xref: /netbsd-src/external/lgpl3/gmp/dist/mpn/generic/hgcd_matrix.c (revision 72c7faa4dbb41dbb0238d6b4a109da0d4b236dd4)
1 /* hgcd_matrix.c.
2 
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.
6 
7 Copyright 2003-2005, 2008, 2012 Free Software Foundation, Inc.
8 
9 This file is part of the GNU MP Library.
10 
11 The GNU MP Library is free software; you can redistribute it and/or modify
12 it under the terms of either:
13 
14   * the GNU Lesser General Public License as published by the Free
15     Software Foundation; either version 3 of the License, or (at your
16     option) any later version.
17 
18 or
19 
20   * the GNU General Public License as published by the Free Software
21     Foundation; either version 2 of the License, or (at your option) any
22     later version.
23 
24 or both in parallel, as here.
25 
26 The GNU MP Library is distributed in the hope that it will be useful, but
27 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
28 or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
29 for more details.
30 
31 You should have received copies of the GNU General Public License and the
32 GNU Lesser General Public License along with the GNU MP Library.  If not,
33 see https://www.gnu.org/licenses/.  */
34 
35 #include "gmp-impl.h"
36 #include "longlong.h"
37 
38 /* For input of size n, matrix elements are of size at most ceil(n/2)
39    - 1, but we need two limbs extra. */
40 void
mpn_hgcd_matrix_init(struct hgcd_matrix * M,mp_size_t n,mp_ptr p)41 mpn_hgcd_matrix_init (struct hgcd_matrix *M, mp_size_t n, mp_ptr p)
42 {
43   mp_size_t s = (n+1)/2 + 1;
44   M->alloc = s;
45   M->n = 1;
46   MPN_ZERO (p, 4 * s);
47   M->p[0][0] = p;
48   M->p[0][1] = p + s;
49   M->p[1][0] = p + 2 * s;
50   M->p[1][1] = p + 3 * s;
51 
52   M->p[0][0][0] = M->p[1][1][0] = 1;
53 }
54 
55 /* Update column COL, adding in Q * column (1-COL). Temporary storage:
56  * qn + n <= M->alloc, where n is the size of the largest element in
57  * column 1 - COL. */
58 void
mpn_hgcd_matrix_update_q(struct hgcd_matrix * M,mp_srcptr qp,mp_size_t qn,unsigned col,mp_ptr tp)59 mpn_hgcd_matrix_update_q (struct hgcd_matrix *M, mp_srcptr qp, mp_size_t qn,
60 			  unsigned col, mp_ptr tp)
61 {
62   ASSERT (col < 2);
63 
64   if (qn == 1)
65     {
66       mp_limb_t q = qp[0];
67       mp_limb_t c0, c1;
68 
69       c0 = mpn_addmul_1 (M->p[0][col], M->p[0][1-col], M->n, q);
70       c1 = mpn_addmul_1 (M->p[1][col], M->p[1][1-col], M->n, q);
71 
72       M->p[0][col][M->n] = c0;
73       M->p[1][col][M->n] = c1;
74 
75       M->n += (c0 | c1) != 0;
76     }
77   else
78     {
79       unsigned row;
80 
81       /* Carries for the unlikely case that we get both high words
82 	 from the multiplication and carries from the addition. */
83       mp_limb_t c[2];
84       mp_size_t n;
85 
86       /* The matrix will not necessarily grow in size by qn, so we
87 	 need normalization in order not to overflow M. */
88 
89       for (n = M->n; n + qn > M->n; n--)
90 	{
91 	  ASSERT (n > 0);
92 	  if (M->p[0][1-col][n-1] > 0 || M->p[1][1-col][n-1] > 0)
93 	    break;
94 	}
95 
96       ASSERT (qn + n <= M->alloc);
97 
98       for (row = 0; row < 2; row++)
99 	{
100 	  if (qn <= n)
101 	    mpn_mul (tp, M->p[row][1-col], n, qp, qn);
102 	  else
103 	    mpn_mul (tp, qp, qn, M->p[row][1-col], n);
104 
105 	  ASSERT (n + qn >= M->n);
106 	  c[row] = mpn_add (M->p[row][col], tp, n + qn, M->p[row][col], M->n);
107 	}
108 
109       n += qn;
110 
111       if (c[0] | c[1])
112 	{
113 	  M->p[0][col][n] = c[0];
114 	  M->p[1][col][n] = c[1];
115 	  n++;
116 	}
117       else
118 	{
119 	  n -= (M->p[0][col][n-1] | M->p[1][col][n-1]) == 0;
120 	  ASSERT (n >= M->n);
121 	}
122       M->n = n;
123     }
124 
125   ASSERT (M->n < M->alloc);
126 }
127 
128 /* Multiply M by M1 from the right. Since the M1 elements fit in
129    GMP_NUMB_BITS - 1 bits, M grows by at most one limb. Needs
130    temporary space M->n */
131 void
mpn_hgcd_matrix_mul_1(struct hgcd_matrix * M,const struct hgcd_matrix1 * M1,mp_ptr tp)132 mpn_hgcd_matrix_mul_1 (struct hgcd_matrix *M, const struct hgcd_matrix1 *M1,
133 		       mp_ptr tp)
134 {
135   mp_size_t n0, n1;
136 
137   /* Could avoid copy by some swapping of pointers. */
138   MPN_COPY (tp, M->p[0][0], M->n);
139   n0 = mpn_hgcd_mul_matrix1_vector (M1, M->p[0][0], tp, M->p[0][1], M->n);
140   MPN_COPY (tp, M->p[1][0], M->n);
141   n1 = mpn_hgcd_mul_matrix1_vector (M1, M->p[1][0], tp, M->p[1][1], M->n);
142 
143   /* Depends on zero initialization */
144   M->n = MAX(n0, n1);
145   ASSERT (M->n < M->alloc);
146 }
147 
148 /* Multiply M by M1 from the right. Needs 3*(M->n + M1->n) + 5 limbs
149    of temporary storage (see mpn_matrix22_mul_itch). */
150 void
mpn_hgcd_matrix_mul(struct hgcd_matrix * M,const struct hgcd_matrix * M1,mp_ptr tp)151 mpn_hgcd_matrix_mul (struct hgcd_matrix *M, const struct hgcd_matrix *M1,
152 		     mp_ptr tp)
153 {
154   mp_size_t n;
155 
156   /* About the new size of M:s elements. Since M1's diagonal elements
157      are > 0, no element can decrease. The new elements are of size
158      M->n + M1->n, one limb more or less. The computation of the
159      matrix product produces elements of size M->n + M1->n + 1. But
160      the true size, after normalization, may be three limbs smaller.
161 
162      The reason that the product has normalized size >= M->n + M1->n -
163      2 is subtle. It depends on the fact that M and M1 can be factored
164      as products of (1,1; 0,1) and (1,0; 1,1), and that we can't have
165      M ending with a large power and M1 starting with a large power of
166      the same matrix. */
167 
168   /* FIXME: Strassen multiplication gives only a small speedup. In FFT
169      multiplication range, this function could be sped up quite a lot
170      using invariance. */
171   ASSERT (M->n + M1->n < M->alloc);
172 
173   ASSERT ((M->p[0][0][M->n-1] | M->p[0][1][M->n-1]
174 	   | M->p[1][0][M->n-1] | M->p[1][1][M->n-1]) > 0);
175 
176   ASSERT ((M1->p[0][0][M1->n-1] | M1->p[0][1][M1->n-1]
177 	   | M1->p[1][0][M1->n-1] | M1->p[1][1][M1->n-1]) > 0);
178 
179   mpn_matrix22_mul (M->p[0][0], M->p[0][1],
180 		    M->p[1][0], M->p[1][1], M->n,
181 		    M1->p[0][0], M1->p[0][1],
182 		    M1->p[1][0], M1->p[1][1], M1->n, tp);
183 
184   /* Index of last potentially non-zero limb, size is one greater. */
185   n = M->n + M1->n;
186 
187   n -= ((M->p[0][0][n] | M->p[0][1][n] | M->p[1][0][n] | M->p[1][1][n]) == 0);
188   n -= ((M->p[0][0][n] | M->p[0][1][n] | M->p[1][0][n] | M->p[1][1][n]) == 0);
189   n -= ((M->p[0][0][n] | M->p[0][1][n] | M->p[1][0][n] | M->p[1][1][n]) == 0);
190 
191   ASSERT ((M->p[0][0][n] | M->p[0][1][n] | M->p[1][0][n] | M->p[1][1][n]) > 0);
192 
193   M->n = n + 1;
194 }
195 
196 /* Multiplies the least significant p limbs of (a;b) by M^-1.
197    Temporary space needed: 2 * (p + M->n)*/
198 mp_size_t
mpn_hgcd_matrix_adjust(const struct hgcd_matrix * M,mp_size_t n,mp_ptr ap,mp_ptr bp,mp_size_t p,mp_ptr tp)199 mpn_hgcd_matrix_adjust (const struct hgcd_matrix *M,
200 			mp_size_t n, mp_ptr ap, mp_ptr bp,
201 			mp_size_t p, mp_ptr tp)
202 {
203   /* M^-1 (a;b) = (r11, -r01; -r10, r00) (a ; b)
204      = (r11 a - r01 b; - r10 a + r00 b */
205 
206   mp_ptr t0 = tp;
207   mp_ptr t1 = tp + p + M->n;
208   mp_limb_t ah, bh;
209   mp_limb_t cy;
210 
211   ASSERT (p + M->n  < n);
212 
213   /* First compute the two values depending on a, before overwriting a */
214 
215   if (M->n >= p)
216     {
217       mpn_mul (t0, M->p[1][1], M->n, ap, p);
218       mpn_mul (t1, M->p[1][0], M->n, ap, p);
219     }
220   else
221     {
222       mpn_mul (t0, ap, p, M->p[1][1], M->n);
223       mpn_mul (t1, ap, p, M->p[1][0], M->n);
224     }
225 
226   /* Update a */
227   MPN_COPY (ap, t0, p);
228   ah = mpn_add (ap + p, ap + p, n - p, t0 + p, M->n);
229 
230   if (M->n >= p)
231     mpn_mul (t0, M->p[0][1], M->n, bp, p);
232   else
233     mpn_mul (t0, bp, p, M->p[0][1], M->n);
234 
235   cy = mpn_sub (ap, ap, n, t0, p + M->n);
236   ASSERT (cy <= ah);
237   ah -= cy;
238 
239   /* Update b */
240   if (M->n >= p)
241     mpn_mul (t0, M->p[0][0], M->n, bp, p);
242   else
243     mpn_mul (t0, bp, p, M->p[0][0], M->n);
244 
245   MPN_COPY (bp, t0, p);
246   bh = mpn_add (bp + p, bp + p, n - p, t0 + p, M->n);
247   cy = mpn_sub (bp, bp, n, t1, p + M->n);
248   ASSERT (cy <= bh);
249   bh -= cy;
250 
251   if (ah > 0 || bh > 0)
252     {
253       ap[n] = ah;
254       bp[n] = bh;
255       n++;
256     }
257   else
258     {
259       /* The subtraction can reduce the size by at most one limb. */
260       if (ap[n-1] == 0 && bp[n-1] == 0)
261 	n--;
262     }
263   ASSERT (ap[n-1] > 0 || bp[n-1] > 0);
264   return n;
265 }
266