comparison libtommath/bn_mp_exptmod_fast.c @ 284:eed26cff980b

propagate from branch 'au.asn.ucc.matt.ltm.dropbear' (head 6c790cad5a7fa866ad062cb3a0c279f7ba788583) to branch 'au.asn.ucc.matt.dropbear' (head fff0894a0399405a9410ea1c6d118f342cf2aa64)
author Matt Johnston <matt@ucc.asn.au>
date Wed, 08 Mar 2006 13:23:49 +0000
parents
children cd14c94fe89c 5ff8218bcee9
comparison
equal deleted inserted replaced
283:bd240aa12ba7 284:eed26cff980b
1 #include <tommath.h>
2 #ifdef BN_MP_EXPTMOD_FAST_C
3 /* LibTomMath, multiple-precision integer library -- Tom St Denis
4 *
5 * LibTomMath is a library that provides multiple-precision
6 * integer arithmetic as well as number theoretic functionality.
7 *
8 * The library was designed directly after the MPI library by
9 * Michael Fromberger but has been written from scratch with
10 * additional optimizations in place.
11 *
12 * The library is free for all purposes without any express
13 * guarantee it works.
14 *
15 * Tom St Denis, [email protected], http://math.libtomcrypt.org
16 */
17
18 /* computes Y == G**X mod P, HAC pp.616, Algorithm 14.85
19 *
20 * Uses a left-to-right k-ary sliding window to compute the modular exponentiation.
21 * The value of k changes based on the size of the exponent.
22 *
23 * Uses Montgomery or Diminished Radix reduction [whichever appropriate]
24 */
25
26 #ifdef MP_LOW_MEM
27 #define TAB_SIZE 32
28 #else
29 #define TAB_SIZE 256
30 #endif
31
32 int mp_exptmod_fast (mp_int * G, mp_int * X, mp_int * P, mp_int * Y, int redmode)
33 {
34 mp_int M[TAB_SIZE], res;
35 mp_digit buf, mp;
36 int err, bitbuf, bitcpy, bitcnt, mode, digidx, x, y, winsize;
37
38 /* use a pointer to the reduction algorithm. This allows us to use
39 * one of many reduction algorithms without modding the guts of
40 * the code with if statements everywhere.
41 */
42 int (*redux)(mp_int*,mp_int*,mp_digit);
43
44 /* find window size */
45 x = mp_count_bits (X);
46 if (x <= 7) {
47 winsize = 2;
48 } else if (x <= 36) {
49 winsize = 3;
50 } else if (x <= 140) {
51 winsize = 4;
52 } else if (x <= 450) {
53 winsize = 5;
54 } else if (x <= 1303) {
55 winsize = 6;
56 } else if (x <= 3529) {
57 winsize = 7;
58 } else {
59 winsize = 8;
60 }
61
62 #ifdef MP_LOW_MEM
63 if (winsize > 5) {
64 winsize = 5;
65 }
66 #endif
67
68 /* init M array */
69 /* init first cell */
70 if ((err = mp_init(&M[1])) != MP_OKAY) {
71 return err;
72 }
73
74 /* now init the second half of the array */
75 for (x = 1<<(winsize-1); x < (1 << winsize); x++) {
76 if ((err = mp_init(&M[x])) != MP_OKAY) {
77 for (y = 1<<(winsize-1); y < x; y++) {
78 mp_clear (&M[y]);
79 }
80 mp_clear(&M[1]);
81 return err;
82 }
83 }
84
85 /* determine and setup reduction code */
86 if (redmode == 0) {
87 #ifdef BN_MP_MONTGOMERY_SETUP_C
88 /* now setup montgomery */
89 if ((err = mp_montgomery_setup (P, &mp)) != MP_OKAY) {
90 goto LBL_M;
91 }
92 #else
93 err = MP_VAL;
94 goto LBL_M;
95 #endif
96
97 /* automatically pick the comba one if available (saves quite a few calls/ifs) */
98 #ifdef BN_FAST_MP_MONTGOMERY_REDUCE_C
99 if (((P->used * 2 + 1) < MP_WARRAY) &&
100 P->used < (1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) {
101 redux = fast_mp_montgomery_reduce;
102 } else
103 #endif
104 {
105 #ifdef BN_MP_MONTGOMERY_REDUCE_C
106 /* use slower baseline Montgomery method */
107 redux = mp_montgomery_reduce;
108 #else
109 err = MP_VAL;
110 goto LBL_M;
111 #endif
112 }
113 } else if (redmode == 1) {
114 #if defined(BN_MP_DR_SETUP_C) && defined(BN_MP_DR_REDUCE_C)
115 /* setup DR reduction for moduli of the form B**k - b */
116 mp_dr_setup(P, &mp);
117 redux = mp_dr_reduce;
118 #else
119 err = MP_VAL;
120 goto LBL_M;
121 #endif
122 } else {
123 #if defined(BN_MP_REDUCE_2K_SETUP_C) && defined(BN_MP_REDUCE_2K_C)
124 /* setup DR reduction for moduli of the form 2**k - b */
125 if ((err = mp_reduce_2k_setup(P, &mp)) != MP_OKAY) {
126 goto LBL_M;
127 }
128 redux = mp_reduce_2k;
129 #else
130 err = MP_VAL;
131 goto LBL_M;
132 #endif
133 }
134
135 /* setup result */
136 if ((err = mp_init (&res)) != MP_OKAY) {
137 goto LBL_M;
138 }
139
140 /* create M table
141 *
142
143 *
144 * The first half of the table is not computed though accept for M[0] and M[1]
145 */
146
147 if (redmode == 0) {
148 #ifdef BN_MP_MONTGOMERY_CALC_NORMALIZATION_C
149 /* now we need R mod m */
150 if ((err = mp_montgomery_calc_normalization (&res, P)) != MP_OKAY) {
151 goto LBL_RES;
152 }
153 #else
154 err = MP_VAL;
155 goto LBL_RES;
156 #endif
157
158 /* now set M[1] to G * R mod m */
159 if ((err = mp_mulmod (G, &res, P, &M[1])) != MP_OKAY) {
160 goto LBL_RES;
161 }
162 } else {
163 mp_set(&res, 1);
164 if ((err = mp_mod(G, P, &M[1])) != MP_OKAY) {
165 goto LBL_RES;
166 }
167 }
168
169 /* compute the value at M[1<<(winsize-1)] by squaring M[1] (winsize-1) times */
170 if ((err = mp_copy (&M[1], &M[1 << (winsize - 1)])) != MP_OKAY) {
171 goto LBL_RES;
172 }
173
174 for (x = 0; x < (winsize - 1); x++) {
175 if ((err = mp_sqr (&M[1 << (winsize - 1)], &M[1 << (winsize - 1)])) != MP_OKAY) {
176 goto LBL_RES;
177 }
178 if ((err = redux (&M[1 << (winsize - 1)], P, mp)) != MP_OKAY) {
179 goto LBL_RES;
180 }
181 }
182
183 /* create upper table */
184 for (x = (1 << (winsize - 1)) + 1; x < (1 << winsize); x++) {
185 if ((err = mp_mul (&M[x - 1], &M[1], &M[x])) != MP_OKAY) {
186 goto LBL_RES;
187 }
188 if ((err = redux (&M[x], P, mp)) != MP_OKAY) {
189 goto LBL_RES;
190 }
191 }
192
193 /* set initial mode and bit cnt */
194 mode = 0;
195 bitcnt = 1;
196 buf = 0;
197 digidx = X->used - 1;
198 bitcpy = 0;
199 bitbuf = 0;
200
201 for (;;) {
202 /* grab next digit as required */
203 if (--bitcnt == 0) {
204 /* if digidx == -1 we are out of digits so break */
205 if (digidx == -1) {
206 break;
207 }
208 /* read next digit and reset bitcnt */
209 buf = X->dp[digidx--];
210 bitcnt = (int)DIGIT_BIT;
211 }
212
213 /* grab the next msb from the exponent */
214 y = (mp_digit)(buf >> (DIGIT_BIT - 1)) & 1;
215 buf <<= (mp_digit)1;
216
217 /* if the bit is zero and mode == 0 then we ignore it
218 * These represent the leading zero bits before the first 1 bit
219 * in the exponent. Technically this opt is not required but it
220 * does lower the # of trivial squaring/reductions used
221 */
222 if (mode == 0 && y == 0) {
223 continue;
224 }
225
226 /* if the bit is zero and mode == 1 then we square */
227 if (mode == 1 && y == 0) {
228 if ((err = mp_sqr (&res, &res)) != MP_OKAY) {
229 goto LBL_RES;
230 }
231 if ((err = redux (&res, P, mp)) != MP_OKAY) {
232 goto LBL_RES;
233 }
234 continue;
235 }
236
237 /* else we add it to the window */
238 bitbuf |= (y << (winsize - ++bitcpy));
239 mode = 2;
240
241 if (bitcpy == winsize) {
242 /* ok window is filled so square as required and multiply */
243 /* square first */
244 for (x = 0; x < winsize; x++) {
245 if ((err = mp_sqr (&res, &res)) != MP_OKAY) {
246 goto LBL_RES;
247 }
248 if ((err = redux (&res, P, mp)) != MP_OKAY) {
249 goto LBL_RES;
250 }
251 }
252
253 /* then multiply */
254 if ((err = mp_mul (&res, &M[bitbuf], &res)) != MP_OKAY) {
255 goto LBL_RES;
256 }
257 if ((err = redux (&res, P, mp)) != MP_OKAY) {
258 goto LBL_RES;
259 }
260
261 /* empty window and reset */
262 bitcpy = 0;
263 bitbuf = 0;
264 mode = 1;
265 }
266 }
267
268 /* if bits remain then square/multiply */
269 if (mode == 2 && bitcpy > 0) {
270 /* square then multiply if the bit is set */
271 for (x = 0; x < bitcpy; x++) {
272 if ((err = mp_sqr (&res, &res)) != MP_OKAY) {
273 goto LBL_RES;
274 }
275 if ((err = redux (&res, P, mp)) != MP_OKAY) {
276 goto LBL_RES;
277 }
278
279 /* get next bit of the window */
280 bitbuf <<= 1;
281 if ((bitbuf & (1 << winsize)) != 0) {
282 /* then multiply */
283 if ((err = mp_mul (&res, &M[1], &res)) != MP_OKAY) {
284 goto LBL_RES;
285 }
286 if ((err = redux (&res, P, mp)) != MP_OKAY) {
287 goto LBL_RES;
288 }
289 }
290 }
291 }
292
293 if (redmode == 0) {
294 /* fixup result if Montgomery reduction is used
295 * recall that any value in a Montgomery system is
296 * actually multiplied by R mod n. So we have
297 * to reduce one more time to cancel out the factor
298 * of R.
299 */
300 if ((err = redux(&res, P, mp)) != MP_OKAY) {
301 goto LBL_RES;
302 }
303 }
304
305 /* swap res with Y */
306 mp_exch (&res, Y);
307 err = MP_OKAY;
308 LBL_RES:mp_clear (&res);
309 LBL_M:
310 mp_clear(&M[1]);
311 for (x = 1<<(winsize-1); x < (1 << winsize); x++) {
312 mp_clear (&M[x]);
313 }
314 return err;
315 }
316 #endif
317