xref: /freebsd/lib/libc/quad/qdivrem.c (revision 4b9d6057)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1992, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This software was developed by the Computer Systems Engineering group
8  * at Lawrence Berkeley Laboratory under DARPA contract BG 91-66 and
9  * contributed to Berkeley.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  */
35 
36 /*
37  * Multiprecision divide.  This algorithm is from Knuth vol. 2 (2nd ed),
38  * section 4.3.1, pp. 257--259.
39  */
40 
41 #include "quad.h"
42 
43 #define	B	(1L << HALF_BITS)	/* digit base */
44 
45 /* Combine two `digits' to make a single two-digit number. */
46 #define	COMBINE(a, b) (((u_long)(a) << HALF_BITS) | (b))
47 
48 /* select a type for digits in base B: use unsigned short if they fit */
49 #if ULONG_MAX == 0xffffffff && USHRT_MAX >= 0xffff
50 typedef unsigned short digit;
51 #else
52 typedef u_long digit;
53 #endif
54 
55 /*
56  * Shift p[0]..p[len] left `sh' bits, ignoring any bits that
57  * `fall out' the left (there never will be any such anyway).
58  * We may assume len >= 0.  NOTE THAT THIS WRITES len+1 DIGITS.
59  */
60 static void
61 shl(digit *p, int len, int sh)
62 {
63 	int i;
64 
65 	for (i = 0; i < len; i++)
66 		p[i] = LHALF(p[i] << sh) | (p[i + 1] >> (HALF_BITS - sh));
67 	p[i] = LHALF(p[i] << sh);
68 }
69 
70 /*
71  * __qdivrem(u, v, rem) returns u/v and, optionally, sets *rem to u%v.
72  *
73  * We do this in base 2-sup-HALF_BITS, so that all intermediate products
74  * fit within u_long.  As a consequence, the maximum length dividend and
75  * divisor are 4 `digits' in this base (they are shorter if they have
76  * leading zeros).
77  */
78 u_quad_t
79 __qdivrem(u_quad_t uq, u_quad_t vq, u_quad_t *arq)
80 {
81 	union uu tmp;
82 	digit *u, *v, *q;
83 	digit v1, v2;
84 	u_long qhat, rhat, t;
85 	int m, n, d, j, i;
86 	digit uspace[5], vspace[5], qspace[5];
87 
88 	/*
89 	 * Take care of special cases: divide by zero, and u < v.
90 	 */
91 	if (__predict_false(vq == 0)) {
92 		/* divide by zero. */
93 		static volatile const unsigned int zero = 0;
94 
95 		tmp.ul[H] = tmp.ul[L] = 1 / zero;
96 		if (arq)
97 			*arq = uq;
98 		return (tmp.q);
99 	}
100 	if (uq < vq) {
101 		if (arq)
102 			*arq = uq;
103 		return (0);
104 	}
105 	u = &uspace[0];
106 	v = &vspace[0];
107 	q = &qspace[0];
108 
109 	/*
110 	 * Break dividend and divisor into digits in base B, then
111 	 * count leading zeros to determine m and n.  When done, we
112 	 * will have:
113 	 *	u = (u[1]u[2]...u[m+n]) sub B
114 	 *	v = (v[1]v[2]...v[n]) sub B
115 	 *	v[1] != 0
116 	 *	1 < n <= 4 (if n = 1, we use a different division algorithm)
117 	 *	m >= 0 (otherwise u < v, which we already checked)
118 	 *	m + n = 4
119 	 * and thus
120 	 *	m = 4 - n <= 2
121 	 */
122 	tmp.uq = uq;
123 	u[0] = 0;
124 	u[1] = HHALF(tmp.ul[H]);
125 	u[2] = LHALF(tmp.ul[H]);
126 	u[3] = HHALF(tmp.ul[L]);
127 	u[4] = LHALF(tmp.ul[L]);
128 	tmp.uq = vq;
129 	v[1] = HHALF(tmp.ul[H]);
130 	v[2] = LHALF(tmp.ul[H]);
131 	v[3] = HHALF(tmp.ul[L]);
132 	v[4] = LHALF(tmp.ul[L]);
133 	for (n = 4; v[1] == 0; v++) {
134 		if (--n == 1) {
135 			u_long rbj;	/* r*B+u[j] (not root boy jim) */
136 			digit q1, q2, q3, q4;
137 
138 			/*
139 			 * Change of plan, per exercise 16.
140 			 *	r = 0;
141 			 *	for j = 1..4:
142 			 *		q[j] = floor((r*B + u[j]) / v),
143 			 *		r = (r*B + u[j]) % v;
144 			 * We unroll this completely here.
145 			 */
146 			t = v[2];	/* nonzero, by definition */
147 			q1 = u[1] / t;
148 			rbj = COMBINE(u[1] % t, u[2]);
149 			q2 = rbj / t;
150 			rbj = COMBINE(rbj % t, u[3]);
151 			q3 = rbj / t;
152 			rbj = COMBINE(rbj % t, u[4]);
153 			q4 = rbj / t;
154 			if (arq)
155 				*arq = rbj % t;
156 			tmp.ul[H] = COMBINE(q1, q2);
157 			tmp.ul[L] = COMBINE(q3, q4);
158 			return (tmp.q);
159 		}
160 	}
161 
162 	/*
163 	 * By adjusting q once we determine m, we can guarantee that
164 	 * there is a complete four-digit quotient at &qspace[1] when
165 	 * we finally stop.
166 	 */
167 	for (m = 4 - n; u[1] == 0; u++)
168 		m--;
169 	for (i = 4 - m; --i >= 0;)
170 		q[i] = 0;
171 	q += 4 - m;
172 
173 	/*
174 	 * Here we run Program D, translated from MIX to C and acquiring
175 	 * a few minor changes.
176 	 *
177 	 * D1: choose multiplier 1 << d to ensure v[1] >= B/2.
178 	 */
179 	d = 0;
180 	for (t = v[1]; t < B / 2; t <<= 1)
181 		d++;
182 	if (d > 0) {
183 		shl(&u[0], m + n, d);		/* u <<= d */
184 		shl(&v[1], n - 1, d);		/* v <<= d */
185 	}
186 	/*
187 	 * D2: j = 0.
188 	 */
189 	j = 0;
190 	v1 = v[1];	/* for D3 -- note that v[1..n] are constant */
191 	v2 = v[2];	/* for D3 */
192 	do {
193 		digit uj0, uj1, uj2;
194 
195 		/*
196 		 * D3: Calculate qhat (\^q, in TeX notation).
197 		 * Let qhat = min((u[j]*B + u[j+1])/v[1], B-1), and
198 		 * let rhat = (u[j]*B + u[j+1]) mod v[1].
199 		 * While rhat < B and v[2]*qhat > rhat*B+u[j+2],
200 		 * decrement qhat and increase rhat correspondingly.
201 		 * Note that if rhat >= B, v[2]*qhat < rhat*B.
202 		 */
203 		uj0 = u[j + 0];	/* for D3 only -- note that u[j+...] change */
204 		uj1 = u[j + 1];	/* for D3 only */
205 		uj2 = u[j + 2];	/* for D3 only */
206 		if (uj0 == v1) {
207 			qhat = B;
208 			rhat = uj1;
209 			goto qhat_too_big;
210 		} else {
211 			u_long n = COMBINE(uj0, uj1);
212 			qhat = n / v1;
213 			rhat = n % v1;
214 		}
215 		while (v2 * qhat > COMBINE(rhat, uj2)) {
216 	qhat_too_big:
217 			qhat--;
218 			if ((rhat += v1) >= B)
219 				break;
220 		}
221 		/*
222 		 * D4: Multiply and subtract.
223 		 * The variable `t' holds any borrows across the loop.
224 		 * We split this up so that we do not require v[0] = 0,
225 		 * and to eliminate a final special case.
226 		 */
227 		for (t = 0, i = n; i > 0; i--) {
228 			t = u[i + j] - v[i] * qhat - t;
229 			u[i + j] = LHALF(t);
230 			t = (B - HHALF(t)) & (B - 1);
231 		}
232 		t = u[j] - t;
233 		u[j] = LHALF(t);
234 		/*
235 		 * D5: test remainder.
236 		 * There is a borrow if and only if HHALF(t) is nonzero;
237 		 * in that (rare) case, qhat was too large (by exactly 1).
238 		 * Fix it by adding v[1..n] to u[j..j+n].
239 		 */
240 		if (HHALF(t)) {
241 			qhat--;
242 			for (t = 0, i = n; i > 0; i--) { /* D6: add back. */
243 				t += u[i + j] + v[i];
244 				u[i + j] = LHALF(t);
245 				t = HHALF(t);
246 			}
247 			u[j] = LHALF(u[j] + t);
248 		}
249 		q[j] = qhat;
250 	} while (++j <= m);		/* D7: loop on j. */
251 
252 	/*
253 	 * If caller wants the remainder, we have to calculate it as
254 	 * u[m..m+n] >> d (this is at most n digits and thus fits in
255 	 * u[m+1..m+n], but we may need more source digits).
256 	 */
257 	if (arq) {
258 		if (d) {
259 			for (i = m + n; i > m; --i)
260 				u[i] = (u[i] >> d) |
261 				    LHALF(u[i - 1] << (HALF_BITS - d));
262 			u[i] = 0;
263 		}
264 		tmp.ul[H] = COMBINE(uspace[1], uspace[2]);
265 		tmp.ul[L] = COMBINE(uspace[3], uspace[4]);
266 		*arq = tmp.q;
267 	}
268 
269 	tmp.ul[H] = COMBINE(qspace[1], qspace[2]);
270 	tmp.ul[L] = COMBINE(qspace[3], qspace[4]);
271 	return (tmp.q);
272 }
273