Mercurial > dropbear
comparison bn_mp_toom_mul.c @ 1:22d5cf7d4b1a libtommath
Renaming branch
author | Matt Johnston <matt@ucc.asn.au> |
---|---|
date | Mon, 31 May 2004 18:23:46 +0000 |
parents | |
children | d29b64170cf0 |
comparison
equal
deleted
inserted
replaced
-1:000000000000 | 1:22d5cf7d4b1a |
---|---|
1 /* LibTomMath, multiple-precision integer library -- Tom St Denis | |
2 * | |
3 * LibTomMath is a library that provides multiple-precision | |
4 * integer arithmetic as well as number theoretic functionality. | |
5 * | |
6 * The library was designed directly after the MPI library by | |
7 * Michael Fromberger but has been written from scratch with | |
8 * additional optimizations in place. | |
9 * | |
10 * The library is free for all purposes without any express | |
11 * guarantee it works. | |
12 * | |
13 * Tom St Denis, [email protected], http://math.libtomcrypt.org | |
14 */ | |
15 #include <tommath.h> | |
16 | |
17 /* multiplication using the Toom-Cook 3-way algorithm */ | |
18 int mp_toom_mul(mp_int *a, mp_int *b, mp_int *c) | |
19 { | |
20 mp_int w0, w1, w2, w3, w4, tmp1, tmp2, a0, a1, a2, b0, b1, b2; | |
21 int res, B; | |
22 | |
23 /* init temps */ | |
24 if ((res = mp_init_multi(&w0, &w1, &w2, &w3, &w4, | |
25 &a0, &a1, &a2, &b0, &b1, | |
26 &b2, &tmp1, &tmp2, NULL)) != MP_OKAY) { | |
27 return res; | |
28 } | |
29 | |
30 /* B */ | |
31 B = MIN(a->used, b->used) / 3; | |
32 | |
33 /* a = a2 * B**2 + a1 * B + a0 */ | |
34 if ((res = mp_mod_2d(a, DIGIT_BIT * B, &a0)) != MP_OKAY) { | |
35 goto ERR; | |
36 } | |
37 | |
38 if ((res = mp_copy(a, &a1)) != MP_OKAY) { | |
39 goto ERR; | |
40 } | |
41 mp_rshd(&a1, B); | |
42 mp_mod_2d(&a1, DIGIT_BIT * B, &a1); | |
43 | |
44 if ((res = mp_copy(a, &a2)) != MP_OKAY) { | |
45 goto ERR; | |
46 } | |
47 mp_rshd(&a2, B*2); | |
48 | |
49 /* b = b2 * B**2 + b1 * B + b0 */ | |
50 if ((res = mp_mod_2d(b, DIGIT_BIT * B, &b0)) != MP_OKAY) { | |
51 goto ERR; | |
52 } | |
53 | |
54 if ((res = mp_copy(b, &b1)) != MP_OKAY) { | |
55 goto ERR; | |
56 } | |
57 mp_rshd(&b1, B); | |
58 mp_mod_2d(&b1, DIGIT_BIT * B, &b1); | |
59 | |
60 if ((res = mp_copy(b, &b2)) != MP_OKAY) { | |
61 goto ERR; | |
62 } | |
63 mp_rshd(&b2, B*2); | |
64 | |
65 /* w0 = a0*b0 */ | |
66 if ((res = mp_mul(&a0, &b0, &w0)) != MP_OKAY) { | |
67 goto ERR; | |
68 } | |
69 | |
70 /* w4 = a2 * b2 */ | |
71 if ((res = mp_mul(&a2, &b2, &w4)) != MP_OKAY) { | |
72 goto ERR; | |
73 } | |
74 | |
75 /* w1 = (a2 + 2(a1 + 2a0))(b2 + 2(b1 + 2b0)) */ | |
76 if ((res = mp_mul_2(&a0, &tmp1)) != MP_OKAY) { | |
77 goto ERR; | |
78 } | |
79 if ((res = mp_add(&tmp1, &a1, &tmp1)) != MP_OKAY) { | |
80 goto ERR; | |
81 } | |
82 if ((res = mp_mul_2(&tmp1, &tmp1)) != MP_OKAY) { | |
83 goto ERR; | |
84 } | |
85 if ((res = mp_add(&tmp1, &a2, &tmp1)) != MP_OKAY) { | |
86 goto ERR; | |
87 } | |
88 | |
89 if ((res = mp_mul_2(&b0, &tmp2)) != MP_OKAY) { | |
90 goto ERR; | |
91 } | |
92 if ((res = mp_add(&tmp2, &b1, &tmp2)) != MP_OKAY) { | |
93 goto ERR; | |
94 } | |
95 if ((res = mp_mul_2(&tmp2, &tmp2)) != MP_OKAY) { | |
96 goto ERR; | |
97 } | |
98 if ((res = mp_add(&tmp2, &b2, &tmp2)) != MP_OKAY) { | |
99 goto ERR; | |
100 } | |
101 | |
102 if ((res = mp_mul(&tmp1, &tmp2, &w1)) != MP_OKAY) { | |
103 goto ERR; | |
104 } | |
105 | |
106 /* w3 = (a0 + 2(a1 + 2a2))(b0 + 2(b1 + 2b2)) */ | |
107 if ((res = mp_mul_2(&a2, &tmp1)) != MP_OKAY) { | |
108 goto ERR; | |
109 } | |
110 if ((res = mp_add(&tmp1, &a1, &tmp1)) != MP_OKAY) { | |
111 goto ERR; | |
112 } | |
113 if ((res = mp_mul_2(&tmp1, &tmp1)) != MP_OKAY) { | |
114 goto ERR; | |
115 } | |
116 if ((res = mp_add(&tmp1, &a0, &tmp1)) != MP_OKAY) { | |
117 goto ERR; | |
118 } | |
119 | |
120 if ((res = mp_mul_2(&b2, &tmp2)) != MP_OKAY) { | |
121 goto ERR; | |
122 } | |
123 if ((res = mp_add(&tmp2, &b1, &tmp2)) != MP_OKAY) { | |
124 goto ERR; | |
125 } | |
126 if ((res = mp_mul_2(&tmp2, &tmp2)) != MP_OKAY) { | |
127 goto ERR; | |
128 } | |
129 if ((res = mp_add(&tmp2, &b0, &tmp2)) != MP_OKAY) { | |
130 goto ERR; | |
131 } | |
132 | |
133 if ((res = mp_mul(&tmp1, &tmp2, &w3)) != MP_OKAY) { | |
134 goto ERR; | |
135 } | |
136 | |
137 | |
138 /* w2 = (a2 + a1 + a0)(b2 + b1 + b0) */ | |
139 if ((res = mp_add(&a2, &a1, &tmp1)) != MP_OKAY) { | |
140 goto ERR; | |
141 } | |
142 if ((res = mp_add(&tmp1, &a0, &tmp1)) != MP_OKAY) { | |
143 goto ERR; | |
144 } | |
145 if ((res = mp_add(&b2, &b1, &tmp2)) != MP_OKAY) { | |
146 goto ERR; | |
147 } | |
148 if ((res = mp_add(&tmp2, &b0, &tmp2)) != MP_OKAY) { | |
149 goto ERR; | |
150 } | |
151 if ((res = mp_mul(&tmp1, &tmp2, &w2)) != MP_OKAY) { | |
152 goto ERR; | |
153 } | |
154 | |
155 /* now solve the matrix | |
156 | |
157 0 0 0 0 1 | |
158 1 2 4 8 16 | |
159 1 1 1 1 1 | |
160 16 8 4 2 1 | |
161 1 0 0 0 0 | |
162 | |
163 using 12 subtractions, 4 shifts, | |
164 2 small divisions and 1 small multiplication | |
165 */ | |
166 | |
167 /* r1 - r4 */ | |
168 if ((res = mp_sub(&w1, &w4, &w1)) != MP_OKAY) { | |
169 goto ERR; | |
170 } | |
171 /* r3 - r0 */ | |
172 if ((res = mp_sub(&w3, &w0, &w3)) != MP_OKAY) { | |
173 goto ERR; | |
174 } | |
175 /* r1/2 */ | |
176 if ((res = mp_div_2(&w1, &w1)) != MP_OKAY) { | |
177 goto ERR; | |
178 } | |
179 /* r3/2 */ | |
180 if ((res = mp_div_2(&w3, &w3)) != MP_OKAY) { | |
181 goto ERR; | |
182 } | |
183 /* r2 - r0 - r4 */ | |
184 if ((res = mp_sub(&w2, &w0, &w2)) != MP_OKAY) { | |
185 goto ERR; | |
186 } | |
187 if ((res = mp_sub(&w2, &w4, &w2)) != MP_OKAY) { | |
188 goto ERR; | |
189 } | |
190 /* r1 - r2 */ | |
191 if ((res = mp_sub(&w1, &w2, &w1)) != MP_OKAY) { | |
192 goto ERR; | |
193 } | |
194 /* r3 - r2 */ | |
195 if ((res = mp_sub(&w3, &w2, &w3)) != MP_OKAY) { | |
196 goto ERR; | |
197 } | |
198 /* r1 - 8r0 */ | |
199 if ((res = mp_mul_2d(&w0, 3, &tmp1)) != MP_OKAY) { | |
200 goto ERR; | |
201 } | |
202 if ((res = mp_sub(&w1, &tmp1, &w1)) != MP_OKAY) { | |
203 goto ERR; | |
204 } | |
205 /* r3 - 8r4 */ | |
206 if ((res = mp_mul_2d(&w4, 3, &tmp1)) != MP_OKAY) { | |
207 goto ERR; | |
208 } | |
209 if ((res = mp_sub(&w3, &tmp1, &w3)) != MP_OKAY) { | |
210 goto ERR; | |
211 } | |
212 /* 3r2 - r1 - r3 */ | |
213 if ((res = mp_mul_d(&w2, 3, &w2)) != MP_OKAY) { | |
214 goto ERR; | |
215 } | |
216 if ((res = mp_sub(&w2, &w1, &w2)) != MP_OKAY) { | |
217 goto ERR; | |
218 } | |
219 if ((res = mp_sub(&w2, &w3, &w2)) != MP_OKAY) { | |
220 goto ERR; | |
221 } | |
222 /* r1 - r2 */ | |
223 if ((res = mp_sub(&w1, &w2, &w1)) != MP_OKAY) { | |
224 goto ERR; | |
225 } | |
226 /* r3 - r2 */ | |
227 if ((res = mp_sub(&w3, &w2, &w3)) != MP_OKAY) { | |
228 goto ERR; | |
229 } | |
230 /* r1/3 */ | |
231 if ((res = mp_div_3(&w1, &w1, NULL)) != MP_OKAY) { | |
232 goto ERR; | |
233 } | |
234 /* r3/3 */ | |
235 if ((res = mp_div_3(&w3, &w3, NULL)) != MP_OKAY) { | |
236 goto ERR; | |
237 } | |
238 | |
239 /* at this point shift W[n] by B*n */ | |
240 if ((res = mp_lshd(&w1, 1*B)) != MP_OKAY) { | |
241 goto ERR; | |
242 } | |
243 if ((res = mp_lshd(&w2, 2*B)) != MP_OKAY) { | |
244 goto ERR; | |
245 } | |
246 if ((res = mp_lshd(&w3, 3*B)) != MP_OKAY) { | |
247 goto ERR; | |
248 } | |
249 if ((res = mp_lshd(&w4, 4*B)) != MP_OKAY) { | |
250 goto ERR; | |
251 } | |
252 | |
253 if ((res = mp_add(&w0, &w1, c)) != MP_OKAY) { | |
254 goto ERR; | |
255 } | |
256 if ((res = mp_add(&w2, &w3, &tmp1)) != MP_OKAY) { | |
257 goto ERR; | |
258 } | |
259 if ((res = mp_add(&w4, &tmp1, &tmp1)) != MP_OKAY) { | |
260 goto ERR; | |
261 } | |
262 if ((res = mp_add(&tmp1, c, c)) != MP_OKAY) { | |
263 goto ERR; | |
264 } | |
265 | |
266 ERR: | |
267 mp_clear_multi(&w0, &w1, &w2, &w3, &w4, | |
268 &a0, &a1, &a2, &b0, &b1, | |
269 &b2, &tmp1, &tmp2, NULL); | |
270 return res; | |
271 } | |
272 |