Coverage Report

Created: 2024-11-25 06:29

/src/gmp/mpn/hgcd2.c
Line
Count
Source (jump to first uncovered line)
1
/* hgcd2.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 1996, 1998, 2000-2004, 2008, 2012, 2019 Free Software Foundation,
8
Inc.
9
10
This file is part of the GNU MP Library.
11
12
The GNU MP Library is free software; you can redistribute it and/or modify
13
it under the terms of either:
14
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.
18
19
or
20
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.
24
25
or both in parallel, as here.
26
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.
31
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/.  */
35
36
#include "gmp-impl.h"
37
#include "longlong.h"
38
39
#include "mpn/generic/hgcd2-div.h"
40
41
#if GMP_NAIL_BITS != 0
42
#error Nails not implemented
43
#endif
44
45
/* Reduces a,b until |a-b| (almost) fits in one limb + 1 bit. Constructs
46
   matrix M. Returns 1 if we make progress, i.e. can perform at least
47
   one subtraction. Otherwise returns zero. */
48
49
/* FIXME: Possible optimizations:
50
51
   The div2 function starts with checking the most significant bit of
52
   the numerator. We can maintained normalized operands here, call
53
   hgcd with normalized operands only, which should make the code
54
   simpler and possibly faster.
55
56
   Experiment with table lookups on the most significant bits.
57
58
   This function is also a candidate for assembler implementation.
59
*/
60
int
61
mpn_hgcd2 (mp_limb_t ah, mp_limb_t al, mp_limb_t bh, mp_limb_t bl,
62
     struct hgcd_matrix1 *M)
63
67.8k
{
64
67.8k
  mp_limb_t u00, u01, u10, u11;
65
66
67.8k
  if (ah < 2 || bh < 2)
67
0
    return 0;
68
69
67.8k
  if (ah > bh || (ah == bh && al > bl))
70
46.6k
    {
71
46.6k
      sub_ddmmss (ah, al, ah, al, bh, bl);
72
46.6k
      if (ah < 2)
73
0
  return 0;
74
75
46.6k
      u00 = u01 = u11 = 1;
76
46.6k
      u10 = 0;
77
46.6k
    }
78
21.2k
  else
79
21.2k
    {
80
21.2k
      sub_ddmmss (bh, bl, bh, bl, ah, al);
81
21.2k
      if (bh < 2)
82
0
  return 0;
83
84
21.2k
      u00 = u10 = u11 = 1;
85
21.2k
      u01 = 0;
86
21.2k
    }
87
88
67.8k
  if (ah < bh)
89
46.6k
    goto subtract_a;
90
91
21.2k
  for (;;)
92
644k
    {
93
644k
      ASSERT (ah >= bh);
94
644k
      if (ah == bh)
95
0
  goto done;
96
97
644k
      if (ah < (CNST_LIMB(1) << (GMP_LIMB_BITS / 2)))
98
33.9k
  {
99
33.9k
    ah = (ah << (GMP_LIMB_BITS / 2) ) + (al >> (GMP_LIMB_BITS / 2));
100
33.9k
    bh = (bh << (GMP_LIMB_BITS / 2) ) + (bl >> (GMP_LIMB_BITS / 2));
101
102
33.9k
    break;
103
33.9k
  }
104
105
      /* Subtract a -= q b, and multiply M from the right by (1 q ; 0
106
   1), affecting the second column of M. */
107
610k
      ASSERT (ah > bh);
108
610k
      sub_ddmmss (ah, al, ah, al, bh, bl);
109
110
610k
      if (ah < 2)
111
0
  goto done;
112
113
610k
      if (ah <= bh)
114
263k
  {
115
    /* Use q = 1 */
116
263k
    u01 += u00;
117
263k
    u11 += u10;
118
263k
  }
119
347k
      else
120
347k
  {
121
347k
    mp_limb_t r[2];
122
347k
    mp_limb_t q = div2 (r, ah, al, bh, bl);
123
347k
    al = r[0]; ah = r[1];
124
347k
    if (ah < 2)
125
0
      {
126
        /* A is too small, but q is correct. */
127
0
        u01 += q * u00;
128
0
        u11 += q * u10;
129
0
        goto done;
130
0
      }
131
347k
    q++;
132
347k
    u01 += q * u00;
133
347k
    u11 += q * u10;
134
347k
  }
135
657k
    subtract_a:
136
657k
      ASSERT (bh >= ah);
137
657k
      if (ah == bh)
138
0
  goto done;
139
140
657k
      if (bh < (CNST_LIMB(1) << (GMP_LIMB_BITS / 2)))
141
33.9k
  {
142
33.9k
    ah = (ah << (GMP_LIMB_BITS / 2) ) + (al >> (GMP_LIMB_BITS / 2));
143
33.9k
    bh = (bh << (GMP_LIMB_BITS / 2) ) + (bl >> (GMP_LIMB_BITS / 2));
144
145
33.9k
    goto subtract_a1;
146
33.9k
  }
147
148
      /* Subtract b -= q a, and multiply M from the right by (1 0 ; q
149
   1), affecting the first column of M. */
150
623k
      sub_ddmmss (bh, bl, bh, bl, ah, al);
151
152
623k
      if (bh < 2)
153
0
  goto done;
154
155
623k
      if (bh <= ah)
156
224k
  {
157
    /* Use q = 1 */
158
224k
    u00 += u01;
159
224k
    u10 += u11;
160
224k
  }
161
398k
      else
162
398k
  {
163
398k
    mp_limb_t r[2];
164
398k
    mp_limb_t q = div2 (r, bh, bl, ah, al);
165
398k
    bl = r[0]; bh = r[1];
166
398k
    if (bh < 2)
167
0
      {
168
        /* B is too small, but q is correct. */
169
0
        u00 += q * u01;
170
0
        u10 += q * u11;
171
0
        goto done;
172
0
      }
173
398k
    q++;
174
398k
    u00 += q * u01;
175
398k
    u10 += q * u11;
176
398k
  }
177
623k
    }
178
179
  /* NOTE: Since we discard the least significant half limb, we don't get a
180
     truly maximal M (corresponding to |a - b| < 2^{GMP_LIMB_BITS +1}). */
181
  /* Single precision loop */
182
33.9k
  for (;;)
183
632k
    {
184
632k
      ASSERT (ah >= bh);
185
186
632k
      ah -= bh;
187
632k
      if (ah < (CNST_LIMB (1) << (GMP_LIMB_BITS / 2 + 1)))
188
29.7k
  break;
189
190
602k
      if (ah <= bh)
191
297k
  {
192
    /* Use q = 1 */
193
297k
    u01 += u00;
194
297k
    u11 += u10;
195
297k
  }
196
305k
      else
197
305k
  {
198
305k
    mp_double_limb_t rq = div1 (ah, bh);
199
305k
    mp_limb_t q = rq.d1;
200
305k
    ah = rq.d0;
201
202
305k
    if (ah < (CNST_LIMB(1) << (GMP_LIMB_BITS / 2 + 1)))
203
16.9k
      {
204
        /* A is too small, but q is correct. */
205
16.9k
        u01 += q * u00;
206
16.9k
        u11 += q * u10;
207
16.9k
        break;
208
16.9k
      }
209
288k
    q++;
210
288k
    u01 += q * u00;
211
288k
    u11 += q * u10;
212
288k
  }
213
619k
    subtract_a1:
214
619k
      ASSERT (bh >= ah);
215
216
619k
      bh -= ah;
217
619k
      if (bh < (CNST_LIMB (1) << (GMP_LIMB_BITS / 2 + 1)))
218
4.24k
  break;
219
220
615k
      if (bh <= ah)
221
220k
  {
222
    /* Use q = 1 */
223
220k
    u00 += u01;
224
220k
    u10 += u11;
225
220k
  }
226
394k
      else
227
394k
  {
228
394k
    mp_double_limb_t rq = div1 (bh, ah);
229
394k
    mp_limb_t q = rq.d1;
230
394k
    bh = rq.d0;
231
232
394k
    if (bh < (CNST_LIMB(1) << (GMP_LIMB_BITS / 2 + 1)))
233
16.9k
      {
234
        /* B is too small, but q is correct. */
235
16.9k
        u00 += q * u01;
236
16.9k
        u10 += q * u11;
237
16.9k
        break;
238
16.9k
      }
239
377k
    q++;
240
377k
    u00 += q * u01;
241
377k
    u10 += q * u11;
242
377k
  }
243
615k
    }
244
245
67.8k
 done:
246
67.8k
  M->u[0][0] = u00; M->u[0][1] = u01;
247
67.8k
  M->u[1][0] = u10; M->u[1][1] = u11;
248
249
67.8k
  return 1;
250
33.9k
}
251
252
/* Sets (r;b) = (a;b) M, with M = (u00, u01; u10, u11). Vector must
253
 * have space for n + 1 limbs. Uses three buffers to avoid a copy*/
254
mp_size_t
255
mpn_hgcd_mul_matrix1_vector (const struct hgcd_matrix1 *M,
256
           mp_ptr rp, mp_srcptr ap, mp_ptr bp, mp_size_t n)
257
67.8k
{
258
67.8k
  mp_limb_t ah, bh;
259
260
  /* Compute (r,b) <-- (u00 a + u10 b, u01 a + u11 b) as
261
262
     r  = u00 * a
263
     r += u10 * b
264
     b *= u11
265
     b += u01 * a
266
  */
267
268
#if HAVE_NATIVE_mpn_addaddmul_1msb0
269
  ah = mpn_addaddmul_1msb0 (rp, ap, bp, n, M->u[0][0], M->u[1][0]);
270
  bh = mpn_addaddmul_1msb0 (bp, bp, ap, n, M->u[1][1], M->u[0][1]);
271
#else
272
67.8k
  ah =     mpn_mul_1 (rp, ap, n, M->u[0][0]);
273
67.8k
  ah += mpn_addmul_1 (rp, bp, n, M->u[1][0]);
274
275
67.8k
  bh =     mpn_mul_1 (bp, bp, n, M->u[1][1]);
276
67.8k
  bh += mpn_addmul_1 (bp, ap, n, M->u[0][1]);
277
67.8k
#endif
278
67.8k
  rp[n] = ah;
279
67.8k
  bp[n] = bh;
280
281
67.8k
  n += (ah | bh) > 0;
282
67.8k
  return n;
283
67.8k
}