line |
stmt |
bran |
cond |
sub |
pod |
time |
code |
1
|
|
|
|
|
|
|
#include |
2
|
|
|
|
|
|
|
#include |
3
|
|
|
|
|
|
|
#include |
4
|
|
|
|
|
|
|
#include |
5
|
|
|
|
|
|
|
#include |
6
|
|
|
|
|
|
|
|
7
|
|
|
|
|
|
|
/* The AKS primality algorithm for native integers. |
8
|
|
|
|
|
|
|
* |
9
|
|
|
|
|
|
|
* There are three versions here: |
10
|
|
|
|
|
|
|
* V6 The v6 algorithm from the latest AKS paper. |
11
|
|
|
|
|
|
|
* BORNEMANN Improvements from Bernstein, Voloch, and a clever r/s |
12
|
|
|
|
|
|
|
* selection from Folkmar Bornemann. Similar to Bornemann's |
13
|
|
|
|
|
|
|
* 2003 Pari/GP implementation |
14
|
|
|
|
|
|
|
* BERN41 My implementation of theorem 4.1 from Bernstein's 2003 paper. |
15
|
|
|
|
|
|
|
* |
16
|
|
|
|
|
|
|
* Each one is orders of magnitude faster than the previous, and by default |
17
|
|
|
|
|
|
|
* we use Bernstein 4.1 as it is by far the fastest. |
18
|
|
|
|
|
|
|
* |
19
|
|
|
|
|
|
|
* Note that AKS is very, very slow compared to other methods. It is, however, |
20
|
|
|
|
|
|
|
* polynomial in log(N), and log-log performance graphs show nice straight |
21
|
|
|
|
|
|
|
* lines for both implementations. However APR-CL and ECPP both start out |
22
|
|
|
|
|
|
|
* much faster and the slope will be less for any sizes of N that we're |
23
|
|
|
|
|
|
|
* interested in. |
24
|
|
|
|
|
|
|
* |
25
|
|
|
|
|
|
|
* For native 64-bit integers this is purely a coding exercise, as BPSW is |
26
|
|
|
|
|
|
|
* a million times faster and gives proven results. |
27
|
|
|
|
|
|
|
* |
28
|
|
|
|
|
|
|
* |
29
|
|
|
|
|
|
|
* When n < 2^(wordbits/2)-1, we can do a straightforward intermediate: |
30
|
|
|
|
|
|
|
* r = (r + a * b) % n |
31
|
|
|
|
|
|
|
* If n is larger, then these are replaced with: |
32
|
|
|
|
|
|
|
* r = addmod( r, mulmod(a, b, n), n) |
33
|
|
|
|
|
|
|
* which is a lot more work, but keeps us correct. |
34
|
|
|
|
|
|
|
* |
35
|
|
|
|
|
|
|
* Software that does polynomial convolutions followed by a modulo can be |
36
|
|
|
|
|
|
|
* very fast, but will fail when n >= (2^wordbits)/r. |
37
|
|
|
|
|
|
|
* |
38
|
|
|
|
|
|
|
* This is all much easier in GMP. |
39
|
|
|
|
|
|
|
* |
40
|
|
|
|
|
|
|
* Copyright 2012-2016, Dana Jacobsen. |
41
|
|
|
|
|
|
|
*/ |
42
|
|
|
|
|
|
|
|
43
|
|
|
|
|
|
|
#define SQRTN_SHORTCUT 1 |
44
|
|
|
|
|
|
|
|
45
|
|
|
|
|
|
|
#define IMPL_V6 0 /* From the primality_v6 paper */ |
46
|
|
|
|
|
|
|
#define IMPL_BORNEMANN 0 /* From Bornemann's 2002 implementation */ |
47
|
|
|
|
|
|
|
#define IMPL_BERN41 1 /* From Bernstein's early 2003 paper */ |
48
|
|
|
|
|
|
|
|
49
|
|
|
|
|
|
|
#include "ptypes.h" |
50
|
|
|
|
|
|
|
#include "aks.h" |
51
|
|
|
|
|
|
|
#define FUNC_isqrt 1 |
52
|
|
|
|
|
|
|
#define FUNC_gcd_ui 1 |
53
|
|
|
|
|
|
|
#include "util.h" |
54
|
|
|
|
|
|
|
#include "cache.h" |
55
|
|
|
|
|
|
|
#include "mulmod.h" |
56
|
|
|
|
|
|
|
#include "factor.h" |
57
|
|
|
|
|
|
|
|
58
|
|
|
|
|
|
|
#if IMPL_BORNEMANN || IMPL_BERN41 |
59
|
|
|
|
|
|
|
/* We could use lgamma, but it isn't in MSVC and not in pre-C99. The only |
60
|
|
|
|
|
|
|
* sure way to find if it is available is test compilation (ala autoconf). |
61
|
|
|
|
|
|
|
* Instead, we'll just use our own implementation. |
62
|
|
|
|
|
|
|
* See http://mrob.com/pub/ries/lanczos-gamma.html for alternates. */ |
63
|
|
|
|
|
|
|
static double lanczos_coef[8+1] = |
64
|
|
|
|
|
|
|
{ 0.99999999999980993, 676.5203681218851, -1259.1392167224028, |
65
|
|
|
|
|
|
|
771.32342877765313, -176.61502916214059, 12.507343278686905, |
66
|
|
|
|
|
|
|
-0.13857109526572012, 9.9843695780195716e-6, 1.5056327351493116e-7 }; |
67
|
|
|
|
|
|
|
static double log_sqrt_two_pi = 0.91893853320467274178; |
68
|
552
|
|
|
|
|
|
static double log_gamma(double x) |
69
|
|
|
|
|
|
|
{ |
70
|
552
|
|
|
|
|
|
double base = x + 7 + 0.5; |
71
|
552
|
|
|
|
|
|
double sum = 0; |
72
|
|
|
|
|
|
|
int i; |
73
|
4968
|
100
|
|
|
|
|
for (i = 8; i >= 1; i--) |
74
|
4416
|
|
|
|
|
|
sum += lanczos_coef[i] / (x + (double)i); |
75
|
552
|
|
|
|
|
|
sum += lanczos_coef[0]; |
76
|
552
|
|
|
|
|
|
sum = log_sqrt_two_pi + log(sum/x) + ( (x+0.5)*log(base) - base ); |
77
|
552
|
|
|
|
|
|
return sum; |
78
|
|
|
|
|
|
|
} |
79
|
|
|
|
|
|
|
#undef lgamma |
80
|
|
|
|
|
|
|
#define lgamma(x) log_gamma(x) |
81
|
|
|
|
|
|
|
#endif |
82
|
|
|
|
|
|
|
|
83
|
|
|
|
|
|
|
#if IMPL_BERN41 |
84
|
184
|
|
|
|
|
|
static double log_binomial(UV n, UV k) |
85
|
|
|
|
|
|
|
{ |
86
|
184
|
|
|
|
|
|
return log_gamma(n+1) - log_gamma(k+1) - log_gamma(n-k+1); |
87
|
|
|
|
|
|
|
} |
88
|
46
|
|
|
|
|
|
static double log_bern41_binomial(UV r, UV d, UV i, UV j, UV s) |
89
|
|
|
|
|
|
|
{ |
90
|
92
|
|
|
|
|
|
return log_binomial( 2*s, i) |
91
|
46
|
|
|
|
|
|
+ log_binomial( d, i) |
92
|
46
|
|
|
|
|
|
+ log_binomial( 2*s-i, j) |
93
|
46
|
|
|
|
|
|
+ log_binomial( r-2-d, j); |
94
|
|
|
|
|
|
|
} |
95
|
46
|
|
|
|
|
|
static int bern41_acceptable(UV n, UV r, UV s) |
96
|
|
|
|
|
|
|
{ |
97
|
46
|
|
|
|
|
|
double scmp = ceil(sqrt( (r-1)/3.0 )) * log(n); |
98
|
46
|
|
|
|
|
|
UV d = (UV) (0.5 * (r-1)); |
99
|
46
|
|
|
|
|
|
UV i = (UV) (0.475 * (r-1)); |
100
|
46
|
|
|
|
|
|
UV j = i; |
101
|
46
|
50
|
|
|
|
|
if (d > r-2) d = r-2; |
102
|
46
|
50
|
|
|
|
|
if (i > d) i = d; |
103
|
46
|
50
|
|
|
|
|
if (j > (r-2-d)) j = r-2-d; |
104
|
46
|
|
|
|
|
|
return (log_bern41_binomial(r,d,i,j,s) >= scmp); |
105
|
|
|
|
|
|
|
} |
106
|
|
|
|
|
|
|
#endif |
107
|
|
|
|
|
|
|
|
108
|
|
|
|
|
|
|
#if 0 |
109
|
|
|
|
|
|
|
/* Naive znorder. Works well if limit is small. Note arguments. */ |
110
|
|
|
|
|
|
|
static UV order(UV r, UV n, UV limit) { |
111
|
|
|
|
|
|
|
UV j; |
112
|
|
|
|
|
|
|
UV t = 1; |
113
|
|
|
|
|
|
|
for (j = 1; j <= limit; j++) { |
114
|
|
|
|
|
|
|
t = mulmod(t, n, r); |
115
|
|
|
|
|
|
|
if (t == 1) |
116
|
|
|
|
|
|
|
break; |
117
|
|
|
|
|
|
|
} |
118
|
|
|
|
|
|
|
return j; |
119
|
|
|
|
|
|
|
} |
120
|
|
|
|
|
|
|
static void poly_print(UV* poly, UV r) |
121
|
|
|
|
|
|
|
{ |
122
|
|
|
|
|
|
|
int i; |
123
|
|
|
|
|
|
|
for (i = r-1; i >= 1; i--) { |
124
|
|
|
|
|
|
|
if (poly[i] != 0) |
125
|
|
|
|
|
|
|
printf("%lux^%d + ", poly[i], i); |
126
|
|
|
|
|
|
|
} |
127
|
|
|
|
|
|
|
if (poly[0] != 0) printf("%lu", poly[0]); |
128
|
|
|
|
|
|
|
printf("\n"); |
129
|
|
|
|
|
|
|
} |
130
|
|
|
|
|
|
|
#endif |
131
|
|
|
|
|
|
|
|
132
|
0
|
|
|
|
|
|
static void poly_mod_mul(UV* px, UV* py, UV* res, UV r, UV mod) |
133
|
|
|
|
|
|
|
{ |
134
|
|
|
|
|
|
|
UV degpx, degpy; |
135
|
|
|
|
|
|
|
UV i, j, pxi, pyj, rindex; |
136
|
|
|
|
|
|
|
|
137
|
|
|
|
|
|
|
/* Determine max degree of px and py */ |
138
|
0
|
0
|
|
|
|
|
for (degpx = r-1; degpx > 0 && !px[degpx]; degpx--) ; /* */ |
|
|
0
|
|
|
|
|
|
139
|
0
|
0
|
|
|
|
|
for (degpy = r-1; degpy > 0 && !py[degpy]; degpy--) ; /* */ |
|
|
0
|
|
|
|
|
|
140
|
|
|
|
|
|
|
/* We can sum at least j values at once */ |
141
|
0
|
0
|
|
|
|
|
j = (mod >= HALF_WORD) ? 0 : (UV_MAX / ((mod-1)*(mod-1))); |
142
|
|
|
|
|
|
|
|
143
|
0
|
0
|
|
|
|
|
if (j >= degpx || j >= degpy) { |
|
|
0
|
|
|
|
|
|
144
|
|
|
|
|
|
|
/* res will be written completely, so no need to set */ |
145
|
0
|
0
|
|
|
|
|
for (rindex = 0; rindex < r; rindex++) { |
146
|
0
|
|
|
|
|
|
UV sum = 0; |
147
|
0
|
|
|
|
|
|
j = rindex; |
148
|
0
|
0
|
|
|
|
|
for (i = 0; i <= degpx; i++) { |
149
|
0
|
0
|
|
|
|
|
if (j <= degpy) |
150
|
0
|
|
|
|
|
|
sum += px[i] * py[j]; |
151
|
0
|
0
|
|
|
|
|
j = (j == 0) ? r-1 : j-1; |
152
|
|
|
|
|
|
|
} |
153
|
0
|
|
|
|
|
|
res[rindex] = sum % mod; |
154
|
|
|
|
|
|
|
} |
155
|
|
|
|
|
|
|
} else { |
156
|
0
|
|
|
|
|
|
memset(res, 0, r * sizeof(UV)); /* Zero result accumulator */ |
157
|
0
|
0
|
|
|
|
|
for (i = 0; i <= degpx; i++) { |
158
|
0
|
|
|
|
|
|
pxi = px[i]; |
159
|
0
|
0
|
|
|
|
|
if (pxi == 0) continue; |
160
|
0
|
0
|
|
|
|
|
if (mod < HALF_WORD) { |
161
|
0
|
0
|
|
|
|
|
for (j = 0; j <= degpy; j++) { |
162
|
0
|
|
|
|
|
|
pyj = py[j]; |
163
|
0
|
0
|
|
|
|
|
rindex = i+j; if (rindex >= r) rindex -= r; |
164
|
0
|
|
|
|
|
|
res[rindex] = (res[rindex] + (pxi*pyj) ) % mod; |
165
|
|
|
|
|
|
|
} |
166
|
|
|
|
|
|
|
} else { |
167
|
0
|
0
|
|
|
|
|
for (j = 0; j <= degpy; j++) { |
168
|
0
|
|
|
|
|
|
pyj = py[j]; |
169
|
0
|
0
|
|
|
|
|
rindex = i+j; if (rindex >= r) rindex -= r; |
170
|
0
|
|
|
|
|
|
res[rindex] = muladdmod(pxi, pyj, res[rindex], mod); |
171
|
|
|
|
|
|
|
} |
172
|
|
|
|
|
|
|
} |
173
|
|
|
|
|
|
|
} |
174
|
|
|
|
|
|
|
} |
175
|
0
|
|
|
|
|
|
memcpy(px, res, r * sizeof(UV)); /* put result in px */ |
176
|
0
|
|
|
|
|
|
} |
177
|
0
|
|
|
|
|
|
static void poly_mod_sqr(UV* px, UV* res, UV r, UV mod) |
178
|
|
|
|
|
|
|
{ |
179
|
|
|
|
|
|
|
UV c, d, s, sum, rindex, maxpx; |
180
|
0
|
|
|
|
|
|
UV degree = r-1; |
181
|
0
|
|
|
|
|
|
int native_sqr = (mod > isqrt(UV_MAX/(2*r))) ? 0 : 1; |
182
|
|
|
|
|
|
|
|
183
|
0
|
|
|
|
|
|
memset(res, 0, r * sizeof(UV)); /* zero out sums */ |
184
|
|
|
|
|
|
|
/* Discover index of last non-zero value in px */ |
185
|
0
|
0
|
|
|
|
|
for (s = degree; s > 0; s--) |
186
|
0
|
0
|
|
|
|
|
if (px[s] != 0) |
187
|
0
|
|
|
|
|
|
break; |
188
|
0
|
|
|
|
|
|
maxpx = s; |
189
|
|
|
|
|
|
|
/* 1D convolution */ |
190
|
0
|
0
|
|
|
|
|
for (d = 0; d <= 2*degree; d++) { |
191
|
|
|
|
|
|
|
UV *pp1, *pp2, *ppend; |
192
|
0
|
0
|
|
|
|
|
UV s_beg = (d <= degree) ? 0 : d-degree; |
193
|
0
|
|
|
|
|
|
UV s_end = ((d/2) <= maxpx) ? d/2 : maxpx; |
194
|
0
|
0
|
|
|
|
|
if (s_end < s_beg) continue; |
195
|
0
|
|
|
|
|
|
sum = 0; |
196
|
0
|
|
|
|
|
|
pp1 = px + s_beg; |
197
|
0
|
|
|
|
|
|
pp2 = px + d - s_beg; |
198
|
0
|
|
|
|
|
|
ppend = px + s_end; |
199
|
0
|
0
|
|
|
|
|
if (native_sqr) { |
200
|
0
|
0
|
|
|
|
|
while (pp1 < ppend) |
201
|
0
|
|
|
|
|
|
sum += 2 * *pp1++ * *pp2--; |
202
|
|
|
|
|
|
|
/* Special treatment for last point */ |
203
|
0
|
|
|
|
|
|
c = px[s_end]; |
204
|
0
|
0
|
|
|
|
|
sum += (s_end*2 == d) ? c*c : 2*c*px[d-s_end]; |
205
|
0
|
0
|
|
|
|
|
rindex = (d < r) ? d : d-r; /* d % r */ |
206
|
0
|
|
|
|
|
|
res[rindex] = (res[rindex] + sum) % mod; |
207
|
|
|
|
|
|
|
#if HAVE_UINT128 |
208
|
|
|
|
|
|
|
} else { |
209
|
0
|
|
|
|
|
|
uint128_t max = ((uint128_t)1 << 127) - 1; |
210
|
0
|
|
|
|
|
|
uint128_t c128, sum128 = 0; |
211
|
|
|
|
|
|
|
|
212
|
0
|
0
|
|
|
|
|
while (pp1 < ppend) { |
213
|
0
|
|
|
|
|
|
c128 = ((uint128_t)*pp1++) * ((uint128_t)*pp2--); |
214
|
0
|
0
|
|
|
|
|
if (c128 > max) c128 %= mod; |
215
|
0
|
|
|
|
|
|
c128 <<= 1; |
216
|
0
|
0
|
|
|
|
|
if (c128 > max) c128 %= mod; |
217
|
0
|
|
|
|
|
|
sum128 += c128; |
218
|
0
|
0
|
|
|
|
|
if (sum128 > max) sum128 %= mod; |
219
|
|
|
|
|
|
|
} |
220
|
0
|
|
|
|
|
|
c128 = px[s_end]; |
221
|
0
|
0
|
|
|
|
|
if (s_end*2 == d) { |
222
|
0
|
|
|
|
|
|
c128 *= c128; |
223
|
|
|
|
|
|
|
} else { |
224
|
0
|
|
|
|
|
|
c128 *= px[d-s_end]; |
225
|
0
|
0
|
|
|
|
|
if (c128 > max) c128 %= mod; |
226
|
0
|
|
|
|
|
|
c128 <<= 1; |
227
|
|
|
|
|
|
|
} |
228
|
0
|
0
|
|
|
|
|
if (c128 > max) c128 %= mod; |
229
|
0
|
|
|
|
|
|
sum128 += c128; |
230
|
0
|
0
|
|
|
|
|
if (sum128 > max) sum128 %= mod; |
231
|
0
|
0
|
|
|
|
|
rindex = (d < r) ? d : d-r; /* d % r */ |
232
|
0
|
|
|
|
|
|
res[rindex] = ((uint128_t)res[rindex] + sum128) % mod; |
233
|
|
|
|
|
|
|
#else |
234
|
|
|
|
|
|
|
} else { |
235
|
|
|
|
|
|
|
while (pp1 < ppend) { |
236
|
|
|
|
|
|
|
UV p1 = *pp1++; |
237
|
|
|
|
|
|
|
UV p2 = *pp2--; |
238
|
|
|
|
|
|
|
sum = addmod(sum, mulmod(2, mulmod(p1, p2, mod), mod), mod); |
239
|
|
|
|
|
|
|
} |
240
|
|
|
|
|
|
|
c = px[s_end]; |
241
|
|
|
|
|
|
|
if (s_end*2 == d) |
242
|
|
|
|
|
|
|
sum = addmod(sum, sqrmod(c, mod), mod); |
243
|
|
|
|
|
|
|
else |
244
|
|
|
|
|
|
|
sum = addmod(sum, mulmod(2, mulmod(c, px[d-s_end], mod), mod), mod); |
245
|
|
|
|
|
|
|
rindex = (d < r) ? d : d-r; /* d % r */ |
246
|
|
|
|
|
|
|
res[rindex] = addmod(res[rindex], sum, mod); |
247
|
|
|
|
|
|
|
#endif |
248
|
|
|
|
|
|
|
} |
249
|
|
|
|
|
|
|
} |
250
|
0
|
|
|
|
|
|
memcpy(px, res, r * sizeof(UV)); /* put result in px */ |
251
|
0
|
|
|
|
|
|
} |
252
|
|
|
|
|
|
|
|
253
|
0
|
|
|
|
|
|
static UV* poly_mod_pow(UV* pn, UV power, UV r, UV mod) |
254
|
|
|
|
|
|
|
{ |
255
|
|
|
|
|
|
|
UV *res, *temp; |
256
|
|
|
|
|
|
|
|
257
|
0
|
0
|
|
|
|
|
Newz(0, res, r, UV); |
258
|
0
|
0
|
|
|
|
|
New(0, temp, r, UV); |
259
|
0
|
|
|
|
|
|
res[0] = 1; |
260
|
|
|
|
|
|
|
|
261
|
0
|
0
|
|
|
|
|
while (power) { |
262
|
0
|
0
|
|
|
|
|
if (power & 1) poly_mod_mul(res, pn, temp, r, mod); |
263
|
0
|
|
|
|
|
|
power >>= 1; |
264
|
0
|
0
|
|
|
|
|
if (power) poly_mod_sqr(pn, temp, r, mod); |
265
|
|
|
|
|
|
|
} |
266
|
0
|
|
|
|
|
|
Safefree(temp); |
267
|
0
|
|
|
|
|
|
return res; |
268
|
|
|
|
|
|
|
} |
269
|
|
|
|
|
|
|
|
270
|
0
|
|
|
|
|
|
static int test_anr(UV a, UV n, UV r) |
271
|
|
|
|
|
|
|
{ |
272
|
|
|
|
|
|
|
UV* pn; |
273
|
|
|
|
|
|
|
UV* res; |
274
|
|
|
|
|
|
|
UV i; |
275
|
0
|
|
|
|
|
|
int retval = 1; |
276
|
|
|
|
|
|
|
|
277
|
0
|
0
|
|
|
|
|
Newz(0, pn, r, UV); |
278
|
0
|
|
|
|
|
|
a %= r; |
279
|
0
|
|
|
|
|
|
pn[0] = a; |
280
|
0
|
|
|
|
|
|
pn[1] = 1; |
281
|
0
|
|
|
|
|
|
res = poly_mod_pow(pn, n, r, n); |
282
|
0
|
|
|
|
|
|
res[n % r] = addmod(res[n % r], n - 1, n); |
283
|
0
|
|
|
|
|
|
res[0] = addmod(res[0], n - a, n); |
284
|
|
|
|
|
|
|
|
285
|
0
|
0
|
|
|
|
|
for (i = 0; i < r; i++) |
286
|
0
|
0
|
|
|
|
|
if (res[i] != 0) |
287
|
0
|
|
|
|
|
|
retval = 0; |
288
|
0
|
|
|
|
|
|
Safefree(res); |
289
|
0
|
|
|
|
|
|
Safefree(pn); |
290
|
0
|
|
|
|
|
|
return retval; |
291
|
|
|
|
|
|
|
} |
292
|
|
|
|
|
|
|
|
293
|
|
|
|
|
|
|
/* |
294
|
|
|
|
|
|
|
* Avanzi and MihÇilescu, 2007 |
295
|
|
|
|
|
|
|
* http://www.uni-math.gwdg.de/preda/mihailescu-papers/ouraks3.pdf |
296
|
|
|
|
|
|
|
* "As a consequence, one cannot expect the present variants of AKS to |
297
|
|
|
|
|
|
|
* compete with the earlier primality proving methods like ECPP and |
298
|
|
|
|
|
|
|
* cyclotomy." - conclusion regarding memory consumption |
299
|
|
|
|
|
|
|
*/ |
300
|
7
|
|
|
|
|
|
int is_aks_prime(UV n) |
301
|
|
|
|
|
|
|
{ |
302
|
7
|
|
|
|
|
|
UV r, s, a, starta = 1; |
303
|
|
|
|
|
|
|
int verbose; |
304
|
|
|
|
|
|
|
|
305
|
7
|
100
|
|
|
|
|
if (n < 2) |
306
|
2
|
|
|
|
|
|
return 0; |
307
|
5
|
100
|
|
|
|
|
if (n == 2) |
308
|
1
|
|
|
|
|
|
return 1; |
309
|
|
|
|
|
|
|
|
310
|
4
|
50
|
|
|
|
|
if (is_power(n, 0)) |
311
|
0
|
|
|
|
|
|
return 0; |
312
|
|
|
|
|
|
|
|
313
|
4
|
50
|
|
|
|
|
if (n > 11 && ( !(n%2) || !(n%3) || !(n%5) || !(n%7) || !(n%11) )) return 0; |
|
|
50
|
|
|
|
|
|
|
|
50
|
|
|
|
|
|
|
|
50
|
|
|
|
|
|
|
|
50
|
|
|
|
|
|
|
|
50
|
|
|
|
|
|
314
|
|
|
|
|
|
|
/* if (!is_prob_prime(n)) return 0; */ |
315
|
|
|
|
|
|
|
|
316
|
4
|
|
|
|
|
|
verbose = _XS_get_verbose(); |
317
|
|
|
|
|
|
|
#if IMPL_V6 |
318
|
|
|
|
|
|
|
{ |
319
|
|
|
|
|
|
|
UV sqrtn = isqrt(n); |
320
|
|
|
|
|
|
|
double log2n = log(n) / log(2); /* C99 has a log2() function */ |
321
|
|
|
|
|
|
|
UV limit = (UV) floor(log2n * log2n); |
322
|
|
|
|
|
|
|
|
323
|
|
|
|
|
|
|
if (verbose) { printf("# aks limit is %lu\n", (unsigned long) limit); } |
324
|
|
|
|
|
|
|
|
325
|
|
|
|
|
|
|
for (r = 2; r < n; r++) { |
326
|
|
|
|
|
|
|
if ((n % r) == 0) |
327
|
|
|
|
|
|
|
return 0; |
328
|
|
|
|
|
|
|
#if SQRTN_SHORTCUT |
329
|
|
|
|
|
|
|
if (r > sqrtn) |
330
|
|
|
|
|
|
|
return 1; |
331
|
|
|
|
|
|
|
#endif |
332
|
|
|
|
|
|
|
if (znorder(n, r) > limit) |
333
|
|
|
|
|
|
|
break; |
334
|
|
|
|
|
|
|
} |
335
|
|
|
|
|
|
|
|
336
|
|
|
|
|
|
|
if (r >= n) |
337
|
|
|
|
|
|
|
return 1; |
338
|
|
|
|
|
|
|
|
339
|
|
|
|
|
|
|
s = (UV) floor(sqrt(r-1) * log2n); |
340
|
|
|
|
|
|
|
} |
341
|
|
|
|
|
|
|
#endif |
342
|
|
|
|
|
|
|
#if IMPL_BORNEMANN |
343
|
|
|
|
|
|
|
{ |
344
|
|
|
|
|
|
|
UV fac[MPU_MAX_FACTORS+1]; |
345
|
|
|
|
|
|
|
UV slim; |
346
|
|
|
|
|
|
|
double c1, c2, x; |
347
|
|
|
|
|
|
|
double const t = 48; |
348
|
|
|
|
|
|
|
double const t1 = (1.0/((t+1)*log(t+1)-t*log(t))); |
349
|
|
|
|
|
|
|
double const dlogn = log(n); |
350
|
|
|
|
|
|
|
r = next_prime( (UV) (t1*t1 * dlogn*dlogn) ); |
351
|
|
|
|
|
|
|
while (!is_primitive_root(n,r,1)) |
352
|
|
|
|
|
|
|
r = next_prime(r); |
353
|
|
|
|
|
|
|
|
354
|
|
|
|
|
|
|
slim = (UV) (2*t*(r-1)); |
355
|
|
|
|
|
|
|
c1 = lgamma(r-1); |
356
|
|
|
|
|
|
|
c2 = dlogn * floor(sqrt(r)); |
357
|
|
|
|
|
|
|
{ /* Binary search for first s in [1,slim] where x >= 0 */ |
358
|
|
|
|
|
|
|
UV i = 1; |
359
|
|
|
|
|
|
|
UV j = slim; |
360
|
|
|
|
|
|
|
while (i < j) { |
361
|
|
|
|
|
|
|
s = i + (j-i)/2; |
362
|
|
|
|
|
|
|
x = (lgamma(r-1+s) - c1 - lgamma(s+1)) / c2 - 1.0; |
363
|
|
|
|
|
|
|
if (x < 0) i = s+1; |
364
|
|
|
|
|
|
|
else j = s; |
365
|
|
|
|
|
|
|
} |
366
|
|
|
|
|
|
|
s = i-1; |
367
|
|
|
|
|
|
|
} |
368
|
|
|
|
|
|
|
s = (s+3) >> 1; |
369
|
|
|
|
|
|
|
/* Bornemann checks factors up to (s-1)^2, we check to max(r,s) */ |
370
|
|
|
|
|
|
|
/* slim = (s-1)*(s-1); */ |
371
|
|
|
|
|
|
|
slim = (r > s) ? r : s; |
372
|
|
|
|
|
|
|
if (verbose > 1) printf("# aks trial to %lu\n", slim); |
373
|
|
|
|
|
|
|
if (trial_factor(n, fac, 2, slim) > 1) |
374
|
|
|
|
|
|
|
return 0; |
375
|
|
|
|
|
|
|
if (slim >= HALF_WORD || (slim*slim) >= n) |
376
|
|
|
|
|
|
|
return 1; |
377
|
|
|
|
|
|
|
} |
378
|
|
|
|
|
|
|
#endif |
379
|
|
|
|
|
|
|
#if IMPL_BERN41 |
380
|
|
|
|
|
|
|
{ |
381
|
|
|
|
|
|
|
UV slim, fac[MPU_MAX_FACTORS+1]; |
382
|
4
|
|
|
|
|
|
double const log2n = log(n) / log(2); |
383
|
|
|
|
|
|
|
/* Tuning: Initial 'r' selection. Search limit for 's'. */ |
384
|
4
|
50
|
|
|
|
|
double const r0 = ((log2n > 32) ? 0.010 : 0.003) * log2n * log2n; |
385
|
4
|
50
|
|
|
|
|
UV const rmult = (log2n > 32) ? 6 : 30; |
386
|
|
|
|
|
|
|
|
387
|
4
|
100
|
|
|
|
|
r = next_prime(r0 < 2 ? 2 : (UV)r0); /* r must be at least 3 */ |
388
|
20
|
100
|
|
|
|
|
while ( !is_primitive_root(n,r,1) || !bern41_acceptable(n,r,rmult*(r-1)) ) |
|
|
100
|
|
|
|
|
|
389
|
16
|
|
|
|
|
|
r = next_prime(r); |
390
|
|
|
|
|
|
|
|
391
|
|
|
|
|
|
|
{ /* Binary search for first s in [1,slim] where conditions met */ |
392
|
4
|
|
|
|
|
|
UV bi = 1; |
393
|
4
|
|
|
|
|
|
UV bj = rmult * (r-1); |
394
|
38
|
100
|
|
|
|
|
while (bi < bj) { |
395
|
34
|
|
|
|
|
|
s = bi + (bj-bi)/2; |
396
|
34
|
100
|
|
|
|
|
if (!bern41_acceptable(n, r, s)) bi = s+1; |
397
|
23
|
|
|
|
|
|
else bj = s; |
398
|
|
|
|
|
|
|
} |
399
|
4
|
|
|
|
|
|
s = bj; |
400
|
4
|
50
|
|
|
|
|
if (!bern41_acceptable(n, r, s)) croak("AKS: bad s selected"); |
401
|
|
|
|
|
|
|
/* S goes from 2 to s+1 */ |
402
|
4
|
|
|
|
|
|
starta = 2; |
403
|
4
|
|
|
|
|
|
s = s+1; |
404
|
|
|
|
|
|
|
} |
405
|
|
|
|
|
|
|
/* Check divisibility to s * (s-1) to cover both gcd conditions */ |
406
|
4
|
|
|
|
|
|
slim = s * (s-1); |
407
|
4
|
50
|
|
|
|
|
if (verbose > 1) printf("# aks trial to %lu\n", (unsigned long)slim); |
408
|
4
|
100
|
|
|
|
|
if (trial_factor(n, fac, 2, slim) > 1) |
409
|
4
|
|
|
|
|
|
return 0; |
410
|
2
|
50
|
|
|
|
|
if (slim >= HALF_WORD || (slim*slim) >= n) |
|
|
50
|
|
|
|
|
|
411
|
2
|
|
|
|
|
|
return 1; |
412
|
|
|
|
|
|
|
/* Check b^(n-1) = 1 mod n for b in [2..s] */ |
413
|
0
|
0
|
|
|
|
|
for (a = 2; a <= s; a++) { |
414
|
0
|
0
|
|
|
|
|
if (powmod(a, n-1, n) != 1) |
415
|
0
|
|
|
|
|
|
return 0; |
416
|
|
|
|
|
|
|
} |
417
|
|
|
|
|
|
|
} |
418
|
|
|
|
|
|
|
#endif |
419
|
|
|
|
|
|
|
|
420
|
0
|
0
|
|
|
|
|
if (verbose) { printf("# aks r = %lu s = %lu\n", (unsigned long) r, (unsigned long) s); } |
421
|
|
|
|
|
|
|
|
422
|
|
|
|
|
|
|
/* Almost every composite will get recognized by the first test. |
423
|
|
|
|
|
|
|
* However, we need to run 's' tests to have the result proven for all n |
424
|
|
|
|
|
|
|
* based on the theorems we have available at this time. */ |
425
|
0
|
0
|
|
|
|
|
for (a = starta; a <= s; a++) { |
426
|
0
|
0
|
|
|
|
|
if (! test_anr(a, n, r) ) |
427
|
0
|
|
|
|
|
|
return 0; |
428
|
0
|
0
|
|
|
|
|
if (verbose>1) { printf("."); fflush(stdout); } |
429
|
|
|
|
|
|
|
} |
430
|
0
|
0
|
|
|
|
|
if (verbose>1) { printf("\n"); } |
431
|
0
|
|
|
|
|
|
return 1; |
432
|
|
|
|
|
|
|
} |