Coverage Report

Created: 2024-11-25 06:29

/src/gmp/mpn/toom3_sqr.c
Line
Count
Source (jump to first uncovered line)
1
/* mpn_toom3_sqr -- Square {ap,an}.
2
3
   Contributed to the GNU project by Torbjorn Granlund.
4
   Additional improvements by Marco Bodrato.
5
6
   THE FUNCTION IN THIS FILE IS INTERNAL WITH A MUTABLE INTERFACE.  IT IS ONLY
7
   SAFE TO REACH IT THROUGH DOCUMENTED INTERFACES.  IN FACT, IT IS ALMOST
8
   GUARANTEED THAT IT WILL CHANGE OR DISAPPEAR IN A FUTURE GNU MP RELEASE.
9
10
Copyright 2006-2010, 2012, 2015, 2021 Free Software Foundation, Inc.
11
12
This file is part of the GNU MP Library.
13
14
The GNU MP Library is free software; you can redistribute it and/or modify
15
it under the terms of either:
16
17
  * the GNU Lesser General Public License as published by the Free
18
    Software Foundation; either version 3 of the License, or (at your
19
    option) any later version.
20
21
or
22
23
  * the GNU General Public License as published by the Free Software
24
    Foundation; either version 2 of the License, or (at your option) any
25
    later version.
26
27
or both in parallel, as here.
28
29
The GNU MP Library is distributed in the hope that it will be useful, but
30
WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
31
or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
32
for more details.
33
34
You should have received copies of the GNU General Public License and the
35
GNU Lesser General Public License along with the GNU MP Library.  If not,
36
see https://www.gnu.org/licenses/.  */
37
38
39
#include "gmp-impl.h"
40
41
/* Evaluate in: -1, 0, +1, +2, +inf
42
43
  <-s--><--n--><--n-->
44
   ____ ______ ______
45
  |_a2_|___a1_|___a0_|
46
47
  v0  =  a0         ^2 #   A(0)^2
48
  v1  = (a0+ a1+ a2)^2 #   A(1)^2    ah  <= 2
49
  vm1 = (a0- a1+ a2)^2 #  A(-1)^2   |ah| <= 1
50
  v2  = (a0+2a1+4a2)^2 #   A(2)^2    ah  <= 6
51
  vinf=          a2 ^2 # A(inf)^2
52
*/
53
54
#if TUNE_PROGRAM_BUILD || WANT_FAT_BINARY
55
#define MAYBE_sqr_basecase 1
56
#define MAYBE_sqr_toom3   1
57
#else
58
#define MAYBE_sqr_basecase            \
59
8.50M
  (SQR_TOOM3_THRESHOLD < 3 * SQR_TOOM2_THRESHOLD)
60
#define MAYBE_sqr_toom3             \
61
0
  (SQR_TOOM4_THRESHOLD >= 3 * SQR_TOOM3_THRESHOLD)
62
#endif
63
64
#define TOOM3_SQR_REC(p, a, n, ws)          \
65
4.25M
  do {                 \
66
4.25M
    if (MAYBE_sqr_basecase            \
67
4.25M
  && BELOW_THRESHOLD (n, SQR_TOOM2_THRESHOLD))      \
68
4.25M
      mpn_sqr_basecase (p, a, n);         \
69
4.25M
    else if (! MAYBE_sqr_toom3            \
70
0
       || BELOW_THRESHOLD (n, SQR_TOOM3_THRESHOLD))    \
71
0
      mpn_toom2_sqr (p, a, n, ws);         \
72
0
    else                \
73
0
      mpn_toom3_sqr (p, a, n, ws);         \
74
4.25M
  } while (0)
75
76
void
77
mpn_toom3_sqr (mp_ptr pp,
78
         mp_srcptr ap, mp_size_t an,
79
         mp_ptr scratch)
80
850k
{
81
850k
  const int __gmpn_cpuvec_initialized = 1;
82
850k
  mp_size_t n, s;
83
850k
  mp_limb_t cy, vinf0;
84
850k
  mp_ptr gp;
85
850k
  mp_ptr as1, asm1, as2;
86
87
1.70M
#define a0  ap
88
2.55M
#define a1  (ap + n)
89
1.70M
#define a2  (ap + 2*n)
90
91
850k
  n = (an + 2) / (size_t) 3;
92
93
850k
  s = an - 2 * n;
94
95
850k
  ASSERT (0 < s && s <= n);
96
97
850k
  as1 = scratch + 4 * n + 4;
98
850k
  asm1 = scratch + 2 * n + 2;
99
850k
  as2 = pp + n + 1;
100
101
850k
  gp = scratch;
102
103
  /* Compute as1 and asm1.  */
104
850k
  cy = mpn_add (gp, a0, n, a2, s);
105
#if HAVE_NATIVE_mpn_add_n_sub_n
106
  if (cy == 0 && mpn_cmp (gp, a1, n) < 0)
107
    {
108
      cy = mpn_add_n_sub_n (as1, asm1, a1, gp, n);
109
      as1[n] = cy >> 1;
110
      asm1[n] = 0;
111
    }
112
  else
113
    {
114
      mp_limb_t cy2;
115
      cy2 = mpn_add_n_sub_n (as1, asm1, gp, a1, n);
116
      as1[n] = cy + (cy2 >> 1);
117
      asm1[n] = cy - (cy2 & 1);
118
    }
119
#else
120
850k
  as1[n] = cy + mpn_add_n (as1, gp, a1, n);
121
850k
  if (cy == 0 && mpn_cmp (gp, a1, n) < 0)
122
427k
    {
123
427k
      mpn_sub_n (asm1, a1, gp, n);
124
427k
      asm1[n] = 0;
125
427k
    }
126
423k
  else
127
423k
    {
128
423k
      cy -= mpn_sub_n (asm1, gp, a1, n);
129
423k
      asm1[n] = cy;
130
423k
    }
131
850k
#endif
132
133
  /* Compute as2.  */
134
#if HAVE_NATIVE_mpn_rsblsh1_n
135
  cy = mpn_add_n (as2, a2, as1, s);
136
  if (s != n)
137
    cy = mpn_add_1 (as2 + s, as1 + s, n - s, cy);
138
  cy += as1[n];
139
  cy = 2 * cy + mpn_rsblsh1_n (as2, a0, as2, n);
140
#else
141
#if HAVE_NATIVE_mpn_addlsh1_n
142
  cy  = mpn_addlsh1_n (as2, a1, a2, s);
143
  if (s != n)
144
    cy = mpn_add_1 (as2 + s, a1 + s, n - s, cy);
145
  cy = 2 * cy + mpn_addlsh1_n (as2, a0, as2, n);
146
#else
147
850k
  cy = mpn_add_n (as2, a2, as1, s);
148
850k
  if (s != n)
149
850k
    cy = mpn_add_1 (as2 + s, as1 + s, n - s, cy);
150
850k
  cy += as1[n];
151
850k
  cy = 2 * cy + mpn_lshift (as2, as2, n, 1);
152
850k
  cy -= mpn_sub_n (as2, as2, a0, n);
153
850k
#endif
154
850k
#endif
155
850k
  as2[n] = cy;
156
157
850k
  ASSERT (as1[n] <= 2);
158
850k
  ASSERT (asm1[n] <= 1);
159
160
850k
#define v0    pp        /* 2n */
161
850k
#define v1    (pp + 2 * n)      /* 2n+1 */
162
2.55M
#define vinf  (pp + 4 * n)      /* s+s */
163
1.70M
#define vm1   scratch        /* 2n+1 */
164
850k
#define v2    (scratch + 2 * n + 1)    /* 2n+2 */
165
850k
#define scratch_out  (scratch + 5 * n + 5)
166
167
  /* vm1, 2n+1 limbs */
168
#ifdef SMALLER_RECURSION
169
  TOOM3_SQR_REC (vm1, asm1, n, scratch_out);
170
  cy = asm1[n];
171
  if (cy != 0)
172
    {
173
#if HAVE_NATIVE_mpn_addlsh1_n_ip1
174
      cy += mpn_addlsh1_n_ip1 (vm1 + n, asm1, n);
175
#else
176
      cy += mpn_addmul_1 (vm1 + n, asm1, n, CNST_LIMB(2));
177
#endif
178
    }
179
  vm1[2 * n] = cy;
180
#else
181
850k
  vm1[2 * n] = 0;
182
850k
  TOOM3_SQR_REC (vm1, asm1, n + asm1[n], scratch_out);
183
850k
#endif
184
185
850k
  TOOM3_SQR_REC (v2, as2, n + 1, scratch_out); /* v2, 2n+1 limbs */
186
187
850k
  TOOM3_SQR_REC (vinf, a2, s, scratch_out);  /* vinf, s+s limbs */
188
189
850k
  vinf0 = vinf[0];        /* v1 overlaps with this */
190
191
#ifdef SMALLER_RECURSION
192
  /* v1, 2n+1 limbs */
193
  TOOM3_SQR_REC (v1, as1, n, scratch_out);
194
  cy = as1[n];
195
  if (cy == 1)
196
    {
197
#if HAVE_NATIVE_mpn_addlsh1_n_ip1
198
      cy += mpn_addlsh1_n_ip1 (v1 + n, as1, n);
199
#else
200
      cy += mpn_addmul_1 (v1 + n, as1, n, CNST_LIMB(2));
201
#endif
202
    }
203
  else if (cy != 0)
204
    {
205
#if HAVE_NATIVE_mpn_addlsh2_n_ip1
206
      cy = 4 + mpn_addlsh2_n_ip1 (v1 + n, as1, n);
207
#else
208
      cy = 4 + mpn_addmul_1 (v1 + n, as1, n, CNST_LIMB(4));
209
#endif
210
    }
211
  v1[2 * n] = cy;
212
#else
213
850k
  cy = vinf[1];
214
850k
  TOOM3_SQR_REC (v1, as1, n + 1, scratch_out);
215
850k
  vinf[1] = cy;
216
850k
#endif
217
218
850k
  TOOM3_SQR_REC (v0, ap, n, scratch_out);  /* v0, 2n limbs */
219
220
850k
  mpn_toom_interpolate_5pts (pp, v2, vm1, n, s + s, 0, vinf0);
221
850k
}