		memset(buffer, 0, BC_BASE_DIGS * sizeof(size_t));
1657 
1658 		for (j = 0; n9 && j < BC_BASE_DIGS; ++j) {
1659 			buffer[j] = n9 % BC_BASE;
1660 			n9 /= BC_BASE;
1661 		}
1662 
1663 		for (j = BC_BASE_DIGS - 1; j < BC_BASE_DIGS && j >= temp; --j) {
1664 			bool print_rdx = (irdx & (j == BC_BASE_DIGS - 1));
1665 			zero = (zero && buffer[j] == 0);
1666 			if (!zero) bc_num_printHex(buffer[j], 1, print_rdx);
1667 		}
1668 	}
1669 }
1670 
1671 #if BC_ENABLE_EXTRA_MATH
1672 static void bc_num_printExponent(const BcNum *restrict n, bool eng) {
1673 
1674 	bool neg = (n->len <= n->rdx);
1675 	BcNum temp, exp;
1676 	size_t places, mod;
1677 	BcDig digs[BC_NUM_BIGDIG_LOG10];
1678 
1679 	BC_SIG_LOCK;
1680 
1681 	bc_num_createCopy(&temp, n);
1682 
1683 	BC_SETJMP_LOCKED(exit);
1684 
1685 	BC_SIG_UNLOCK;
1686 
1687 	if (neg) {
1688 
1689 		size_t i, idx = bc_num_nonzeroLen(n) - 1;
1690 
1691 		places = 1;
1692 
1693 		for (i = BC_BASE_DIGS - 1; i < BC_BASE_DIGS; --i) {
1694 			if (bc_num_pow10[i] > (BcBigDig) n->num[idx]) places += 1;
1695 			else break;
1696 		}
1697 
1698 		places += (n->rdx - (idx + 1)) * BC_BASE_DIGS;
1699 		mod = places % 3;
1700 
1701 		if (eng && mod != 0) places += 3 - mod;
1702 		bc_num_shiftLeft(&temp, places);
1703 	}
1704 	else {
1705 		places = bc_num_intDigits(n) - 1;
1706 		mod = places % 3;
1707 		if (eng && mod != 0) places -= 3 - (3 - mod);
1708 		bc_num_shiftRight(&temp, places);
1709 	}
1710 
1711 	bc_num_printDecimal(&temp);
1712 	bc_num_putchar('e');
1713 
1714 	if (!places) {
1715 		bc_num_printHex(0, 1, false);
1716 		goto exit;
1717 	}
1718 
1719 	if (neg) bc_num_putchar('-');
1720 
1721 	bc_num_setup(&exp, digs, BC_NUM_BIGDIG_LOG10);
1722 	bc_num_bigdig2num(&exp, (BcBigDig) places);
1723 
1724 	bc_num_printDecimal(&exp);
1725 
1726 exit:
1727 	BC_SIG_MAYLOCK;
1728 	bc_num_free(&temp);
1729 	BC_LONGJMP_CONT;
1730 }
1731 #endif // BC_ENABLE_EXTRA_MATH
1732 
1733 static void bc_num_printFixup(BcNum *restrict n, BcBigDig rem,
1734                               BcBigDig pow, size_t idx)
1735 {
1736 	size_t i, len = n->len - idx;
1737 	BcBigDig acc;
1738 	BcDig *a = n->num + idx;
1739 
1740 	if (len < 2) return;
1741 
1742 	for (i = len - 1; i > 0; --i) {
1743 
1744 		acc = ((BcBigDig) a[i]) * rem + ((BcBigDig) a[i - 1]);
1745 		a[i - 1] = (BcDig) (acc % pow);
1746 		acc /= pow;
1747 		acc += (BcBigDig) a[i];
1748 
1749 		if (acc >= BC_BASE_POW) {
1750 
1751 			if (i == len - 1) {
1752 				len = bc_vm_growSize(len, 1);
1753 				bc_num_expand(n, bc_vm_growSize(len, idx));
1754 				a = n->num + idx;
1755 				a[len - 1] = 0;
1756 			}
1757 
1758 			a[i + 1] += acc / BC_BASE_POW;
1759 			acc %= BC_BASE_POW;
1760 		}
1761 
1762 		assert(acc < BC_BASE_POW);
1763 		a[i] = (BcDig) acc;
1764 	}
1765 
1766 	n->len = len + idx;
1767 }
1768 
1769 static void bc_num_printPrepare(BcNum *restrict n, BcBigDig rem,
1770                                 BcBigDig pow)
1771 {
1772 	size_t i;
1773 
1774 	for (i = 0; i < n->len; ++i) bc_num_printFixup(n, rem, pow, i);
1775 
1776 	for (i = 0; i < n->len; ++i) {
1777 
1778 		assert(pow == ((BcBigDig) ((BcDig) pow)));
1779 
1780 		if (n->num[i] >= (BcDig) pow) {
1781 
1782 			if (i + 1 == n->len) {
1783 				n->len = bc_vm_growSize(n->len, 1);
1784 				bc_num_expand(n, n->len);
1785 				n->num[i + 1] = 0;
1786 			}
1787 
1788 			assert(pow < BC_BASE_POW);
1789 			n->num[i + 1] += n->num[i] / ((BcDig) pow);
1790 			n->num[i] %= (BcDig) pow;
1791 		}
1792 	}
1793 }
1794 
1795 static void bc_num_printNum(BcNum *restrict n, BcBigDig base,
1796                             size_t len, BcNumDigitOp print)
1797 {
1798 	BcVec stack;
1799 	BcNum intp, fracp1, fracp2, digit, flen1, flen2, *n1, *n2, *temp;
1800 	BcBigDig dig = 0, *ptr, acc, exp;
1801 	size_t i, j;
1802 	bool radix;
1803 	BcDig digit_digs[BC_NUM_BIGDIG_LOG10 + 1];
1804 
1805 	assert(base > 1);
1806 
1807 	if (BC_NUM_ZERO(n)) {
1808 		print(0, len, false);
1809 		return;
1810 	}
1811 
1812 	// This function uses an algorithm that Stefan Esser <se@freebsd.org> came
1813 	// up with to print the integer part of a number. What it does is convert
1814 	// intp into a number of the specified base, but it does it directly,
1815 	// instead of just doing a series of divisions and printing the remainders
1816 	// in reverse order.
1817 	//
1818 	// Let me explain in a bit more detail:
1819 	//
1820 	// The algorithm takes the current least significant digit (after intp has
1821 	// been converted to an integer) and the next to least significant digit,
1822 	// and it converts the least significant digit into one of the specified
1823 	// base, putting any overflow into the next to least significant digit. It
1824 	// iterates through the whole number, from least significant to most
1825 	// significant, doing this conversion. At the end of that iteration, the
1826 	// least significant digit is converted, but the others are not, so it
1827 	// iterates again, starting at the next to least significant digit. It keeps
1828 	// doing that conversion, skipping one more digit than the last time, until
1829 	// all digits have been converted. Then it prints them in reverse order.
1830 	//
1831 	// That is the gist of the algorithm. It leaves out several things, such as
1832 	// the fact that digits are not always converted into the specified base,
1833 	// but into something close, basically a power of the specified base. In
1834 	// Stefan's words, "You could consider BcDigs to be of base 10^BC_BASE_DIGS
1835 	// in the normal case and obase^N for the largest value of N that satisfies
1836 	// obase^N <= 10^BC_BASE_DIGS. [This means that] the result is not in base
1837 	// "obase", but in base "obase^N", which happens to be printable as a number
1838 	// of base "obase" without consideration for neighbouring BcDigs." This fact
1839 	// is what necessitates the existence of the loop later in this function.
1840 	//
1841 	// The conversion happens in bc_num_printPrepare() where the outer loop
1842 	// happens and bc_num_printFixup() where the inner loop, or actual
1843 	// conversion, happens.
1844 
1845 	BC_SIG_LOCK;
1846 
1847 	bc_vec_init(&stack, sizeof(BcBigDig), NULL);
1848 	bc_num_init(&fracp1, n->rdx);
1849 
1850 	bc_num_createCopy(&intp, n);
1851 
1852 	BC_SETJMP_LOCKED(err);
1853 
1854 	BC_SIG_UNLOCK;
1855 
1856 	bc_num_truncate(&intp, intp.scale);
1857 
1858 	bc_num_sub(n, &intp, &fracp1, 0);
1859 
1860 	if (base != vm.last_base) {
1861 
1862 		vm.last_pow = 1;
1863 		vm.last_exp = 0;
1864 
1865 		while (vm.last_pow * base <= BC_BASE_POW) {
1866 			vm.last_pow *= base;
1867 			vm.last_exp += 1;
1868 		}
1869 
1870 		vm.last_rem = BC_BASE_POW - vm.last_pow;
1871 		vm.last_base = base;
1872 	}
1873 
1874 	exp = vm.last_exp;
1875 
1876 	if (vm.last_rem != 0) bc_num_printPrepare(&intp, vm.last_rem, vm.last_pow);
1877 
1878 	for (i = 0; i < intp.len; ++i) {
1879 
1880 		acc = (BcBigDig) intp.num[i];
1881 
1882 		for (j = 0; j < exp && (i < intp.len - 1 || acc != 0); ++j)
1883 		{
1884 			if (j != exp - 1) {
1885 				dig = acc % base;
1886 				acc /= base;
1887 			}
1888 			else {
1889 				dig = acc;
1890 				acc = 0;
1891 			}
1892 
1893 			assert(dig < base);
1894 
1895 			bc_vec_push(&stack, &dig);
1896 		}
1897 
1898 		assert(acc == 0);
1899 	}
1900 
1901 	for (i = 0; i < stack.len; ++i) {
1902 		ptr = bc_vec_item_rev(&stack, i);
1903 		assert(ptr != NULL);
1904 		print(*ptr, len, false);
1905 	}
1906 
1907 	if (!n->scale) goto err;
1908 
1909 	BC_SIG_LOCK;
1910 
1911 	BC_UNSETJMP;
1912 
1913 	bc_num_init(&fracp2, n->rdx);
1914 	bc_num_setup(&digit, digit_digs, sizeof(digit_digs) / sizeof(BcDig));
1915 	bc_num_init(&flen1, BC_NUM_BIGDIG_LOG10);
1916 	bc_num_init(&flen2, BC_NUM_BIGDIG_LOG10);
1917 
1918 	BC_SETJMP_LOCKED(frac_err);
1919 
1920 	BC_SIG_UNLOCK;
1921 
1922 	bc_num_one(&flen1);
1923 
1924 	radix = true;
1925 	n1 = &flen1;
1926 	n2 = &flen2;
1927 
1928 	fracp2.scale = n->scale;
1929 	fracp2.rdx = BC_NUM_RDX(fracp2.scale);
1930 
1931 	while (bc_num_intDigits(n1) < n->scale + 1) {
1932 
1933 		bc_num_expand(&fracp2, fracp1.len + 1);
1934 		bc_num_mulArray(&fracp1, base, &fracp2);
1935 		if (fracp2.len < fracp2.rdx) fracp2.len = fracp2.rdx;
1936 
1937 		// fracp is guaranteed to be non-negative and small enough.
1938 		bc_num_bigdig2(&fracp2, &dig);
1939 
1940 		bc_num_bigdig2num(&digit, dig);
1941 		bc_num_sub(&fracp2, &digit, &fracp1, 0);
1942 
1943 		print(dig, len, radix);
1944 		bc_num_mulArray(n1, base, n2);
1945 
1946 		radix = false;
1947 		temp = n1;
1948 		n1 = n2;
1949 		n2 = temp;
1950 	}
1951 
1952 frac_err:
1953 	BC_SIG_MAYLOCK;
1954 	bc_num_free(&flen2);
1955 	bc_num_free(&flen1);
1956 	bc_num_free(&fracp2);
1957 err:
1958 	BC_SIG_MAYLOCK;
1959 	bc_num_free(&fracp1);
1960 	bc_num_free(&intp);
1961 	bc_vec_free(&stack);
1962 	BC_LONGJMP_CONT;
1963 }
1964 
1965 static void bc_num_printBase(BcNum *restrict n, BcBigDig base) {
1966 
1967 	size_t width;
1968 	BcNumDigitOp print;
1969 	bool neg = n->neg;
1970 
1971 	if (neg) bc_num_putchar('-');
1972 
1973 	n->neg = false;
1974 
1975 	if (base <= BC_NUM_MAX_POSIX_IBASE) {
1976 		width = 1;
1977 		print = bc_num_printHex;
1978 	}
1979 	else {
1980 		assert(base <= BC_BASE_POW);
1981 		width = bc_num_log10(base - 1);
1982 		print = bc_num_printDigits;
1983 	}
1984 
1985 	bc_num_printNum(n, base, width, print);
1986 	n->neg = neg;
1987 }
1988 
1989 #if DC_ENABLED
1990 void bc_num_stream(BcNum *restrict n, BcBigDig base) {
1991 	bc_num_printNum(n, base, 1, bc_num_printChar);
1992 }
1993 #endif // DC_ENABLED
1994 
1995 void bc_num_setup(BcNum *restrict n, BcDig *restrict num, size_t cap) {
1996 	assert(n != NULL);
1997 	n->num = num;
1998 	n->cap = cap;
1999 	bc_num_zero(n);
2000 }
2001 
2002 void bc_num_init(BcNum *restrict n, size_t req) {
2003 
2004 	BcDig *num;
2005 
2006 	BC_SIG_ASSERT_LOCKED;
2007 
2008 	assert(n != NULL);
2009 
2010 	req = req >= BC_NUM_DEF_SIZE ? req : BC_NUM_DEF_SIZE;
2011 
2012 	if (req == BC_NUM_DEF_SIZE && vm.temps.len) {
2013 		BcNum *nptr = bc_vec_top(&vm.temps);
2014 		num = nptr->num;
2015 		bc_vec_pop(&vm.temps);
2016 	}
2017 	else num = bc_vm_malloc(BC_NUM_SIZE(req));
2018 
2019 	bc_num_setup(n, num, req);
2020 }
2021 
2022 void bc_num_clear(BcNum *restrict n) {
2023 	n->num = NULL;
2024 	n->cap = 0;
2025 }
2026 
2027 void bc_num_free(void *num) {
2028 
2029 	BcNum *n = (BcNum*) num;
2030 
2031 	BC_SIG_ASSERT_LOCKED;
2032 
2033 	assert(n != NULL);
2034 
2035 	if (n->cap == BC_NUM_DEF_SIZE) bc_vec_push(&vm.temps, n);
2036 	else free(n->num);
2037 }
2038 
2039 void bc_num_copy(BcNum *d, const BcNum *s) {
2040 	assert(d != NULL && s != NULL);
2041 	if (d == s) return;
2042 	bc_num_expand(d, s->len);
2043 	d->len = s->len;
2044 	d->neg = s->neg;
2045 	d->rdx = s->rdx;
2046 	d->scale = s->scale;
2047 	memcpy(d->num, s->num, BC_NUM_SIZE(d->len));
2048 }
2049 
2050 void bc_num_createCopy(BcNum *d, const BcNum *s) {
2051 	BC_SIG_ASSERT_LOCKED;
2052 	bc_num_init(d, s->len);
2053 	bc_num_copy(d, s);
2054 }
2055 
2056 void bc_num_createFromBigdig(BcNum *n, BcBigDig val) {
2057 	BC_SIG_ASSERT_LOCKED;
2058 	bc_num_init(n, (BC_NUM_BIGDIG_LOG10 - 1) / BC_BASE_DIGS + 1);
2059 	bc_num_bigdig2num(n, val);
2060 }
2061 
2062 size_t bc_num_scale(const BcNum *restrict n) {
2063 	return n->scale;
2064 }
2065 
2066 size_t bc_num_len(const BcNum *restrict n) {
2067 
2068 	size_t len = n->len;
2069 
2070 	if (BC_NUM_ZERO(n)) return 0;
2071 
2072 	if (n->rdx == len) {
2073 
2074 		size_t zero, scale;
2075 
2076 		len = bc_num_nonzeroLen(n);
2077 
2078 		scale = n->scale % BC_BASE_DIGS;
2079 		scale = scale ? scale : BC_BASE_DIGS;
2080 
2081 		zero = bc_num_zeroDigits(n->num + len - 1);
2082 
2083 		len = len * BC_BASE_DIGS - zero - (BC_BASE_DIGS - scale);
2084 	}
2085 	else len = bc_num_intDigits(n) + n->scale;
2086 
2087 	return len;
2088 }
2089 
2090 void bc_num_parse(BcNum *restrict n, const char *restrict val,
2091                   BcBigDig base, bool letter)
2092 {
2093 	assert(n != NULL && val != NULL && base);
2094 	assert(base >= BC_NUM_MIN_BASE && base <= vm.maxes[BC_PROG_GLOBALS_IBASE]);
2095 	assert(bc_num_strValid(val));
2096 
2097 	if (letter) {
2098 		BcBigDig dig = bc_num_parseChar(val[0], BC_NUM_MAX_LBASE);
2099 		bc_num_bigdig2num(n, dig);
2100 	}
2101 	else if (base == BC_BASE) bc_num_parseDecimal(n, val);
2102 	else bc_num_parseBase(n, val, base);
2103 }
2104 
2105 void bc_num_print(BcNum *restrict n, BcBigDig base, bool newline) {
2106 
2107 	assert(n != NULL);
2108 	assert(BC_ENABLE_EXTRA_MATH || base >= BC_NUM_MIN_BASE);
2109 
2110 	bc_num_printNewline();
2111 
2112 	if (BC_NUM_ZERO(n)) bc_num_printHex(0, 1, false);
2113 	else if (base == BC_BASE) bc_num_printDecimal(n);
2114 #if BC_ENABLE_EXTRA_MATH
2115 	else if (base == 0 || base == 1)
2116 		bc_num_printExponent(n, base != 0);
2117 #endif // BC_ENABLE_EXTRA_MATH
2118 	else bc_num_printBase(n, base);
2119 
2120 	if (newline) bc_num_putchar('\n');
2121 }
2122 
2123 void bc_num_bigdig2(const BcNum *restrict n, BcBigDig *result) {
2124 
2125 	// This function returns no errors because it's guaranteed to succeed if
2126 	// its preconditions are met. Those preconditions include both parameters
2127 	// being non-NULL, n being non-negative, and n being less than vm.max. If
2128 	// all of that is true, then we can just convert without worrying about
2129 	// negative errors or overflow. We also don't care about signals because
2130 	// this function should execute in only a few iterations, meaning that
2131 	// ignoring signals here should be fine.
2132 
2133 	BcBigDig r = 0;
2134 
2135 	assert(n != NULL && result != NULL);
2136 	assert(!n->neg);
2137 	assert(bc_num_cmp(n, &vm.max) < 0);
2138 	assert(n->len - n->rdx <= 3);
2139 
2140 	// There is a small speed win from unrolling the loop here, and since it
2141 	// only adds 53 bytes, I decided that it was worth it.
2142 	switch (n->len - n->rdx) {
2143 		case 3:
2144 			r = (BcBigDig) n->num[n->rdx + 2];
2145 			// Fallthrough.
2146 		case 2:
2147 			r = r * BC_BASE_POW + (BcBigDig) n->num[n->rdx + 1];
2148 			// Fallthrough.
2149 		case 1:
2150 			r = r * BC_BASE_POW + (BcBigDig) n->num[n->rdx];
2151 	}
2152 
2153 	*result = r;
2154 }
2155 
2156 void bc_num_bigdig(const BcNum *restrict n, BcBigDig *result) {
2157 
2158 	assert(n != NULL && result != NULL);
2159 
2160 	if (BC_ERR(n->neg)) bc_vm_err(BC_ERROR_MATH_NEGATIVE);
2161 	if (BC_ERR(bc_num_cmp(n, &vm.max) >= 0))
2162 		bc_vm_err(BC_ERROR_MATH_OVERFLOW);
2163 
2164 	bc_num_bigdig2(n, result);
2165 }
2166 
2167 void bc_num_bigdig2num(BcNum *restrict n, BcBigDig val) {
2168 
2169 	BcDig *ptr;
2170 	size_t i;
2171 
2172 	assert(n != NULL);
2173 
2174 	bc_num_zero(n);
2175 
2176 	if (!val) return;
2177 
2178 	bc_num_expand(n, BC_NUM_BIGDIG_LOG10);
2179 
2180 	for (ptr = n->num, i = 0; val; ++i, val /= BC_BASE_POW)
2181 		ptr[i] = val % BC_BASE_POW;
2182 
2183 	n->len = i;
2184 }
2185 
2186 #if BC_ENABLE_EXTRA_MATH && BC_ENABLE_RAND
2187 void bc_num_rng(const BcNum *restrict n, BcRNG *rng) {
2188 
2189 	BcNum pow, temp, temp2, intn, frac;
2190 	BcRand state1, state2, inc1, inc2;
2191 	BcDig pow_num[BC_RAND_NUM_SIZE];
2192 
2193 	bc_num_setup(&pow, pow_num, sizeof(pow_num) / sizeof(BcDig));
2194 
2195 	BC_SIG_LOCK;
2196 
2197 	bc_num_init(&temp, n->len);
2198 	bc_num_init(&temp2, n->len);
2199 	bc_num_init(&frac, n->rdx);
2200 	bc_num_init(&intn, bc_num_int(n));
2201 
2202 	BC_SETJMP_LOCKED(err);
2203 
2204 	BC_SIG_UNLOCK;
2205 
2206 	bc_num_mul(&vm.max, &vm.max, &pow, 0);
2207 
2208 	memcpy(frac.num, n->num, BC_NUM_SIZE(n->rdx));
2209 	frac.len = n->rdx;
2210 	frac.rdx = n->rdx;
2211 	frac.scale = n->scale;
2212 
2213 	bc_num_mul(&frac, &pow, &temp, 0);
2214 
2215 	bc_num_truncate(&temp, temp.scale);
2216 	bc_num_copy(&frac, &temp);
2217 
2218 	memcpy(intn.num, n->num + n->rdx, BC_NUM_SIZE(bc_num_int(n)));
2219 	intn.len = bc_num_int(n);
2220 
2221 	// This assert is here because it has to be true. It is also here to justify
2222 	// the use of BC_ERROR_SIGNAL_ONLY() on each of the divmod's and mod's
2223 	// below.
2224 	assert(BC_NUM_NONZERO(&vm.max));
2225 
2226 	if (BC_NUM_NONZERO(&frac)) {
2227 
2228 		bc_num_divmod(&frac, &vm.max, &temp, &temp2, 0);
2229 
2230 		// frac is guaranteed to be smaller than vm.max * vm.max (pow).
2231 		// This means that when dividing frac by vm.max, as above, the
2232 		// quotient and remainder are both guaranteed to be less than vm.max,
2233 		// which means we can use bc_num_bigdig2() here and not worry about
2234 		// overflow.
2235 		bc_num_bigdig2(&temp2, (BcBigDig*) &state1);
2236 		bc_num_bigdig2(&temp, (BcBigDig*) &state2);
2237 	}
2238 	else state1 = state2 = 0;
2239 
2240 	if (BC_NUM_NONZERO(&intn)) {
2241 
2242 		bc_num_divmod(&intn, &vm.max, &temp, &temp2, 0);
2243 
2244 		// Because temp2 is the mod of vm.max, from above, it is guaranteed
2245 		// to be small enough to use bc_num_bigdig2().
2246 		bc_num_bigdig2(&temp2, (BcBigDig*) &inc1);
2247 
2248 		if (bc_num_cmp(&temp, &vm.max) >= 0) {
2249 			bc_num_copy(&temp2, &temp);
2250 			bc_num_mod(&temp2, &vm.max, &temp, 0);
2251 		}
2252 
2253 		// The if statement above ensures that temp is less than vm.max, which
2254 		// means that we can use bc_num_bigdig2() here.
2255 		bc_num_bigdig2(&temp, (BcBigDig*) &inc2);
2256 	}
2257 	else inc1 = inc2 = 0;
2258 
2259 	bc_rand_seed(rng, state1, state2, inc1, inc2);
2260 
2261 err:
2262 	BC_SIG_MAYLOCK;
2263 	bc_num_free(&intn);
2264 	bc_num_free(&frac);
2265 	bc_num_free(&temp2);
2266 	bc_num_free(&temp);
2267 	BC_LONGJMP_CONT;
2268 }
2269 
2270 void bc_num_createFromRNG(BcNum *restrict n, BcRNG *rng) {
2271 
2272 	BcRand s1, s2, i1, i2;
2273 	BcNum pow, conv, temp1, temp2, temp3;
2274 	BcDig pow_num[BC_RAND_NUM_SIZE];
2275 	BcDig temp1_num[BC_RAND_NUM_SIZE], temp2_num[BC_RAND_NUM_SIZE];
2276 	BcDig conv_num[BC_NUM_BIGDIG_LOG10];
2277 
2278 	BC_SIG_LOCK;
2279 
2280 	bc_num_init(&temp3, 2 * BC_RAND_NUM_SIZE);
2281 
2282 	BC_SETJMP_LOCKED(err);
2283 
2284 	BC_SIG_UNLOCK;
2285 
2286 	bc_num_setup(&pow, pow_num, sizeof(pow_num) / sizeof(BcDig));
2287 	bc_num_setup(&temp1, temp1_num, sizeof(temp1_num) / sizeof(BcDig));
2288 	bc_num_setup(&temp2, temp2_num, sizeof(temp2_num) / sizeof(BcDig));
2289 	bc_num_setup(&conv, conv_num, sizeof(conv_num) / sizeof(BcDig));
2290 
2291 	// This assert is here because it has to be true. It is also here to justify
2292 	// the assumption that pow is not zero.
2293 	assert(BC_NUM_NONZERO(&vm.max));
2294 
2295 	bc_num_mul(&vm.max, &vm.max, &pow, 0);
2296 
2297 	// Because this is true, we can just use BC_ERROR_SIGNAL_ONLY() below when
2298 	// dividing by pow.
2299 	assert(BC_NUM_NONZERO(&pow));
2300 
2301 	bc_rand_getRands(rng, &s1, &s2, &i1, &i2);
2302 
2303 	bc_num_bigdig2num(&conv, (BcBigDig) s2);
2304 
2305 	bc_num_mul(&conv, &vm.max, &temp1, 0);
2306 
2307 	bc_num_bigdig2num(&conv, (BcBigDig) s1);
2308 
2309 	bc_num_add(&conv, &temp1, &temp2, 0);
2310 
2311 	bc_num_div(&temp2, &pow, &temp3, BC_RAND_STATE_BITS);
2312 
2313 	bc_num_bigdig2num(&conv, (BcBigDig) i2);
2314 
2315 	bc_num_mul(&conv, &vm.max, &temp1, 0);
2316 
2317 	bc_num_bigdig2num(&conv, (BcBigDig) i1);
2318 
2319 	bc_num_add(&conv, &temp1, &temp2, 0);
2320 
2321 	bc_num_add(&temp2, &temp3, n, 0);
2322 
2323 err:
2324 	BC_SIG_MAYLOCK;
2325 	bc_num_free(&temp3);
2326 	BC_LONGJMP_CONT;
2327 }
2328 
2329 void bc_num_irand(const BcNum *restrict a, BcNum *restrict b,
2330                   BcRNG *restrict rng)
2331 {
2332 	BcRand r;
2333 	BcBigDig modl;
2334 	BcNum pow, pow2, cp, cp2, mod, temp1, temp2, rand;
2335 	BcNum *p1, *p2, *t1, *t2, *c1, *c2, *tmp;
2336 	BcDig rand_num[BC_NUM_BIGDIG_LOG10];
2337 	bool carry;
2338 	ssize_t cmp;
2339 
2340 	assert(a != b);
2341 
2342 	if (BC_ERR(a->neg)) bc_vm_err(BC_ERROR_MATH_NEGATIVE);
2343 	if (BC_ERR(a->rdx)) bc_vm_err(BC_ERROR_MATH_NON_INTEGER);
2344 	if (BC_NUM_ZERO(a) || BC_NUM_ONE(a)) return;
2345 
2346 	cmp = bc_num_cmp(a, &vm.max);
2347 
2348 	if (cmp <= 0) {
2349 
2350 		BcRand bits = 0;
2351 
2352 		if (cmp < 0) bc_num_bigdig2(a, (BcBigDig*) &bits);
2353 
2354 		// This condition means that bits is a power of 2. In that case, we
2355 		// can just grab a full-size int and mask out the unneeded bits.
2356 		// Also, this condition says that 0 is a power of 2, which works for
2357 		// us, since a value of 0 means a == rng->max. The bitmask will mask
2358 		// nothing in that case as well.
2359 		if (!(bits & (bits - 1))) r = bc_rand_int(rng) & (bits - 1);
2360 		else r = bc_rand_bounded(rng, bits);
2361 
2362 		// We made sure that r is less than vm.max,
2363 		// so we can use bc_num_bigdig2() here.
2364 		bc_num_bigdig2num(b, r);
2365 
2366 		return;
2367 	}
2368 
2369 	// In the case where a is less than rng->max, we have to make sure we have
2370 	// an exclusive bound. This ensures that it happens. (See below.)
2371 	carry = (cmp < 0);
2372 
2373 	BC_SIG_LOCK;
2374 
2375 	bc_num_createCopy(&cp, a);
2376 
2377 	bc_num_init(&cp2, cp.len);
2378 	bc_num_init(&mod, BC_NUM_BIGDIG_LOG10);
2379 	bc_num_init(&temp1, BC_NUM_DEF_SIZE);
2380 	bc_num_init(&temp2, BC_NUM_DEF_SIZE);
2381 	bc_num_init(&pow2, BC_NUM_DEF_SIZE);
2382 	bc_num_init(&pow, BC_NUM_DEF_SIZE);
2383 	bc_num_one(&pow);
2384 	bc_num_setup(&rand, rand_num, sizeof(rand_num) / sizeof(BcDig));
2385 
2386 	BC_SETJMP_LOCKED(err);
2387 
2388 	BC_SIG_UNLOCK;
2389 
2390 	p1 = &pow;
2391 	p2 = &pow2;
2392 	t1 = &temp1;
2393 	t2 = &temp2;
2394 	c1 = &cp;
2395 	c2 = &cp2;
2396 
2397 	// This assert is here because it has to be true. It is also here to justify
2398 	// the use of BC_ERROR_SIGNAL_ONLY() on each of the divmod's and mod's
2399 	// below.
2400 	assert(BC_NUM_NONZERO(&vm.max));
2401 
2402 	while (BC_NUM_NONZERO(c1)) {
2403 
2404 		bc_num_divmod(c1, &vm.max, c2, &mod, 0);
2405 
2406 		// Because mod is the mod of vm.max, it is guaranteed to be smaller,
2407 		// which means we can use bc_num_bigdig2() here.
2408 		bc_num_bigdig(&mod, &modl);
2409 
2410 		if (bc_num_cmp(c1, &vm.max) < 0) {
2411 
2412 			// In this case, if there is no carry, then we know we can generate
2413 			// an integer *equal* to modl. Thus, we add one if there is no
2414 			// carry. Otherwise, we add zero, and we are still bounded properly.
2415 			// Since the last portion is guaranteed to be greater than 1, we
2416 			// know modl isn't 0 unless there is no carry.
2417 			modl += !carry;
2418 
2419 			if (modl == 1) r = 0;
2420 			else if (!modl) r = bc_rand_int(rng);
2421 			else r = bc_rand_bounded(rng, (BcRand) modl);
2422 		}
2423 		else {
2424 			if (modl) modl -= carry;
2425 			r = bc_rand_int(rng);
2426 			carry = (r >= (BcRand) modl);
2427 		}
2428 
2429 		bc_num_bigdig2num(&rand, r);
2430 
2431 		bc_num_mul(&rand, p1, p2, 0);
2432 		bc_num_add(p2, t1, t2, 0);
2433 
2434 		if (BC_NUM_NONZERO(c2)) {
2435 
2436 			bc_num_mul(&vm.max, p1, p2, 0);
2437 
2438 			tmp = p1;
2439 			p1 = p2;
2440 			p2 = tmp;
2441 
2442 			tmp = c1;
2443 			c1 = c2;
2444 			c2 = tmp;
2445 		}
2446 		else c1 = c2;
2447 
2448 		tmp = t1;
2449 		t1 = t2;
2450 		t2 = tmp;
2451 	}
2452 
2453 	bc_num_copy(b, t1);
2454 	bc_num_clean(b);
2455 
2456 err:
2457 	BC_SIG_MAYLOCK;
2458 	bc_num_free(&pow);
2459 	bc_num_free(&pow2);
2460 	bc_num_free(&temp2);
2461 	bc_num_free(&temp1);
2462 	bc_num_free(&mod);
2463 	bc_num_free(&cp2);
2464 	bc_num_free(&cp);
2465 	BC_LONGJMP_CONT;
2466 }
2467 #endif // BC_ENABLE_EXTRA_MATH && BC_ENABLE_RAND
2468 
2469 size_t bc_num_addReq(const BcNum *a, const BcNum *b, size_t scale) {
2470 
2471 	size_t aint, bint, ardx, brdx;
2472 
2473 	BC_UNUSED(scale);
2474 
2475 	ardx = a->rdx;
2476 	aint = bc_num_int(a);
2477 	assert(aint <= a->len && ardx <= a->len);
2478 
2479 	brdx = b->rdx;
2480 	bint = bc_num_int(b);
2481 	assert(bint <= b->len && brdx <= b->len);
2482 
2483 	ardx = BC_MAX(ardx, brdx);
2484 	aint = BC_MAX(aint, bint);
2485 
2486 	return bc_vm_growSize(bc_vm_growSize(ardx, aint), 1);
2487 }
2488 
2489 size_t bc_num_mulReq(const BcNum *a, const BcNum *b, size_t scale) {
2490 	size_t max, rdx;
2491 	rdx = bc_vm_growSize(a->rdx, b->rdx);
2492 	max = BC_NUM_RDX(scale);
2493 	max = bc_vm_growSize(BC_MAX(max, rdx), 1);
2494 	rdx = bc_vm_growSize(bc_vm_growSize(bc_num_int(a), bc_num_int(b)), max);
2495 	return rdx;
2496 }
2497 
2498 size_t bc_num_powReq(const BcNum *a, const BcNum *b, size_t scale) {
2499 	BC_UNUSED(scale);
2500 	return bc_vm_growSize(bc_vm_growSize(a->len, b->len), 1);
2501 }
2502 
2503 #if BC_ENABLE_EXTRA_MATH
2504 size_t bc_num_placesReq(const BcNum *a, const BcNum *b, size_t scale) {
2505 	BC_UNUSED(scale);
2506 	return a->len + b->len - a->rdx - b->rdx;
2507 }
2508 #endif // BC_ENABLE_EXTRA_MATH
2509 
2510 void bc_num_add(BcNum *a, BcNum *b, BcNum *c, size_t scale) {
2511 	bc_num_binary(a, b, c, false, bc_num_as, bc_num_addReq(a, b, scale));
2512 }
2513 
2514 void bc_num_sub(BcNum *a, BcNum *b, BcNum *c, size_t scale) {
2515 	bc_num_binary(a, b, c, true, bc_num_as, bc_num_addReq(a, b, scale));
2516 }
2517 
2518 void bc_num_mul(BcNum *a, BcNum *b, BcNum *c, size_t scale) {
2519 	bc_num_binary(a, b, c, scale, bc_num_m, bc_num_mulReq(a, b, scale));
2520 }
2521 
2522 void bc_num_div(BcNum *a, BcNum *b, BcNum *c, size_t scale) {
2523 	bc_num_binary(a, b, c, scale, bc_num_d, bc_num_mulReq(a, b, scale));
2524 }
2525 
2526 void bc_num_mod(BcNum *a, BcNum *b, BcNum *c, size_t scale) {
2527 	bc_num_binary(a, b, c, scale, bc_num_rem, bc_num_mulReq(a, b, scale));
2528 }
2529 
2530 void bc_num_pow(BcNum *a, BcNum *b, BcNum *c, size_t scale) {
2531 	bc_num_binary(a, b, c, scale, bc_num_p, bc_num_powReq(a, b, scale));
2532 }
2533 
2534 #if BC_ENABLE_EXTRA_MATH
2535 void bc_num_places(BcNum *a, BcNum *b, BcNum *c, size_t scale) {
2536 	bc_num_binary(a, b, c, scale, bc_num_place, bc_num_placesReq(a, b, scale));
2537 }
2538 
2539 void bc_num_lshift(BcNum *a, BcNum *b, BcNum *c, size_t scale) {
2540 	bc_num_binary(a, b, c, scale, bc_num_left, bc_num_placesReq(a, b, scale));
2541 }
2542 
2543 void bc_num_rshift(BcNum *a, BcNum *b, BcNum *c, size_t scale) {
2544 	bc_num_binary(a, b, c, scale, bc_num_right, bc_num_placesReq(a, b, scale));
2545 }
2546 #endif // BC_ENABLE_EXTRA_MATH
2547 
2548 void bc_num_sqrt(BcNum *restrict a, BcNum *restrict b, size_t scale) {
2549 
2550 	BcNum num1, num2, half, f, fprime, *x0, *x1, *temp;
2551 	size_t pow, len, rdx, req, digs, digs1, digs2, resscale;
2552 	BcDig half_digs[1];
2553 
2554 	assert(a != NULL && b != NULL && a != b);
2555 
2556 	if (BC_ERR(a->neg)) bc_vm_err(BC_ERROR_MATH_NEGATIVE);
2557 
2558 	if (a->scale > scale) scale = a->scale;
2559 
2560 	len = bc_vm_growSize(bc_num_intDigits(a), 1);
2561 	rdx = BC_NUM_RDX(scale);
2562 	req = bc_vm_growSize(BC_MAX(rdx, a->rdx), len >> 1);
2563 
2564 	BC_SIG_LOCK;
2565 
2566 	bc_num_init(b, bc_vm_growSize(req, 1));
2567 
2568 	BC_SIG_UNLOCK;
2569 
2570 	if (BC_NUM_ZERO(a)) {
2571 		bc_num_setToZero(b, scale);
2572 		return;
2573 	}
2574 	if (BC_NUM_ONE(a)) {
2575 		bc_num_one(b);
2576 		bc_num_extend(b, scale);
2577 		return;
2578 	}
2579 
2580 	rdx = BC_NUM_RDX(scale);
2581 	rdx = BC_MAX(rdx, a->rdx);
2582 	len = bc_vm_growSize(a->len, rdx);
2583 
2584 	BC_SIG_LOCK;
2585 
2586 	bc_num_init(&num1, len);
2587 	bc_num_init(&num2, len);
2588 	bc_num_setup(&half, half_digs, sizeof(half_digs) / sizeof(BcDig));
2589 
2590 	bc_num_one(&half);
2591 	half.num[0] = BC_BASE_POW / 2;
2592 	half.len = 1;
2593 	half.rdx = 1;
2594 	half.scale = 1;
2595 
2596 	bc_num_init(&f, len);
2597 	bc_num_init(&fprime, len);
2598 
2599 	BC_SETJMP_LOCKED(err);
2600 
2601 	BC_SIG_UNLOCK;
2602 
2603 	x0 = &num1;
2604 	x1 = &num2;
2605 
2606 	bc_num_one(x0);
2607 	pow = bc_num_intDigits(a);
2608 
2609 	if (pow) {
2610 
2611 		if (pow & 1) x0->num[0] = 2;
2612 		else x0->num[0] = 6;
2613 
2614 		pow -= 2 - (pow & 1);
2615 		bc_num_shiftLeft(x0, pow / 2);
2616 	}
2617 
2618 	x0->scale = x0->rdx = digs = digs1 = digs2 = 0;
2619 	resscale = (scale + BC_BASE_DIGS) + 2;
2620 
2621 	while (bc_num_cmp(x1, x0)) {
2622 
2623 		assert(BC_NUM_NONZERO(x0));
2624 
2625 		bc_num_div(a, x0, &f, resscale);
2626 		bc_num_add(x0, &f, &fprime, resscale);
2627 		bc_num_mul(&fprime, &half, x1, resscale);
2628 
2629 		temp = x0;
2630 		x0 = x1;
2631 		x1 = temp;
2632 	}
2633 
2634 	bc_num_copy(b, x0);
2635 	if (b->scale > scale) bc_num_truncate(b, b->scale - scale);
2636 
2637 	assert(!b->neg || BC_NUM_NONZERO(b));
2638 	assert(b->rdx <= b->len || !b->len);
2639 	assert(!b->len || b->num[b->len - 1] || b->rdx == b->len);
2640 
2641 err:
2642 	BC_SIG_MAYLOCK;
2643 	bc_num_free(&fprime);
2644 	bc_num_free(&f);
2645 	bc_num_free(&num2);
2646 	bc_num_free(&num1);
2647 	BC_LONGJMP_CONT;
2648 }
2649 
2650 void bc_num_divmod(BcNum *a, BcNum *b, BcNum *c, BcNum *d, size_t scale) {
2651 
2652 	BcNum num2, *ptr_a;
2653 	bool init = false;
2654 	size_t ts, len;
2655 
2656 	ts = BC_MAX(scale + b->scale, a->scale);
2657 	len = bc_num_mulReq(a, b, ts);
2658 
2659 	assert(a != NULL && b != NULL && c != NULL && d != NULL);
2660 	assert(c != d && a != d && b != d && b != c);
2661 
2662 	if (c == a) {
2663 
2664 		memcpy(&num2, c, sizeof(BcNum));
2665 		ptr_a = &num2;
2666 
2667 		BC_SIG_LOCK;
2668 
2669 		bc_num_init(c, len);
2670 
2671 		init = true;
2672 
2673 		BC_SETJMP_LOCKED(err);
2674 
2675 		BC_SIG_UNLOCK;
2676 	}
2677 	else {
2678 		ptr_a = a;
2679 		bc_num_expand(c, len);
2680 	}
2681 
2682 	if (BC_NUM_NONZERO(a) && !a->rdx && !b->rdx && b->len == 1 && !scale) {
2683 
2684 		BcBigDig rem;
2685 
2686 		bc_num_divArray(ptr_a, (BcBigDig) b->num[0], c, &rem);
2687 
2688 		assert(rem < BC_BASE_POW);
2689 
2690 		d->num[0] = (BcDig) rem;
2691 		d->len = (rem != 0);
2692 	}
2693 	else bc_num_r(ptr_a, b, c, d, scale, ts);
2694 
2695 	assert(!c->neg || BC_NUM_NONZERO(c));
2696 	assert(c->rdx <= c->len || !c->len);
2697 	assert(!c->len || c->num[c->len - 1] || c->rdx == c->len);
2698 	assert(!d->neg || BC_NUM_NONZERO(d));
2699 	assert(d->rdx <= d->len || !d->len);
2700 	assert(!d->len || d->num[d->len - 1] || d->rdx == d->len);
2701 
2702 err:
2703 	if (init) {
2704 		BC_SIG_MAYLOCK;
2705 		bc_num_free(&num2);
2706 		BC_LONGJMP_CONT;
2707 	}
2708 }
2709 
2710 #if DC_ENABLED
2711 void bc_num_modexp(BcNum *a, BcNum *b, BcNum *c, BcNum *restrict d) {
2712 
2713 	BcNum base, exp, two, temp;
2714 	BcDig two_digs[2];
2715 
2716 	assert(a != NULL && b != NULL && c != NULL && d != NULL);
2717 	assert(a != d && b != d && c != d);
2718 
2719 	if (BC_ERR(BC_NUM_ZERO(c))) bc_vm_err(BC_ERROR_MATH_DIVIDE_BY_ZERO);
2720 	if (BC_ERR(b->neg)) bc_vm_err(BC_ERROR_MATH_NEGATIVE);
2721 	if (BC_ERR(a->rdx || b->rdx || c->rdx))
2722 		bc_vm_err(BC_ERROR_MATH_NON_INTEGER);
2723 
2724 	bc_num_expand(d, c->len);
2725 
2726 	BC_SIG_LOCK;
2727 
2728 	bc_num_init(&base, c->len);
2729 	bc_num_setup(&two, two_digs, sizeof(two_digs) / sizeof(BcDig));
2730 	bc_num_init(&temp, b->len + 1);
2731 	bc_num_createCopy(&exp, b);
2732 
2733 	BC_SETJMP_LOCKED(err);
2734 
2735 	BC_SIG_UNLOCK;
2736 
2737 	bc_num_one(&two);
2738 	two.num[0] = 2;
2739 	bc_num_one(d);
2740 
2741 	// We already checked for 0.
2742 	bc_num_rem(a, c, &base, 0);
2743 
2744 	while (BC_NUM_NONZERO(&exp)) {
2745 
2746 		// Num two cannot be 0, so no errors.
2747 		bc_num_divmod(&exp, &two, &exp, &temp, 0);
2748 
2749 		if (BC_NUM_ONE(&temp) && !temp.neg) {
2750 
2751 			bc_num_mul(d, &base, &temp, 0);
2752 
2753 			// We already checked for 0.
2754 			bc_num_rem(&temp, c, d, 0);
2755 		}
2756 
2757 		bc_num_mul(&base, &base, &temp, 0);
2758 
2759 		// We already checked for 0.
2760 		bc_num_rem(&temp, c, &base, 0);
2761 	}
2762 
2763 err:
2764 	BC_SIG_MAYLOCK;
2765 	bc_num_free(&exp);
2766 	bc_num_free(&temp);
2767 	bc_num_free(&base);
2768 	BC_LONGJMP_CONT;
2769 	assert(!d->neg || d->len);
2770 	assert(!d->len || d->num[d->len - 1] || d->rdx == d->len);
2771 }
2772 #endif // DC_ENABLED
2773 
2774 #if BC_DEBUG_CODE
2775 void bc_num_printDebug(const BcNum *n, const char *name, bool emptyline) {
2776 	bc_file_puts(&vm.fout, name);
2777 	bc_file_puts(&vm.fout, ": ");
2778 	bc_num_printDecimal(n);
2779 	bc_file_putchar(&vm.fout, '\n');
2780 	if (emptyline) bc_file_putchar(&vm.fout, '\n');
2781 	vm.nchars = 0;
2782 }
2783 
2784 void bc_num_printDigs(const BcDig *n, size_t len, bool emptyline) {
2785 
2786 	size_t i;
2787 
2788 	for (i = len - 1; i < len; --i)
2789 		bc_file_printf(&vm.fout, " %lu", (unsigned long) n[i]);
2790 
2791 	bc_file_putchar(&vm.fout, '\n');
2792 	if (emptyline) bc_file_putchar(&vm.fout, '\n');
2793 	vm.nchars = 0;
2794 }
2795 
2796 void bc_num_printWithDigs(const BcNum *n, const char *name, bool emptyline) {
2797 	bc_file_puts(&vm.fout, name);
2798 	bc_file_printf(&vm.fout, " len: %zu, rdx: %zu, scale: %zu\n",
2799 	               name, n->len, n->rdx, n->scale);
2800 	bc_num_printDigs(n->num, n->len, emptyline);
2801 }
2802 
2803 void bc_num_dump(const char *varname, const BcNum *n) {
2804 
2805 	ulong i, scale = n->scale;
2806 
2807 	bc_file_printf(&vm.ferr, "\n%s = %s", varname,
2808 	               n->len ? (n->neg ? "-" : "+") : "0 ");
2809 
2810 	for (i = n->len - 1; i < n->len; --i) {
2811 
2812 		if (i + 1 == n->rdx) bc_file_puts(&vm.ferr, ". ");
2813 
2814 		if (scale / BC_BASE_DIGS != n->rdx - i - 1)
2815 			bc_file_printf(&vm.ferr, "%lu ", (unsigned long) n->num[i]);
2816 		else {
2817 
2818 			int mod = scale % BC_BASE_DIGS;
2819 			int d = BC_BASE_DIGS - mod;
2820 			BcDig div;
2821 
2822 			if (mod != 0) {
2823 				div = n->num[i] / ((BcDig) bc_num_pow10[(ulong) d]);
2824 				bc_file_printf(&vm.ferr, "%lu", (unsigned long) div);
2825 			}
2826 
2827 			div = n->num[i] % ((BcDig) bc_num_pow10[(ulong) d]);
2828 			bc_file_printf(&vm.ferr, " ' %lu ", (unsigned long) div);
2829 		}
2830 	}
2831 
2832 	bc_file_printf(&vm.ferr, "(%zu | %zu.%zu / %zu) %lu\n",
2833 	               n->scale, n->len, n->rdx, n->cap,
2834 	               (unsigned long) (void*) n->num);
2835 }
2836 #endif // BC_DEBUG_CODE
2837