• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (c) 2008-2016 Stefan Krah. All rights reserved.
3  *
4  * Redistribution and use in source and binary forms, with or without
5  * modification, are permitted provided that the following conditions
6  * are met:
7  *
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  *
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  *
15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS "AS IS" AND
16  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25  * SUCH DAMAGE.
26  */
27 
28 
29 #include "mpdecimal.h"
30 #include <stdio.h>
31 #include <assert.h>
32 #include "numbertheory.h"
33 #include "umodarith.h"
34 #include "crt.h"
35 
36 
37 /* Bignum: Chinese Remainder Theorem, extends the maximum transform length. */
38 
39 
40 /* Multiply P1P2 by v, store result in w. */
41 static inline void
_crt_mulP1P2_3(mpd_uint_t w[3],mpd_uint_t v)42 _crt_mulP1P2_3(mpd_uint_t w[3], mpd_uint_t v)
43 {
44     mpd_uint_t hi1, hi2, lo;
45 
46     _mpd_mul_words(&hi1, &lo, LH_P1P2, v);
47     w[0] = lo;
48 
49     _mpd_mul_words(&hi2, &lo, UH_P1P2, v);
50     lo = hi1 + lo;
51     if (lo < hi1) hi2++;
52 
53     w[1] = lo;
54     w[2] = hi2;
55 }
56 
57 /* Add 3 words from v to w. The result is known to fit in w. */
58 static inline void
_crt_add3(mpd_uint_t w[3],mpd_uint_t v[3])59 _crt_add3(mpd_uint_t w[3], mpd_uint_t v[3])
60 {
61     mpd_uint_t carry;
62     mpd_uint_t s;
63 
64     s = w[0] + v[0];
65     carry = (s < w[0]);
66     w[0] = s;
67 
68     s = w[1] + (v[1] + carry);
69     carry = (s < w[1]);
70     w[1] = s;
71 
72     w[2] = w[2] + (v[2] + carry);
73 }
74 
75 /* Divide 3 words in u by v, store result in w, return remainder. */
76 static inline mpd_uint_t
_crt_div3(mpd_uint_t * w,const mpd_uint_t * u,mpd_uint_t v)77 _crt_div3(mpd_uint_t *w, const mpd_uint_t *u, mpd_uint_t v)
78 {
79     mpd_uint_t r1 = u[2];
80     mpd_uint_t r2;
81 
82     if (r1 < v) {
83         w[2] = 0;
84     }
85     else {
86         _mpd_div_word(&w[2], &r1, u[2], v); /* GCOV_NOT_REACHED */
87     }
88 
89     _mpd_div_words(&w[1], &r2, r1, u[1], v);
90     _mpd_div_words(&w[0], &r1, r2, u[0], v);
91 
92     return r1;
93 }
94 
95 
96 /*
97  * Chinese Remainder Theorem:
98  * Algorithm from Joerg Arndt, "Matters Computational",
99  * Chapter 37.4.1 [http://www.jjj.de/fxt/]
100  *
101  * See also Knuth, TAOCP, Volume 2, 4.3.2, exercise 7.
102  */
103 
104 /*
105  * CRT with carry: x1, x2, x3 contain numbers modulo p1, p2, p3. For each
106  * triple of members of the arrays, find the unique z modulo p1*p2*p3, with
107  * zmax = p1*p2*p3 - 1.
108  *
109  * In each iteration of the loop, split z into result[i] = z % MPD_RADIX
110  * and carry = z / MPD_RADIX. Let N be the size of carry[] and cmax the
111  * maximum carry.
112  *
113  * Limits for the 32-bit build:
114  *
115  *   N    = 2**96
116  *   cmax = 7711435591312380274
117  *
118  * Limits for the 64 bit build:
119  *
120  *   N    = 2**192
121  *   cmax = 627710135393475385904124401220046371710
122  *
123  * The following statements hold for both versions:
124  *
125  *   1) cmax + zmax < N, so the addition does not overflow.
126  *
127  *   2) (cmax + zmax) / MPD_RADIX == cmax.
128  *
129  *   3) If c <= cmax, then c_next = (c + zmax) / MPD_RADIX <= cmax.
130  */
131 void
crt3(mpd_uint_t * x1,mpd_uint_t * x2,mpd_uint_t * x3,mpd_size_t rsize)132 crt3(mpd_uint_t *x1, mpd_uint_t *x2, mpd_uint_t *x3, mpd_size_t rsize)
133 {
134     mpd_uint_t p1 = mpd_moduli[P1];
135     mpd_uint_t umod;
136 #ifdef PPRO
137     double dmod;
138     uint32_t dinvmod[3];
139 #endif
140     mpd_uint_t a1, a2, a3;
141     mpd_uint_t s;
142     mpd_uint_t z[3], t[3];
143     mpd_uint_t carry[3] = {0,0,0};
144     mpd_uint_t hi, lo;
145     mpd_size_t i;
146 
147     for (i = 0; i < rsize; i++) {
148 
149         a1 = x1[i];
150         a2 = x2[i];
151         a3 = x3[i];
152 
153         SETMODULUS(P2);
154         s = ext_submod(a2, a1, umod);
155         s = MULMOD(s, INV_P1_MOD_P2);
156 
157         _mpd_mul_words(&hi, &lo, s, p1);
158         lo = lo + a1;
159         if (lo < a1) hi++;
160 
161         SETMODULUS(P3);
162         s = dw_submod(a3, hi, lo, umod);
163         s = MULMOD(s, INV_P1P2_MOD_P3);
164 
165         z[0] = lo;
166         z[1] = hi;
167         z[2] = 0;
168 
169         _crt_mulP1P2_3(t, s);
170         _crt_add3(z, t);
171         _crt_add3(carry, z);
172 
173         x1[i] = _crt_div3(carry, carry, MPD_RADIX);
174     }
175 
176     assert(carry[0] == 0 && carry[1] == 0 && carry[2] == 0);
177 }
178 
179 
180