2
* Version: MPL 1.1/GPL 2.0/LGPL 2.1
4
* The contents of this file are subject to the Mozilla Public License Version
5
* 1.1 (the "License"); you may not use this file except in compliance with
6
* the License. You may obtain a copy of the License at
7
* http://www.mozilla.org/MPL/
9
* Software distributed under the License is distributed on an "AS IS" basis,
10
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
11
* for the specific language governing rights and limitations under the
14
* The Original Code is the Multi-precision Binary Polynomial Arithmetic
17
* The Initial Developer of the Original Code is Sun Microsystems, Inc.
18
* Portions created by Sun Microsystems, Inc. are Copyright (C) 2003
19
* Sun Microsystems, Inc. All Rights Reserved.
22
* Sheueling Chang Shantz <sheueling.chang@sun.com> and
23
* Douglas Stebila <douglas@stebila.ca> of Sun Laboratories.
25
* Alternatively, the contents of this file may be used under the terms of
26
* either the GNU General Public License Version 2 or later (the "GPL"), or
27
* the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
28
* in which case the provisions of the GPL or the LGPL are applicable instead
29
* of those above. If you wish to allow use of your version of this file only
30
* under the terms of either the GPL or the LGPL, and not to allow others to
31
* use your version of this file under the terms of the MPL, indicate your
32
* decision by deleting the provisions above and replace them with the notice
33
* and other provisions required by the GPL or the LGPL. If you do not delete
34
* the provisions above, a recipient may use your version of this file under
35
* the terms of any one of the MPL, the GPL or the LGPL.
40
#include "mp_gf2m-priv.h"
44
const mp_digit mp_gf2m_sqr_tb[16] =
46
0, 1, 4, 5, 16, 17, 20, 21,
47
64, 65, 68, 69, 80, 81, 84, 85
50
/* Multiply two binary polynomials mp_digits a, b.
51
* Result is a polynomial with degree < 2 * MP_DIGIT_BITS - 1.
52
* Output in two mp_digits rh, rl.
54
#if MP_DIGIT_BITS == 32
56
s_bmul_1x1(mp_digit *rh, mp_digit *rl, const mp_digit a, const mp_digit b)
58
register mp_digit h, l, s;
59
mp_digit tab[8], top2b = a >> 30;
60
register mp_digit a1, a2, a4;
62
a1 = a & (0x3FFFFFFF); a2 = a1 << 1; a4 = a2 << 1;
64
tab[0] = 0; tab[1] = a1; tab[2] = a2; tab[3] = a1^a2;
65
tab[4] = a4; tab[5] = a1^a4; tab[6] = a2^a4; tab[7] = a1^a2^a4;
67
s = tab[b & 0x7]; l = s;
68
s = tab[b >> 3 & 0x7]; l ^= s << 3; h = s >> 29;
69
s = tab[b >> 6 & 0x7]; l ^= s << 6; h ^= s >> 26;
70
s = tab[b >> 9 & 0x7]; l ^= s << 9; h ^= s >> 23;
71
s = tab[b >> 12 & 0x7]; l ^= s << 12; h ^= s >> 20;
72
s = tab[b >> 15 & 0x7]; l ^= s << 15; h ^= s >> 17;
73
s = tab[b >> 18 & 0x7]; l ^= s << 18; h ^= s >> 14;
74
s = tab[b >> 21 & 0x7]; l ^= s << 21; h ^= s >> 11;
75
s = tab[b >> 24 & 0x7]; l ^= s << 24; h ^= s >> 8;
76
s = tab[b >> 27 & 0x7]; l ^= s << 27; h ^= s >> 5;
77
s = tab[b >> 30 ]; l ^= s << 30; h ^= s >> 2;
79
/* compensate for the top two bits of a */
81
if (top2b & 01) { l ^= b << 30; h ^= b >> 2; }
82
if (top2b & 02) { l ^= b << 31; h ^= b >> 1; }
88
s_bmul_1x1(mp_digit *rh, mp_digit *rl, const mp_digit a, const mp_digit b)
90
register mp_digit h, l, s;
91
mp_digit tab[16], top3b = a >> 61;
92
register mp_digit a1, a2, a4, a8;
94
a1 = a & (0x1FFFFFFFFFFFFFFF); a2 = a1 << 1;
95
a4 = a2 << 1; a8 = a4 << 1;
96
tab[ 0] = 0; tab[ 1] = a1; tab[ 2] = a2; tab[ 3] = a1^a2;
97
tab[ 4] = a4; tab[ 5] = a1^a4; tab[ 6] = a2^a4; tab[ 7] = a1^a2^a4;
98
tab[ 8] = a8; tab[ 9] = a1^a8; tab[10] = a2^a8; tab[11] = a1^a2^a8;
99
tab[12] = a4^a8; tab[13] = a1^a4^a8; tab[14] = a2^a4^a8; tab[15] = a1^a2^a4^a8;
101
s = tab[b & 0xF]; l = s;
102
s = tab[b >> 4 & 0xF]; l ^= s << 4; h = s >> 60;
103
s = tab[b >> 8 & 0xF]; l ^= s << 8; h ^= s >> 56;
104
s = tab[b >> 12 & 0xF]; l ^= s << 12; h ^= s >> 52;
105
s = tab[b >> 16 & 0xF]; l ^= s << 16; h ^= s >> 48;
106
s = tab[b >> 20 & 0xF]; l ^= s << 20; h ^= s >> 44;
107
s = tab[b >> 24 & 0xF]; l ^= s << 24; h ^= s >> 40;
108
s = tab[b >> 28 & 0xF]; l ^= s << 28; h ^= s >> 36;
109
s = tab[b >> 32 & 0xF]; l ^= s << 32; h ^= s >> 32;
110
s = tab[b >> 36 & 0xF]; l ^= s << 36; h ^= s >> 28;
111
s = tab[b >> 40 & 0xF]; l ^= s << 40; h ^= s >> 24;
112
s = tab[b >> 44 & 0xF]; l ^= s << 44; h ^= s >> 20;
113
s = tab[b >> 48 & 0xF]; l ^= s << 48; h ^= s >> 16;
114
s = tab[b >> 52 & 0xF]; l ^= s << 52; h ^= s >> 12;
115
s = tab[b >> 56 & 0xF]; l ^= s << 56; h ^= s >> 8;
116
s = tab[b >> 60 ]; l ^= s << 60; h ^= s >> 4;
118
/* compensate for the top three bits of a */
120
if (top3b & 01) { l ^= b << 61; h ^= b >> 3; }
121
if (top3b & 02) { l ^= b << 62; h ^= b >> 2; }
122
if (top3b & 04) { l ^= b << 63; h ^= b >> 1; }
128
/* Compute xor-multiply of two binary polynomials (a1, a0) x (b1, b0)
129
* result is a binary polynomial in 4 mp_digits r[4].
130
* The caller MUST ensure that r has the right amount of space allocated.
133
s_bmul_2x2(mp_digit *r, const mp_digit a1, const mp_digit a0, const mp_digit b1,
137
/* r[3] = h1, r[2] = h0; r[1] = l1; r[0] = l0 */
138
s_bmul_1x1(r+3, r+2, a1, b1);
139
s_bmul_1x1(r+1, r, a0, b0);
140
s_bmul_1x1(&m1, &m0, a0 ^ a1, b0 ^ b1);
141
/* Correction on m1 ^= l1 ^ h1; m0 ^= l0 ^ h0; */
142
r[2] ^= m1 ^ r[1] ^ r[3]; /* h0 ^= m1 ^ l1 ^ h1; */
143
r[1] = r[3] ^ r[2] ^ r[0] ^ m1 ^ m0; /* l1 ^= l0 ^ h0 ^ m0; */
146
/* Compute xor-multiply of two binary polynomials (a2, a1, a0) x (b2, b1, b0)
147
* result is a binary polynomial in 6 mp_digits r[6].
148
* The caller MUST ensure that r has the right amount of space allocated.
151
s_bmul_3x3(mp_digit *r, const mp_digit a2, const mp_digit a1, const mp_digit a0,
152
const mp_digit b2, const mp_digit b1, const mp_digit b0)
156
s_bmul_1x1(r+5, r+4, a2, b2); /* fill top 2 words */
157
s_bmul_2x2(zm, a1, a2^a0, b1, b2^b0); /* fill middle 4 words */
158
s_bmul_2x2(r, a1, a0, b1, b0); /* fill bottom 4 words */
162
zm[1] ^= r[1] ^ r[5];
163
zm[0] ^= r[0] ^ r[4];
171
/* Compute xor-multiply of two binary polynomials (a3, a2, a1, a0) x (b3, b2, b1, b0)
172
* result is a binary polynomial in 8 mp_digits r[8].
173
* The caller MUST ensure that r has the right amount of space allocated.
175
void s_bmul_4x4(mp_digit *r, const mp_digit a3, const mp_digit a2, const mp_digit a1,
176
const mp_digit a0, const mp_digit b3, const mp_digit b2, const mp_digit b1,
181
s_bmul_2x2(r+4, a3, a2, b3, b2); /* fill top 4 words */
182
s_bmul_2x2(zm, a3^a1, a2^a0, b3^b1, b2^b0); /* fill middle 4 words */
183
s_bmul_2x2(r, a1, a0, b1, b0); /* fill bottom 4 words */
185
zm[3] ^= r[3] ^ r[7];
186
zm[2] ^= r[2] ^ r[6];
187
zm[1] ^= r[1] ^ r[5];
188
zm[0] ^= r[0] ^ r[4];
196
/* Compute addition of two binary polynomials a and b,
197
* store result in c; c could be a or b, a and b could be equal;
198
* c is the bitwise XOR of a and b.
201
mp_badd(const mp_int *a, const mp_int *b, mp_int *c)
203
mp_digit *pa, *pb, *pc;
205
mp_size used_pa, used_pb;
206
mp_err res = MP_OKAY;
208
/* Add all digits up to the precision of b. If b had more
209
* precision than a initially, swap a, b first
211
if (MP_USED(a) >= MP_USED(b)) {
214
used_pa = MP_USED(a);
215
used_pb = MP_USED(b);
219
used_pa = MP_USED(b);
220
used_pb = MP_USED(a);
223
/* Make sure c has enough precision for the output value */
224
MP_CHECKOK( s_mp_pad(c, used_pa) );
226
/* Do word-by-word xor */
228
for (ix = 0; ix < used_pb; ix++) {
229
(*pc++) = (*pa++) ^ (*pb++);
232
/* Finish the rest of digits until we're actually done */
233
for (; ix < used_pa; ++ix) {
237
MP_USED(c) = used_pa;
245
#define s_mp_div2(a) MP_CHECKOK( mpl_rsh((a), (a), 1) );
247
/* Compute binary polynomial multiply d = a * b */
249
s_bmul_d(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *d)
251
mp_digit a_i, a0b0, a1b1, carry = 0;
254
s_bmul_1x1(&a1b1, &a0b0, a_i, b);
261
/* Compute binary polynomial xor multiply accumulate d ^= a * b */
263
s_bmul_d_add(const mp_digit *a, mp_size a_len, mp_digit b, mp_digit *d)
265
mp_digit a_i, a0b0, a1b1, carry = 0;
268
s_bmul_1x1(&a1b1, &a0b0, a_i, b);
269
*d++ ^= a0b0 ^ carry;
275
/* Compute binary polynomial xor multiply c = a * b.
276
* All parameters may be identical.
279
mp_bmul(const mp_int *a, const mp_int *b, mp_int *c)
283
mp_size ib, a_used, b_used;
284
mp_err res = MP_OKAY;
288
ARGCHK(a != NULL && b != NULL && c != NULL, MP_BADARG);
291
MP_CHECKOK( mp_init_copy(&tmp, a) );
296
MP_CHECKOK( mp_init_copy(&tmp, b) );
300
if (MP_USED(a) < MP_USED(b)) {
301
const mp_int *xch = b; /* switch a and b if b longer */
306
MP_USED(c) = 1; MP_DIGIT(c, 0) = 0;
307
MP_CHECKOK( s_mp_pad(c, USED(a) + USED(b)) );
310
s_bmul_d(MP_DIGITS(a), MP_USED(a), *pb++, MP_DIGITS(c));
312
/* Outer loop: Digits of b */
315
MP_USED(c) = a_used + b_used;
316
for (ib = 1; ib < b_used; ib++) {
319
/* Inner product: Digits of a */
321
s_bmul_d_add(MP_DIGITS(a), a_used, b_i, MP_DIGITS(c) + ib);
323
MP_DIGIT(c, ib + a_used) = b_i;
336
/* Compute modular reduction of a and store result in r.
338
* For modular arithmetic, the irreducible polynomial f(t) is represented
339
* as an array of int[], where f(t) is of the form:
340
* f(t) = t^p[0] + t^p[1] + ... + t^p[k]
341
* where m = p[0] > p[1] > ... > p[k] = 0.
344
mp_bmod(const mp_int *a, const unsigned int p[], mp_int *r)
348
mp_digit zz, *z, tmp;
350
mp_err res = MP_OKAY;
352
/* The algorithm does the reduction in place in r,
353
* if a != r, copy a into r first so reduction can be done in r
356
MP_CHECKOK( mp_copy(a, r) );
360
/* start reduction */
361
dN = p[0] / MP_DIGIT_BITS;
364
for (j = used - 1; j > dN;) {
372
for (k = 1; p[k] > 0; k++) {
373
/* reducing component t^p[k] */
375
d0 = n % MP_DIGIT_BITS;
376
d1 = MP_DIGIT_BITS - d0;
380
z[j-n-1] ^= (zz<<d1);
383
/* reducing component t^0 */
385
d0 = p[0] % MP_DIGIT_BITS;
386
d1 = MP_DIGIT_BITS - d0;
387
z[j-n] ^= (zz >> d0);
389
z[j-n-1] ^= (zz << d1);
393
/* final round of reduction */
396
d0 = p[0] % MP_DIGIT_BITS;
399
d1 = MP_DIGIT_BITS - d0;
401
/* clear up the top d1 bits */
402
if (d0) z[dN] = (z[dN] << d1) >> d1;
403
*z ^= zz; /* reduction t^0 component */
405
for (k = 1; p[k] > 0; k++) {
406
/* reducing component t^p[k]*/
407
n = p[k] / MP_DIGIT_BITS;
408
d0 = p[k] % MP_DIGIT_BITS;
409
d1 = MP_DIGIT_BITS - d0;
422
/* Compute the product of two polynomials a and b, reduce modulo p,
423
* Store the result in r. r could be a or b; a could be b.
426
mp_bmulmod(const mp_int *a, const mp_int *b, const unsigned int p[], mp_int *r)
430
if (a == b) return mp_bsqrmod(a, p, r);
431
if ((res = mp_bmul(a, b, r) ) != MP_OKAY)
433
return mp_bmod(r, p, r);
436
/* Compute binary polynomial squaring c = a*a mod p .
437
* Parameter r and a can be identical.
441
mp_bsqrmod(const mp_int *a, const unsigned int p[], mp_int *r)
443
mp_digit *pa, *pr, a_i;
448
ARGCHK(a != NULL && r != NULL, MP_BADARG);
452
MP_CHECKOK( mp_init_copy(&tmp, a) );
456
MP_USED(r) = 1; MP_DIGIT(r, 0) = 0;
457
MP_CHECKOK( s_mp_pad(r, 2*USED(a)) );
462
MP_USED(r) = 2 * a_used;
464
for (ia = 0; ia < a_used; ia++) {
466
*pr++ = gf2m_SQR0(a_i);
467
*pr++ = gf2m_SQR1(a_i);
470
MP_CHECKOK( mp_bmod(r, p, r) );
479
/* Compute binary polynomial y/x mod p, y divided by x, reduce modulo p.
480
* Store the result in r. r could be x or y, and x could equal y.
481
* Uses algorithm Modular_Division_GF(2^m) from
482
* Chang-Shantz, S. "From Euclid's GCD to Montgomery Multiplication to
486
mp_bdivmod(const mp_int *y, const mp_int *x, const mp_int *pp,
487
const unsigned int p[], mp_int *r)
490
mp_int *a, *b, *u, *v;
491
mp_err res = MP_OKAY;
497
MP_CHECKOK( mp_init_copy(&aa, x) );
498
MP_CHECKOK( mp_init_copy(&uu, y) );
499
MP_CHECKOK( mp_init_copy(&bb, pp) );
500
MP_CHECKOK( s_mp_pad(r, USED(pp)) );
501
MP_USED(r) = 1; MP_DIGIT(r, 0) = 0;
503
a = &aa; b= &bb; u=&uu; v=r;
504
/* reduce x and y mod p */
505
MP_CHECKOK( mp_bmod(a, p, a) );
506
MP_CHECKOK( mp_bmod(u, p, u) );
508
while (!mp_isodd(a)) {
511
MP_CHECKOK( mp_badd(u, pp, u) );
517
if (mp_cmp_mag(b, a) > 0) {
518
MP_CHECKOK( mp_badd(b, a, b) );
519
MP_CHECKOK( mp_badd(v, u, v) );
523
MP_CHECKOK( mp_badd(v, pp, v) );
526
} while (!mp_isodd(b));
528
else if ((MP_DIGIT(a,0) == 1) && (MP_USED(a) == 1))
531
MP_CHECKOK( mp_badd(a, b, a) );
532
MP_CHECKOK( mp_badd(u, v, u) );
536
MP_CHECKOK( mp_badd(u, pp, u) );
539
} while (!mp_isodd(a));
543
MP_CHECKOK( mp_copy(u, r) );
550
/* Convert the bit-string representation of a polynomial a into an array
551
* of integers corresponding to the bits with non-zero coefficient.
552
* Up to max elements of the array will be filled. Return value is total
553
* number of coefficients that would be extracted if array was large enough.
556
mp_bpoly2arr(const mp_int *a, unsigned int p[], int max)
559
mp_digit top_bit, mask;
562
top_bit <<= MP_DIGIT_BIT - 1;
564
for (k = 0; k < max; k++) p[k] = 0;
567
for (i = MP_USED(a) - 1; i >= 0; i--) {
569
for (j = MP_DIGIT_BIT - 1; j >= 0; j--) {
570
if (MP_DIGITS(a)[i] & mask) {
571
if (k < max) p[k] = MP_DIGIT_BIT * i + j;
581
/* Convert the coefficient array representation of a polynomial to a
582
* bit-string. The array must be terminated by 0.
585
mp_barr2poly(const unsigned int p[], mp_int *a)
588
mp_err res = MP_OKAY;
592
for (i = 0; p[i] > 0; i++) {
593
MP_CHECKOK( mpl_set_bit(a, p[i], 1) );
595
MP_CHECKOK( mpl_set_bit(a, 0, 1) );