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