beta-0.89.2
[luatex.git] / source / libs / gmp / gmp-src / mpn / generic / jacobi.c
blobbdc3ec67da1bc5a929598201435cedd725c192c3
1 /* jacobi.c
3 THE FUNCTIONS IN THIS FILE ARE INTERNAL WITH MUTABLE INTERFACES. IT IS ONLY
4 SAFE TO REACH THEM THROUGH DOCUMENTED INTERFACES. IN FACT, IT IS ALMOST
5 GUARANTEED THAT THEY'LL CHANGE OR DISAPPEAR IN A FUTURE GNU MP RELEASE.
7 Copyright 1996, 1998, 2000-2004, 2008, 2010, 2011 Free Software Foundation,
8 Inc.
10 This file is part of the GNU MP Library.
12 The GNU MP Library is free software; you can redistribute it and/or modify
13 it under the terms of either:
15 * the GNU Lesser General Public License as published by the Free
16 Software Foundation; either version 3 of the License, or (at your
17 option) any later version.
21 * the GNU General Public License as published by the Free Software
22 Foundation; either version 2 of the License, or (at your option) any
23 later version.
25 or both in parallel, as here.
27 The GNU MP Library is distributed in the hope that it will be useful, but
28 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
29 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
30 for more details.
32 You should have received copies of the GNU General Public License and the
33 GNU Lesser General Public License along with the GNU MP Library. If not,
34 see https://www.gnu.org/licenses/. */
36 #include "gmp.h"
37 #include "gmp-impl.h"
38 #include "longlong.h"
40 #ifndef JACOBI_DC_THRESHOLD
41 #define JACOBI_DC_THRESHOLD GCD_DC_THRESHOLD
42 #endif
44 /* Schönhage's rules:
46 * Assume r0 = r1 q1 + r2, with r0 odd, and r1 = q2 r2 + r3
48 * If r1 is odd, then
50 * (r1 | r0) = s(r1, r0) (r0 | r1) = s(r1, r0) (r2, r1)
52 * where s(x,y) = (-1)^{(x-1)(y-1)/4} = (-1)^[x = y = 3 (mod 4)].
54 * If r1 is even, r2 must be odd. We have
56 * (r1 | r0) = (r1 - r0 | r0) = (-1)^(r0-1)/2 (r0 - r1 | r0)
57 * = (-1)^(r0-1)/2 s(r0, r0 - r1) (r0 | r0 - r1)
58 * = (-1)^(r0-1)/2 s(r0, r0 - r1) (r1 | r0 - r1)
60 * Now, if r1 = 0 (mod 4), then the sign factor is +1, and repeating
61 * q1 times gives
63 * (r1 | r0) = (r1 | r2) = (r3 | r2)
65 * On the other hand, if r1 = 2 (mod 4), the sign factor is
66 * (-1)^{(r0-1)/2}, and repeating q1 times gives the exponent
68 * (r0-1)/2 + (r0-r1-1)/2 + ... + (r0 - (q1-1) r1)/2
69 * = q1 (r0-1)/2 + q1 (q1-1)/2
71 * and we can summarize the even case as
73 * (r1 | r0) = t(r1, r0, q1) (r3 | r2)
75 * where t(x,y,q) = (-1)^{[x = 2 (mod 4)] (q(y-1)/2 + y(q-1)/2)}
77 * What about termination? The remainder sequence ends with (0|1) = 1
78 * (or (0 | r) = 0 if r != 1). What are the possible cases? If r1 is
79 * odd, r2 may be zero. If r1 is even, then r2 = r0 - q1 r1 is odd and
80 * hence non-zero. We may have r3 = r1 - q2 r2 = 0.
82 * Examples: (11|15) = - (15|11) = - (4|11)
83 * (4|11) = (4| 3) = (1| 3)
84 * (1| 3) = (3|1) = (0|1) = 1
86 * (2|7) = (2|1) = (0|1) = 1
88 * Detail: (2|7) = (2-7|7) = (-1|7)(5|7) = -(7|5) = -(2|5)
89 * (2|5) = (2-5|5) = (-1|5)(3|5) = (5|3) = (2|3)
90 * (2|3) = (2-3|3) = (-1|3)(1|3) = -(3|1) = -(2|1)
94 /* In principle, the state consists of four variables: e (one bit), a,
95 b (two bits each), d (one bit). Collected factors are (-1)^e. a and
96 b are the least significant bits of the current remainders. d
97 (denominator) is 0 if we're currently subtracting multiplies of a
98 from b, and 1 if we're subtracting b from a.
100 e is stored in the least significant bit, while a, b and d are
101 coded as only 13 distinct values in bits 1-4, according to the
102 following table. For rows not mentioning d, the value is either
103 implied, or it doesn't matter. */
105 #if WANT_ASSERT
106 static const struct
108 unsigned char a;
109 unsigned char b;
110 } decode_table[13] = {
111 /* 0 */ { 0, 1 },
112 /* 1 */ { 0, 3 },
113 /* 2 */ { 1, 1 },
114 /* 3 */ { 1, 3 },
115 /* 4 */ { 2, 1 },
116 /* 5 */ { 2, 3 },
117 /* 6 */ { 3, 1 },
118 /* 7 */ { 3, 3 }, /* d = 1 */
119 /* 8 */ { 1, 0 },
120 /* 9 */ { 1, 2 },
121 /* 10 */ { 3, 0 },
122 /* 11 */ { 3, 2 },
123 /* 12 */ { 3, 3 }, /* d = 0 */
125 #define JACOBI_A(bits) (decode_table[(bits)>>1].a)
126 #define JACOBI_B(bits) (decode_table[(bits)>>1].b)
127 #endif /* WANT_ASSERT */
129 const unsigned char jacobi_table[208] = {
130 #include "jacobitab.h"
133 #define BITS_FAIL 31
135 static void
136 jacobi_hook (void *p, mp_srcptr gp, mp_size_t gn,
137 mp_srcptr qp, mp_size_t qn, int d)
139 unsigned *bitsp = (unsigned *) p;
141 if (gp)
143 ASSERT (gn > 0);
144 if (gn != 1 || gp[0] != 1)
146 *bitsp = BITS_FAIL;
147 return;
151 if (qp)
153 ASSERT (qn > 0);
154 ASSERT (d >= 0);
155 *bitsp = mpn_jacobi_update (*bitsp, d, qp[0] & 3);
159 #define CHOOSE_P(n) (2*(n) / 3)
162 mpn_jacobi_n (mp_ptr ap, mp_ptr bp, mp_size_t n, unsigned bits)
164 mp_size_t scratch;
165 mp_size_t matrix_scratch;
166 mp_ptr tp;
168 TMP_DECL;
170 ASSERT (n > 0);
171 ASSERT ( (ap[n-1] | bp[n-1]) > 0);
172 ASSERT ( (bp[0] | ap[0]) & 1);
174 /* FIXME: Check for small sizes first, before setting up temporary
175 storage etc. */
176 scratch = MPN_GCD_SUBDIV_STEP_ITCH(n);
178 if (ABOVE_THRESHOLD (n, GCD_DC_THRESHOLD))
180 mp_size_t hgcd_scratch;
181 mp_size_t update_scratch;
182 mp_size_t p = CHOOSE_P (n);
183 mp_size_t dc_scratch;
185 matrix_scratch = MPN_HGCD_MATRIX_INIT_ITCH (n - p);
186 hgcd_scratch = mpn_hgcd_itch (n - p);
187 update_scratch = p + n - 1;
189 dc_scratch = matrix_scratch + MAX(hgcd_scratch, update_scratch);
190 if (dc_scratch > scratch)
191 scratch = dc_scratch;
194 TMP_MARK;
195 tp = TMP_ALLOC_LIMBS(scratch);
197 while (ABOVE_THRESHOLD (n, JACOBI_DC_THRESHOLD))
199 struct hgcd_matrix M;
200 mp_size_t p = 2*n/3;
201 mp_size_t matrix_scratch = MPN_HGCD_MATRIX_INIT_ITCH (n - p);
202 mp_size_t nn;
203 mpn_hgcd_matrix_init (&M, n - p, tp);
205 nn = mpn_hgcd_jacobi (ap + p, bp + p, n - p, &M, &bits,
206 tp + matrix_scratch);
207 if (nn > 0)
209 ASSERT (M.n <= (n - p - 1)/2);
210 ASSERT (M.n + p <= (p + n - 1) / 2);
211 /* Temporary storage 2 (p + M->n) <= p + n - 1. */
212 n = mpn_hgcd_matrix_adjust (&M, p + nn, ap, bp, p, tp + matrix_scratch);
214 else
216 /* Temporary storage n */
217 n = mpn_gcd_subdiv_step (ap, bp, n, 0, jacobi_hook, &bits, tp);
218 if (!n)
220 TMP_FREE;
221 return bits == BITS_FAIL ? 0 : mpn_jacobi_finish (bits);
226 while (n > 2)
228 struct hgcd_matrix1 M;
229 mp_limb_t ah, al, bh, bl;
230 mp_limb_t mask;
232 mask = ap[n-1] | bp[n-1];
233 ASSERT (mask > 0);
235 if (mask & GMP_NUMB_HIGHBIT)
237 ah = ap[n-1]; al = ap[n-2];
238 bh = bp[n-1]; bl = bp[n-2];
240 else
242 int shift;
244 count_leading_zeros (shift, mask);
245 ah = MPN_EXTRACT_NUMB (shift, ap[n-1], ap[n-2]);
246 al = MPN_EXTRACT_NUMB (shift, ap[n-2], ap[n-3]);
247 bh = MPN_EXTRACT_NUMB (shift, bp[n-1], bp[n-2]);
248 bl = MPN_EXTRACT_NUMB (shift, bp[n-2], bp[n-3]);
251 /* Try an mpn_nhgcd2 step */
252 if (mpn_hgcd2_jacobi (ah, al, bh, bl, &M, &bits))
254 n = mpn_matrix22_mul1_inverse_vector (&M, tp, ap, bp, n);
255 MP_PTR_SWAP (ap, tp);
257 else
259 /* mpn_hgcd2 has failed. Then either one of a or b is very
260 small, or the difference is very small. Perform one
261 subtraction followed by one division. */
262 n = mpn_gcd_subdiv_step (ap, bp, n, 0, &jacobi_hook, &bits, tp);
263 if (!n)
265 TMP_FREE;
266 return bits == BITS_FAIL ? 0 : mpn_jacobi_finish (bits);
271 if (bits >= 16)
272 MP_PTR_SWAP (ap, bp);
274 ASSERT (bp[0] & 1);
276 if (n == 1)
278 mp_limb_t al, bl;
279 al = ap[0];
280 bl = bp[0];
282 TMP_FREE;
283 if (bl == 1)
284 return 1 - 2*(bits & 1);
285 else
286 return mpn_jacobi_base (al, bl, bits << 1);
289 else
291 int res = mpn_jacobi_2 (ap, bp, bits & 1);
292 TMP_FREE;
293 return res;