1 | /* mpn_divrem -- Divide natural numbers, producing both remainder and |
2 | quotient. |
3 | |
4 | Copyright (C) 1993, 1994, 1995, 1996 Free Software Foundation, Inc. |
5 | |
6 | This file is part of the GNU MP Library. |
7 | |
8 | The GNU MP Library is free software; you can redistribute it and/or modify |
9 | it under the terms of the GNU Lesser General Public License as published by |
10 | the Free Software Foundation; either version 2.1 of the License, or (at your |
11 | option) any later version. |
12 | |
13 | The GNU MP Library is distributed in the hope that it will be useful, but |
14 | WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY |
15 | or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public |
16 | License for more details. |
17 | |
18 | You should have received a copy of the GNU Lesser General Public License |
19 | along with the GNU MP Library; see the file COPYING.LIB. If not, write to |
20 | the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, |
21 | MA 02111-1307, USA. */ |
22 | |
23 | #include <config.h> |
24 | #include "gmp-impl.h" |
25 | |
26 | /* Divide num (NP/NSIZE) by den (DP/DSIZE) and write |
27 | the NSIZE-DSIZE least significant quotient limbs at QP |
28 | and the DSIZE long remainder at NP. If QEXTRA_LIMBS is |
29 | non-zero, generate that many fraction bits and append them after the |
30 | other quotient limbs. |
31 | Return the most significant limb of the quotient, this is always 0 or 1. |
32 | |
33 | Preconditions: |
34 | 0. NSIZE >= DSIZE. |
35 | 1. The most significant bit of the divisor must be set. |
36 | 2. QP must either not overlap with the input operands at all, or |
37 | QP + DSIZE >= NP must hold true. (This means that it's |
38 | possible to put the quotient in the high part of NUM, right after the |
39 | remainder in NUM. |
40 | 3. NSIZE >= DSIZE, even if QEXTRA_LIMBS is non-zero. */ |
41 | |
42 | mp_limb_t |
43 | #if __STDC__ |
44 | mpn_divrem (mp_ptr qp, mp_size_t , |
45 | mp_ptr np, mp_size_t nsize, |
46 | mp_srcptr dp, mp_size_t dsize) |
47 | #else |
48 | mpn_divrem (qp, qextra_limbs, np, nsize, dp, dsize) |
49 | mp_ptr qp; |
50 | mp_size_t qextra_limbs; |
51 | mp_ptr np; |
52 | mp_size_t nsize; |
53 | mp_srcptr dp; |
54 | mp_size_t dsize; |
55 | #endif |
56 | { |
57 | mp_limb_t most_significant_q_limb = 0; |
58 | |
59 | switch (dsize) |
60 | { |
61 | case 0: |
62 | /* We are asked to divide by zero, so go ahead and do it! (To make |
63 | the compiler not remove this statement, return the value.) */ |
64 | return 1 / dsize; |
65 | |
66 | case 1: |
67 | { |
68 | mp_size_t i; |
69 | mp_limb_t n1; |
70 | mp_limb_t d; |
71 | |
72 | d = dp[0]; |
73 | n1 = np[nsize - 1]; |
74 | |
75 | if (n1 >= d) |
76 | { |
77 | n1 -= d; |
78 | most_significant_q_limb = 1; |
79 | } |
80 | |
81 | qp += qextra_limbs; |
82 | for (i = nsize - 2; i >= 0; i--) |
83 | udiv_qrnnd (qp[i], n1, n1, np[i], d); |
84 | qp -= qextra_limbs; |
85 | |
86 | for (i = qextra_limbs - 1; i >= 0; i--) |
87 | udiv_qrnnd (qp[i], n1, n1, 0, d); |
88 | |
89 | np[0] = n1; |
90 | } |
91 | break; |
92 | |
93 | case 2: |
94 | { |
95 | mp_size_t i; |
96 | mp_limb_t n1, n0, n2; |
97 | mp_limb_t d1, d0; |
98 | |
99 | np += nsize - 2; |
100 | d1 = dp[1]; |
101 | d0 = dp[0]; |
102 | n1 = np[1]; |
103 | n0 = np[0]; |
104 | |
105 | if (n1 >= d1 && (n1 > d1 || n0 >= d0)) |
106 | { |
107 | sub_ddmmss (n1, n0, n1, n0, d1, d0); |
108 | most_significant_q_limb = 1; |
109 | } |
110 | |
111 | for (i = qextra_limbs + nsize - 2 - 1; i >= 0; i--) |
112 | { |
113 | mp_limb_t q; |
114 | mp_limb_t r; |
115 | |
116 | if (i >= qextra_limbs) |
117 | np--; |
118 | else |
119 | np[0] = 0; |
120 | |
121 | if (n1 == d1) |
122 | { |
123 | /* Q should be either 111..111 or 111..110. Need special |
124 | treatment of this rare case as normal division would |
125 | give overflow. */ |
126 | q = ~(mp_limb_t) 0; |
127 | |
128 | r = n0 + d1; |
129 | if (r < d1) /* Carry in the addition? */ |
130 | { |
131 | add_ssaaaa (n1, n0, r - d0, np[0], 0, d0); |
132 | qp[i] = q; |
133 | continue; |
134 | } |
135 | n1 = d0 - (d0 != 0); |
136 | n0 = -d0; |
137 | } |
138 | else |
139 | { |
140 | udiv_qrnnd (q, r, n1, n0, d1); |
141 | umul_ppmm (n1, n0, d0, q); |
142 | } |
143 | |
144 | n2 = np[0]; |
145 | q_test: |
146 | if (n1 > r || (n1 == r && n0 > n2)) |
147 | { |
148 | /* The estimated Q was too large. */ |
149 | q--; |
150 | |
151 | sub_ddmmss (n1, n0, n1, n0, 0, d0); |
152 | r += d1; |
153 | if (r >= d1) /* If not carry, test Q again. */ |
154 | goto q_test; |
155 | } |
156 | |
157 | qp[i] = q; |
158 | sub_ddmmss (n1, n0, r, n2, n1, n0); |
159 | } |
160 | np[1] = n1; |
161 | np[0] = n0; |
162 | } |
163 | break; |
164 | |
165 | default: |
166 | { |
167 | mp_size_t i; |
168 | mp_limb_t dX, d1, n0; |
169 | |
170 | np += nsize - dsize; |
171 | dX = dp[dsize - 1]; |
172 | d1 = dp[dsize - 2]; |
173 | n0 = np[dsize - 1]; |
174 | |
175 | if (n0 >= dX) |
176 | { |
177 | if (n0 > dX || mpn_cmp (np, dp, dsize - 1) >= 0) |
178 | { |
179 | mpn_sub_n (np, np, dp, dsize); |
180 | n0 = np[dsize - 1]; |
181 | most_significant_q_limb = 1; |
182 | } |
183 | } |
184 | |
185 | for (i = qextra_limbs + nsize - dsize - 1; i >= 0; i--) |
186 | { |
187 | mp_limb_t q; |
188 | mp_limb_t n1, n2; |
189 | mp_limb_t cy_limb; |
190 | |
191 | if (i >= qextra_limbs) |
192 | { |
193 | np--; |
194 | n2 = np[dsize]; |
195 | } |
196 | else |
197 | { |
198 | n2 = np[dsize - 1]; |
199 | MPN_COPY_DECR (np + 1, np, dsize); |
200 | np[0] = 0; |
201 | } |
202 | |
203 | if (n0 == dX) |
204 | /* This might over-estimate q, but it's probably not worth |
205 | the extra code here to find out. */ |
206 | q = ~(mp_limb_t) 0; |
207 | else |
208 | { |
209 | mp_limb_t r; |
210 | |
211 | udiv_qrnnd (q, r, n0, np[dsize - 1], dX); |
212 | umul_ppmm (n1, n0, d1, q); |
213 | |
214 | while (n1 > r || (n1 == r && n0 > np[dsize - 2])) |
215 | { |
216 | q--; |
217 | r += dX; |
218 | if (r < dX) /* I.e. "carry in previous addition?" */ |
219 | break; |
220 | n1 -= n0 < d1; |
221 | n0 -= d1; |
222 | } |
223 | } |
224 | |
225 | /* Possible optimization: We already have (q * n0) and (1 * n1) |
226 | after the calculation of q. Taking advantage of that, we |
227 | could make this loop make two iterations less. */ |
228 | |
229 | cy_limb = mpn_submul_1 (np, dp, dsize, q); |
230 | |
231 | if (n2 != cy_limb) |
232 | { |
233 | mpn_add_n (np, np, dp, dsize); |
234 | q--; |
235 | } |
236 | |
237 | qp[i] = q; |
238 | n0 = np[dsize - 1]; |
239 | } |
240 | } |
241 | } |
242 | |
243 | return most_significant_q_limb; |
244 | } |
245 | |