xref: /freebsd/sys/libkern/qdivrem.c (revision 783d3ff6d7fae619db8a7990b8a6387de0c677b5)
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 #include <sys/cdefs.h>
37 /*
38  * Multiprecision divide.  This algorithm is from Knuth vol. 2 (2nd ed),
39  * section 4.3.1, pp. 257--259.
40  */
41 
42 #include <libkern/quad.h>
43 
44 #define	B	(1 << HALF_BITS)	/* digit base */
45 
46 /* Combine two `digits' to make a single two-digit number. */
47 #define	COMBINE(a, b) (((u_long)(a) << HALF_BITS) | (b))
48 
49 /* select a type for digits in base B: use unsigned short if they fit */
50 #if ULONG_MAX == 0xffffffff && USHRT_MAX >= 0xffff
51 typedef unsigned short digit;
52 #else
53 typedef u_long digit;
54 #endif
55 
56 /*
57  * Shift p[0]..p[len] left `sh' bits, ignoring any bits that
58  * `fall out' the left (there never will be any such anyway).
59  * We may assume len >= 0.  NOTE THAT THIS WRITES len+1 DIGITS.
60  */
61 static void
62 __shl(digit *p, int len, int sh)
63 {
64 	int i;
65 
66 	for (i = 0; i < len; i++)
67 		p[i] = LHALF(p[i] << sh) | (p[i + 1] >> (HALF_BITS - sh));
68 	p[i] = LHALF(p[i] << sh);
69 }
70 
71 /*
72  * __qdivrem(u, v, rem) returns u/v and, optionally, sets *rem to u%v.
73  *
74  * We do this in base 2-sup-HALF_BITS, so that all intermediate products
75  * fit within u_long.  As a consequence, the maximum length dividend and
76  * divisor are 4 `digits' in this base (they are shorter if they have
77  * leading zeros).
78  */
79 u_quad_t
80 __qdivrem(u_quad_t uq, u_quad_t vq, u_quad_t *arq)
81 {
82 	union uu tmp;
83 	digit *u, *v, *q;
84 	digit v1, v2;
85 	u_long qhat, rhat, t;
86 	int m, n, d, j, i;
87 	digit uspace[5], vspace[5], qspace[5];
88 
89 	/*
90 	 * Take care of special cases: divide by zero, and u < v.
91 	 */
92 	if (__predict_false(vq == 0)) {
93 		/* divide by zero. */
94 		static volatile const unsigned int zero = 0;
95 
96 		tmp.ul[H] = tmp.ul[L] = 1 / zero;
97 		if (arq)
98 			*arq = uq;
99 		return (tmp.q);
100 	}
101 	if (uq < vq) {
102 		if (arq)
103 			*arq = uq;
104 		return (0);
105 	}
106 	u = &uspace[0];
107 	v = &vspace[0];
108 	q = &qspace[0];
109 
110 	/*
111 	 * Break dividend and divisor into digits in base B, then
112 	 * count leading zeros to determine m and n.  When done, we
113 	 * will have:
114 	 *	u = (u[1]u[2]...u[m+n]) sub B
115 	 *	v = (v[1]v[2]...v[n]) sub B
116 	 *	v[1] != 0
117 	 *	1 < n <= 4 (if n = 1, we use a different division algorithm)
118 	 *	m >= 0 (otherwise u < v, which we already checked)
119 	 *	m + n = 4
120 	 * and thus
121 	 *	m = 4 - n <= 2
122 	 */
123 	tmp.uq = uq;
124 	u[0] = 0;
125 	u[1] = HHALF(tmp.ul[H]);
126 	u[2] = LHALF(tmp.ul[H]);
127 	u[3] = HHALF(tmp.ul[L]);
128 	u[4] = LHALF(tmp.ul[L]);
129 	tmp.uq = vq;
130 	v[1] = HHALF(tmp.ul[H]);
131 	v[2] = LHALF(tmp.ul[H]);
132 	v[3] = HHALF(tmp.ul[L]);
133 	v[4] = LHALF(tmp.ul[L]);
134 	for (n = 4; v[1] == 0; v++) {
135 		if (--n == 1) {
136 			u_long rbj;	/* r*B+u[j] (not root boy jim) */
137 			digit q1, q2, q3, q4;
138 
139 			/*
140 			 * Change of plan, per exercise 16.
141 			 *	r = 0;
142 			 *	for j = 1..4:
143 			 *		q[j] = floor((r*B + u[j]) / v),
144 			 *		r = (r*B + u[j]) % v;
145 			 * We unroll this completely here.
146 			 */
147 			t = v[2];	/* nonzero, by definition */
148 			q1 = u[1] / t;
149 			rbj = COMBINE(u[1] % t, u[2]);
150 			q2 = rbj / t;
151 			rbj = COMBINE(rbj % t, u[3]);
152 			q3 = rbj / t;
153 			rbj = COMBINE(rbj % t, u[4]);
154 			q4 = rbj / t;
155 			if (arq)
156 				*arq = rbj % t;
157 			tmp.ul[H] = COMBINE(q1, q2);
158 			tmp.ul[L] = COMBINE(q3, q4);
159 			return (tmp.q);
160 		}
161 	}
162 
163 	/*
164 	 * By adjusting q once we determine m, we can guarantee that
165 	 * there is a complete four-digit quotient at &qspace[1] when
166 	 * we finally stop.
167 	 */
168 	for (m = 4 - n; u[1] == 0; u++)
169 		m--;
170 	for (i = 4 - m; --i >= 0;)
171 		q[i] = 0;
172 	q += 4 - m;
173 
174 	/*
175 	 * Here we run Program D, translated from MIX to C and acquiring
176 	 * a few minor changes.
177 	 *
178 	 * D1: choose multiplier 1 << d to ensure v[1] >= B/2.
179 	 */
180 	d = 0;
181 	for (t = v[1]; t < B / 2; t <<= 1)
182 		d++;
183 	if (d > 0) {
184 		__shl(&u[0], m + n, d);		/* u <<= d */
185 		__shl(&v[1], n - 1, d);		/* v <<= d */
186 	}
187 	/*
188 	 * D2: j = 0.
189 	 */
190 	j = 0;
191 	v1 = v[1];	/* for D3 -- note that v[1..n] are constant */
192 	v2 = v[2];	/* for D3 */
193 	do {
194 		digit uj0, uj1, uj2;
195 
196 		/*
197 		 * D3: Calculate qhat (\^q, in TeX notation).
198 		 * Let qhat = min((u[j]*B + u[j+1])/v[1], B-1), and
199 		 * let rhat = (u[j]*B + u[j+1]) mod v[1].
200 		 * While rhat < B and v[2]*qhat > rhat*B+u[j+2],
201 		 * decrement qhat and increase rhat correspondingly.
202 		 * Note that if rhat >= B, v[2]*qhat < rhat*B.
203 		 */
204 		uj0 = u[j + 0];	/* for D3 only -- note that u[j+...] change */
205 		uj1 = u[j + 1];	/* for D3 only */
206 		uj2 = u[j + 2];	/* for D3 only */
207 		if (uj0 == v1) {
208 			qhat = B;
209 			rhat = uj1;
210 			goto qhat_too_big;
211 		} else {
212 			u_long nn = COMBINE(uj0, uj1);
213 			qhat = nn / v1;
214 			rhat = nn % v1;
215 		}
216 		while (v2 * qhat > COMBINE(rhat, uj2)) {
217 	qhat_too_big:
218 			qhat--;
219 			if ((rhat += v1) >= B)
220 				break;
221 		}
222 		/*
223 		 * D4: Multiply and subtract.
224 		 * The variable `t' holds any borrows across the loop.
225 		 * We split this up so that we do not require v[0] = 0,
226 		 * and to eliminate a final special case.
227 		 */
228 		for (t = 0, i = n; i > 0; i--) {
229 			t = u[i + j] - v[i] * qhat - t;
230 			u[i + j] = LHALF(t);
231 			t = (B - HHALF(t)) & (B - 1);
232 		}
233 		t = u[j] - t;
234 		u[j] = LHALF(t);
235 		/*
236 		 * D5: test remainder.
237 		 * There is a borrow if and only if HHALF(t) is nonzero;
238 		 * in that (rare) case, qhat was too large (by exactly 1).
239 		 * Fix it by adding v[1..n] to u[j..j+n].
240 		 */
241 		if (HHALF(t)) {
242 			qhat--;
243 			for (t = 0, i = n; i > 0; i--) { /* D6: add back. */
244 				t += u[i + j] + v[i];
245 				u[i + j] = LHALF(t);
246 				t = HHALF(t);
247 			}
248 			u[j] = LHALF(u[j] + t);
249 		}
250 		q[j] = qhat;
251 	} while (++j <= m);		/* D7: loop on j. */
252 
253 	/*
254 	 * If caller wants the remainder, we have to calculate it as
255 	 * u[m..m+n] >> d (this is at most n digits and thus fits in
256 	 * u[m+1..m+n], but we may need more source digits).
257 	 */
258 	if (arq) {
259 		if (d) {
260 			for (i = m + n; i > m; --i)
261 				u[i] = (u[i] >> d) |
262 				    LHALF(u[i - 1] << (HALF_BITS - d));
263 			u[i] = 0;
264 		}
265 		tmp.ul[H] = COMBINE(uspace[1], uspace[2]);
266 		tmp.ul[L] = COMBINE(uspace[3], uspace[4]);
267 		*arq = tmp.q;
268 	}
269 
270 	tmp.ul[H] = COMBINE(qspace[1], qspace[2]);
271 	tmp.ul[L] = COMBINE(qspace[3], qspace[4]);
272 	return (tmp.q);
273 }
274